1 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
5 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
7 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
8 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
9 (vector_alignment_reachable_p, vect_get_data_access_cost)
10 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
11 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
12 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
13 (vect_analyze_group_access_1, vect_analyze_group_access)
14 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
15 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
16 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
17 result of vect_dr_stmt and use the stmt_vec_info instead of
18 the associated gimple stmt.
19 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
20 (vect_gen_prolog_loop_niters): Likewise.
21 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
23 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
25 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
26 to a vec<stmt_vec_info>.
27 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
28 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
30 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
31 to a vec<stmt_vec_info>.
32 (bst_traits::value_type, bst_traits::value_type): Likewise.
33 (bst_traits::hash): Update accordingly.
34 (vect_get_and_check_slp_defs): Change the stmts parameter from
35 a vec<gimple *> to a vec<stmt_vec_info>.
36 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
37 (vect_build_slp_tree): Likewise.
38 (vect_build_slp_tree_2): Likewise. Update uses of
39 SLP_TREE_SCALAR_STMTS.
40 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
41 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
42 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
43 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
44 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
45 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
46 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
47 (vect_get_constant_vectors, vect_get_slp_defs)
48 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
49 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
50 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
51 instead of gimple stmts.
52 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
53 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
54 (vect_slp_analyze_instance_dependence): Update uses of
55 SLP_TREE_SCALAR_STMTS.
56 (vect_slp_analyze_and_verify_node_alignment): Likewise.
57 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
58 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
59 (get_initial_defs_for_reduction): Likewise.
60 (vect_create_epilog_for_reduction): Likewise.
61 (vectorize_fold_left_reduction): Likewise.
62 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
63 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
64 (can_vectorize_live_stmts): Likewise.
66 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
68 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
69 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
70 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
72 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
74 (gather_scalar_reductions): Update after above interface changes.
75 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
76 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
78 (vect_force_simple_reduction): Likewise.
79 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
80 LOOP_VINFO_REDUCTIONS.
81 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
83 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
85 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
86 a gimple stmt to a stmt_vec_info.
87 * tree-vect-loop.c (vect_active_double_reduction_p)
88 (vect_force_simple_reduction, vectorizable_reduction): Update
91 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
93 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
94 vec<gimple *> to a vec<stmt_vec_info>.
95 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
96 the reduction_phis argument from a vec<gimple *> to a
98 (vectorizable_reduction): Likewise the phis local variable that
99 is passed to vect_create_epilog_for_reduction. Update for new type
100 of SLP_TREE_VEC_STMTS.
101 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
102 (vectorizable_live_operation): Likewise.
103 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
104 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
106 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
108 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
109 a gimple stmt to a stmt_vec_info.
110 (vectorizable_condition, vectorizable_live_operation)
111 (vectorizable_reduction, vectorizable_induction): Pass back the
112 vectorized statement as a stmt_vec_info.
113 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
114 use of STMT_VINFO_VEC_STMT.
115 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
116 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
117 as stmt_vec_infos rather than gimple stmts.
118 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
120 (vectorizable_live_operation): Likewise.
121 (vectorizable_reduction, vectorizable_induction): Likewise,
122 updating use of STMT_VINFO_VEC_STMT.
123 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
124 of STMT_VINFO_VEC_STMT.
125 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
126 (vectorizable_simd_clone_call, vectorizable_conversion)
127 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
128 (vectorizable_store, vectorizable_load, vectorizable_condition)
129 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
130 from a gimple stmt to a stmt_vec_info.
131 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
132 pointer to a stmt_vec_info to the vectorizable_* routines.
134 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
136 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
137 a gimple stmt to a stmt_vec_info.
138 (is_pattern_stmt_p): Update accordingly.
139 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
140 (vect_record_grouped_load_vectors): Likewise.
141 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
142 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
143 (vect_model_reduction_cost): Likewise.
144 (vect_create_epilog_for_reduction): Likewise.
145 (vectorizable_reduction, vectorizable_induction): Likewise.
146 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
147 Return the stmt_vec_info for the pattern statement.
148 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
149 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
150 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
151 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
152 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
153 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
154 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
155 (free_stmt_vec_info, vect_is_simple_use): Likewise.
157 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
159 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
160 (vect_finish_stmt_generation): Likewise.
161 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
162 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
163 (vect_build_gather_load_calls): Use the return value of the above
164 functions instead of a separate call to vinfo_for_stmt. Use narrow
165 scopes for the input gimple stmt and wider scopes for the associated
166 stmt_vec_info. Use vec_info::lookup_def when setting these
167 stmt_vec_infos from an SSA_NAME definition.
168 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
169 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
170 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
171 (vectorizable_store, vectorizable_load, vectorizable_condition)
172 (vectorizable_comparison): Likewise.
173 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
174 (vectorizable_reduction): Likewise.
176 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
178 * tree-vectorizer.h (vect_is_simple_use): Add an optional
179 stmt_vec_info * parameter before the optional gimple **.
180 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
181 (process_use, vect_get_vec_def_for_operand_1): Update callers.
182 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
183 * tree-vect-loop.c (vectorizable_reduction): Likewise.
184 (vectorizable_live_operation): Likewise.
185 * tree-vect-patterns.c (type_conversion_p): Likewise.
186 (vect_look_through_possible_promotion): Likewise.
187 (vect_recog_rotate_pattern): Likewise.
188 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
190 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
192 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
193 a typedef to a wrapper class.
194 (NULL_STMT_VEC_INFO): New macro.
195 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
196 (stmt_vec_info::operator*): New function.
197 (stmt_vec_info::operator gimple *): Likewise.
198 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
199 (add_stmt_costs): Likewise.
200 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
201 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
202 (vect_get_known_peeling_cost): Likewise.
203 (vect_estimate_min_profitable_iters): Likewise.
204 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
205 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
206 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
207 (vectorizable_store, free_stmt_vec_infos): Likewise.
208 (new_stmt_vec_info): Change return type of xcalloc to
211 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
213 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
214 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
215 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
216 a single_imm_use-based sequence.
217 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
219 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
221 * tree-vectorizer.h (vec_info::lookup_def): Declare.
222 * tree-vectorizer.c (vec_info::lookup_def): New function.
223 * tree-vect-patterns.c (vect_get_internal_def): Use it.
224 (vect_widened_op_tree): Likewise.
225 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
226 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
227 (vectorizable_reduction): Likewise.
228 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
230 (vect_is_slp_reduction): Update calls accordingly. Use
231 vec_info::lookup_def.
232 (vect_is_simple_reduction): Likewise
233 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
235 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
237 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
238 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
239 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
241 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
242 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
243 (vect_update_vf_for_slp, vect_analyze_loop_operations)
244 (vect_is_slp_reduction, vectorizable_induction)
245 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
246 * tree-vect-patterns.c (vect_init_pattern_stmt):
247 (vect_determine_min_output_precision_1, vect_determine_precisions)
248 (vect_pattern_recog): Likewise.
249 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
250 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
251 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
252 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
253 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
254 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
255 info field from a loop to a loop_vec_info.
257 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
259 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
260 (vec_info::add_stmt): Declare.
261 * tree-vectorizer.c (vec_info::add_stmt): New function.
262 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
263 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
264 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
265 (vectorizable_induction): Likewise.
266 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
267 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
268 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
269 (vectorizable_load): Likewise.
270 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
271 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
272 (vect_recog_gather_scatter_pattern): Likewise.
273 (append_pattern_def_seq): Likewise. Remove a check that is
274 performed by add_stmt itself.
276 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
278 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
279 which make_ssa_name was called with new_stmt before new_stmt
282 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
284 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
286 (vect_is_slp_reduction): ...here...
287 (vect_is_simple_reduction): ...and here. Remove repetition of tests
288 that are already known to be false.
290 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
292 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
293 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
294 STMT_VINFO_NUM_SLP_USES when it's true.
295 (vect_free_slp_instance): Add a final_p parameter and pass it to
297 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
298 (vect_analyze_slp_instance): Likewise.
299 (vect_slp_analyze_operations): Likewise.
300 (vect_slp_analyze_bb_1): Likewise.
301 * tree-vectorizer.c (vec_info): Likewise.
302 * tree-vect-loop.c (vect_transform_loop): Likewise.
304 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
306 * tree-vect-loop.c (vectorizable_reduction): Assert that the
307 function is not called for second and subsequent members of
310 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
312 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
313 cases for nested loops from here to ...
314 (vect_create_epilog_for_reduction): ...here. Only call
315 vect_is_simple_use for inner-loop reductions.
317 2018-07-31 Martin Liska <mliska@suse.cz>
319 PR gcov-profile/85338
320 PR gcov-profile/85350
321 PR gcov-profile/85372
322 * profile.c (struct location_triplet): New.
323 (struct location_triplet_hash): Likewise.
324 (output_location): Do not output a BB that
325 is already recorded for a line.
326 (branch_prob): Use streamed_locations.
328 2018-07-31 Martin Liska <mliska@suse.cz>
330 PR gcov-profile/85370
331 * coverage.c (coverage_begin_function): Do not mark target
332 clones as artificial functions.
334 2018-07-31 Martin Liska <mliska@suse.cz>
336 PR gcov-profile/83813
337 PR gcov-profile/84758
338 PR gcov-profile/85217
339 PR gcov-profile/85332
340 * profile.c (branch_prob): Do not record GOTO expressions
341 for GIMPLE statements which locations are already streamed.
343 2018-07-31 Olivier Hainque <hainque@adacore.com>
345 * gcc.c (handle_spec_function): Accept a soft_matched_part
346 argument, as do_spec_1. Pass it down to ...
347 (eval_spec_function): Accept a soft_matched_part argument,
348 and pass it down to ...
349 (do_spec_2): Accept a soft_matched_part argument, and pass
350 it down to do_spec_1.
351 (do_spec_1): Pass soft_matched_part to handle_spec_function.
352 (handle_braces): Update call to handle_spec_function.
353 (driver::set_up_specs): Update calls to do_spec_2.
354 (compare_debug_dump_opt_spec_function): Likewise.
355 (compare_debug_self_opt_spec_function): Likewise.
357 2018-07-31 Olivier Hainque <hainque@adacore.com>
359 * common.opt (nolibc): New option.
360 * doc/invoke.texi (Link Options): Document it.
361 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
362 * config/alpha/linux.h: Likewise.
363 * config/arc/elf.h: Likewise.
364 * config/arm/uclinux-elf.h: Likewise.
365 * config/arm/unknown-elf.h: Likewise.
366 * config/avr/avrlibc.h: Likewise.
367 * config/bfin/bfin.h: Likewise.
368 * config/bfin/linux.h: Likewise.
369 * config/bfin/uclinux.h: Likewise.
370 * config/darwin.h: Likewise.
371 * config/darwin10.h: Likewise.
372 * config/darwin12.h: Likewise.
373 * config/gnu-user.h: Likewise.
374 * config/lm32/uclinux-elf.h: Likewise.
375 * config/pa/pa-hpux11.h: Likewise.
376 * config/pa/pa64-hpux.h: Likewise.
377 * config/sparc/sparc.h: Likewise.
379 2018-07-31 Olivier Hainque <hainque@adacore.com>
381 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
384 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
387 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
390 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
392 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
393 terminated string literal.
395 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
397 PR rtl-optimization/85160
398 * combine.c (is_just_move): New function.
399 (try_combine): Allow combining two instructions into two if neither of
400 the original instructions was a move.
402 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
405 * doc/extend.texi (Global Register Variables): Discourage use of type
407 (Local Register Variables): Likewise.
409 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
411 PR tree-optimization/86506
412 * hwint.h (ceil_log2): Resync with hwint.c implementation.
414 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
417 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
418 hard_regno, make sure no insn between `from` and `to` clobbers it.
420 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
421 Tom de Vries <tdevries@suse.de>
423 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
424 (PTX_DEFAULT_RUNTIME_DIM): ... this.
425 (nvptx_goacc_validate_dims): Set default worker and gang dims to
426 PTX_DEFAULT_RUNTIME_DIM.
427 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
429 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
431 * config/pa/pa.c (pa_output_addr_vec): Align address table.
432 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
433 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
435 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
437 * config/rs6000/constraints.md (wG constraint): Delete, no longer
439 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
440 predicate to reflect toc fusion has been deleted.
441 (toc_fusion_mem_raw): Delete, no longer used.
442 (toc_fusion_mem_wrapped): Likewise.
443 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
445 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
446 Delete, no longer used.
447 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
448 meant to be used for toc fusion.
449 (rs6000_debug_print_mode): Delete toc fusion debugging.
450 (rs6000_debug_reg_global): Likewise.
451 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
452 fusion and secondary reload support that were never used.
453 (rs6000_option_override_internal): Delete TOC fusion, that was only
454 partially defined, and it did not work unless you also used the
456 (rs6000_legitimate_address_p): Delete TOC fusion support.
457 (rs6000_opt_masks): Likewise.
458 (fusion_wrap_memory_address): Delete function, no longer used.
459 (fusion_split_address); Delete TOC fusion support.
460 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
461 longer used with toc fusion being deleted.
462 (TARGET_TOC_FUSION_FP): Likewise.
463 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
465 (toc fusion spliter): Delete TOC fusion support.
466 (toc_fusionload_<mode>): Likewise.
467 (toc_fusionload_di): Likewise.
468 (fusion_gpr_load_<mode>): Delete generator function, this insn no
469 longer needs to be named. Rename predicate to delete TOC fusion.
470 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
471 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
472 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
473 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
474 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
476 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
478 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
479 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
480 __int128 in built-in function prototypes.
481 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
482 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
484 2018-07-27 Martin Sebor <msebor@redhat.com>
486 PR tree-optimization/86696
487 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
488 types, including enums.
489 (handle_char_store): Be prepared for the above function to fail.
491 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
493 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
494 when optimization level is lower than 2 or optimize for size.
496 2018-07-26 Martin Sebor <msebor@redhat.com>
498 PR tree-optimization/86043
499 PR tree-optimization/86042
500 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
501 (get_string_cst_length): Rename...
502 (get_min_string_length): ...to this. Add argument.
503 (handle_char_store): Extend to handle multi-character stores by
505 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
506 * tree.h (initializer_zerop): Add argument.
508 2018-07-26 Jakub Jelinek <jakub@redhat.com>
511 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
512 declare target to variables if they have always,{to,from,tofrom} map
515 2018-07-26 Martin Liska <mliska@suse.cz>
518 * lto-wrapper.c: Add linker_output as prefix
519 for ltrans_output_file.
521 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
523 PR rtl-optimization/85805
524 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
525 value for hard registers if that was written in the same mode.
527 2018-07-26 Martin Liska <mliska@suse.cz>
529 PR gcov-profile/86536
530 * gcov.c (format_gcov): Use printf format %.*f directly
531 and do not handle special values.
533 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
535 * common/config/arc/arc-common.c (arc_option_optimization_table):
536 Update default optimizations for size.
538 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
540 * config/arc/arc.md (movsf_insn): Add short instruction selection.
541 * config/arc/constraints.md (CfZ): New constraint.
542 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
543 (subsf3_fpu): Likewise.
544 (cmpsf_fpu): Likewise.
545 (cmpsf_fpu_uneq): Likewise.
547 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
549 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
551 (arc_is_uncached_mem_p): uncached applies to both the variable and
554 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
556 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
559 2018-07-25 David Malcolm <dmalcolm@redhat.com>
561 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
562 field "m_scopes" from vec to auto_vec.
564 2018-07-25 Martin Liska <mliska@suse.cz>
566 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
569 2018-07-25 Richard Biener <rguenther@suse.de>
572 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
573 special wrt context_die late.
574 (gen_subprogram_die): Re-use DIEs in local scope.
576 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
578 PR tree-optimization/86644
579 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
581 2018-07-25 Martin Liska <mliska@suse.cz>
584 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
585 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
587 2018-07-25 Martin Liska <mliska@suse.cz>
590 * params.def: Explain ASan abbreviation and provide
591 a documentation link.
593 2018-07-24 Martin Sebor <msebor@redhat.com>
595 PR tree-optimization/86622
596 PR tree-optimization/86532
597 * builtins.h (string_length): Declare.
598 * builtins.c (c_strlen): Correct handling of non-constant offsets.
599 (check_access): Be prepared for non-constant length ranges.
600 (string_length): Make extern.
601 * expr.c (string_constant): Only handle the minor non-constant
602 array index. Use string_constant to compute the length of
603 a generic string constant.
605 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
607 PR tree-optimization/86618
608 * tree-vect-stmts.c (vectorizable_call): Don't take the address
609 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
611 2018-07-24 David Malcolm <dmalcolm@redhat.com>
613 PR tree-optimization/86636
614 * json.cc (json::object::set): Fix comment. Add assertions.
615 (json::array::append): Move here from json.h. Add comment and an
617 (json::string::string): Likewise.
618 * json.h (json::array::append): Move to json.cc.
619 (json::string::string): Likewise.
620 * optinfo-emit-json.cc
621 (optrecord_json_writer::impl_location_to_json): Assert that we
622 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
623 wrapper around it. Expand the location once, rather than three
625 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
626 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
628 (optrecord_json_writer::optinfo_to_json): Likewise, in four
629 places. Fix some overlong lines.
631 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
633 * config/aarch64/aarch64-simd.md
634 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
635 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
636 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
637 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
638 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
639 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
640 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
641 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
642 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
644 2018-07-24 Jakub Jelinek <jakub@redhat.com>
647 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
648 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
649 and abs_d == d, do the power of two handling if profitable.
651 2018-07-24 Richard Biener <rguenther@suse.de>
653 * match.pd: Add BIT_FIELD_REF canonicalizations.
655 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
658 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
660 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
662 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
663 terminated STRING_CST object.
665 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
667 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
669 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
671 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
672 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
673 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
674 the elements into a register.
675 (rs6000_split_v4si_init_di_reg): Delete.
676 (rs6000_split_v4si_init): Delete.
677 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
678 (vsx_init_v4si): Rewrite as a define_expand.
680 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
682 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
683 zero_extend argument from memory): New.
685 2018-07-22 Martin Sebor <msebor@redhat.com>
688 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
689 diagnosing calls with unknown arguments unless -Walloca-larger-than
690 is restricted to less than PTRDIFF_MAX bytes.
692 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
694 * doc/gcov.texi (Invoking Gcov): Editorial changes.
696 2018-07-20 David Malcolm <dmalcolm@redhat.com>
698 * pretty-print.c (text_info::set_location): Remove redundant
699 "line_table" parameter from call to rich_location::set_range.
701 2018-07-20 Martin Sebor <msebor@redhat.com>
704 * builtins.c (expand_builtin_alloca): Adjust.
705 * calls.c (alloc_max_size): Simplify.
706 * cgraphunit.c (cgraph_node::expand): Adjust.
707 * common.opt (larger_than_size, warn_frame_larger_than): Remove
709 (frame_larger_than_size): Same.
710 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
711 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
713 * doc/invoke.texi (GCC Command Options): Document option arguments.
714 Explain byte-size arguments and suffixes.
715 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
716 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
717 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
718 * doc/options.texi (UInteger): Expand.
719 (Host_Wide_Int, ByteSize): Document new properties.
720 * final.c (final_start_function_1): Include sizes in an error message.
721 * function.c (frame_offset_overflow): Same.
722 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
723 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
724 Diagnose unbounded alloca calls only for limits of less than
726 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
727 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
729 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
730 only for limits of less than PTRDIFF_MAX.
731 * langhooks-def.h (lhd_handle_option): Change function argument
733 * langhooks.c (lhd_handle_option): Same.
734 * langhooks.h (handle_option): Same.
735 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
737 (var_type, var_type_struct): Same.
738 (var_set): Handle ByteSize flag.
739 * optc-gen.awk: Add comments to output to ease debugging. Make
740 use of HOST_WIDE_INT where appropriate.
741 * opts-gen-save.awk: Use %lx to format unsigned long.
742 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
743 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
744 arguments. Parse bytes-size suffixes.
745 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
746 (enum_value_to_arg): Same.
747 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
748 (handle_option): Adjust.
749 (generate_option): Change function argument to HOST_WIDE_INT.
750 (cmdline_handle_error): Adjust.
751 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
752 (set_option): Change function argument to HOST_WIDE_INT.
753 (option_enabled): Handle cl_host_wide_int.
754 (get_option_state): Handle CLVC_SIZE.
755 (control_warning_option): Same.
756 * opts.c (common_handle_option): Change function argument to
757 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
758 OPT_Wvla_larger_than_.
759 * opts.h (enum cl_var_type): Add an enumerator.
760 * stor-layout.c (layout_decl): Print a more meaningful warning.
761 * toplev.c (output_stack_usage): Adjust.
763 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
765 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
766 call to inline_expand_builtin_string_cmp.
767 (expand_builtin_strcmp): Likewise.
768 (expand_builtin_strncmp): Likewise.
769 (inline_string_cmp): Delete the last parameter, change char_type_node
770 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
772 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
773 the inlining expansion on target where the type of the call has same or
774 narrower precision than unsigned char.
776 2018-07-20 David Malcolm <dmalcolm@redhat.com>
778 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
779 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
780 * common.opt (fsave-optimization-record): New option.
781 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
782 * doc/invoke.texi (-fsave-optimization-record): New option.
783 * dumpfile.c: Include "optinfo-emit-json.h".
784 (struct kv_pair): Move to coretypes.h.
785 (optgroup_options): Make non-static.
786 (dump_context::end_scope): Call
787 optimization_records_maybe_pop_dump_scope.
788 * dumpfile.h (optgroup_options): New decl.
791 * optinfo-emit-json.cc: New file.
792 * optinfo-emit-json.h: New file.
793 * optinfo.cc: Include "optinfo-emit-json.h".
794 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
795 (optinfo_enabled_p): Check optimization_records_enabled_p.
796 (optinfo_wants_inlining_info_p): Likewise.
797 * optinfo.h: Update comment.
798 * profile-count.c (profile_quality_as_string): New function.
799 * profile-count.h (profile_quality_as_string): New decl.
800 (profile_count::quality): New accessor.
801 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
802 and optinfo_emit_json_cc_tests.
803 * selftest.h (selftest::json_cc_tests): New decl.
804 (selftest::optinfo_emit_json_cc_tests): New decl.
805 * toplev.c: Include "optinfo-emit-json.h".
806 (compile_file): Call optimization_records_finish.
807 (do_compile): Call optimization_records_start.
808 * tree-ssa-live.c: Include optinfo.h.
809 (remove_unused_scope_block_p): Retain inlining information if
810 optinfo_wants_inlining_info_p returns true.
812 2018-07-20 Richard Biener <rguenther@suse.de>
815 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
816 to cover -flto-partition=none.
818 2018-07-20 Martin Liska <mliska@suse.cz>
820 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
821 (get_decl_source_range): Remove unused function.
823 2018-07-20 Richard Biener <rguenther@suse.de>
825 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
826 (struct vn_phi_s): Likewise.
827 (struct vn_reference_s): Likewise.
828 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
829 for searching the slot of an entry known to be in the hash itself.
830 (vn_phi_hasher::equal): Likewise.
831 (vn_reference_hasher::equal): Likewise.
832 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
834 (optimistic_info, current_info): Remove, keeping only valid_info.
835 (vn_reference_lookup_1): Remove fallback lookup.
836 (vn_reference_lookup_2): Likewise.
837 (vn_nary_op_lookup_1): Likewise.
838 (vn_phi_lookup): Likewise.
839 (vn_nary_build_or_lookup_1): Make sure to not chain the built
841 (vn_reference_insert): Adjust, chain the inserted hash element
842 at last_inserted_ref.
843 (vn_reference_insert_pieces): Likewise.
844 (visit_reference_op_call): Likewise.
845 (vn_nary_op_insert_into): Chain the inserted hash element at
847 (vn_nary_op_insert_pieces): Adjust.
848 (vn_nary_op_insert): Likewise.
849 (vn_nary_op_insert_stmt): Likewise.
850 (vn_phi_insert): Adjust, chain the inserted hash element at
852 (process_scc): Remove clearing and copying the optimistic
853 table. Instead remove elements inserted during an optimistic
854 iteration from the single table we maintain.
855 (init_scc_vn): Adjust.
856 (free_scc_vn): Likewise.
857 (sccvn_dom_walker::record_cond): Likewise.
858 (sccvn_dom_walker::after_dom_children): Likewise.
860 2018-07-19 Martin Sebor <msebor@redhat.com>
862 PR tree-optimization/84047
863 PR tree-optimization/83776
864 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
865 (check_array_bounds): Call it.
867 2018-07-19 Martin Sebor <msebor@redhat.com>
869 * align.h (align_flags): Use member initialization.
871 2018-07-19 David Malcolm <dmalcolm@redhat.com>
873 * Makefile.in (OBJS): Add optinfo.o.
874 * coretypes.h (class symtab_node): New forward decl.
875 (struct cgraph_node): New forward decl.
876 (class varpool_node): New forward decl.
877 * dump-context.h: New file.
878 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
880 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
881 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
882 (set_alt_dump_file): Likewise.
883 (dump_context::~dump_context): New dtor.
884 (dump_gimple_stmt): Move implementation to...
885 (dump_context::dump_gimple_stmt): ...this new member function.
886 Add the stmt to any pending optinfo, creating one if need be.
887 (dump_gimple_stmt_loc): Move implementation to...
888 (dump_context::dump_gimple_stmt_loc): ...this new member function.
889 Start a new optinfo and add the stmt to it.
890 (dump_gimple_expr): Move implementation to...
891 (dump_context::dump_gimple_expr): ...this new member function.
892 Add the stmt to any pending optinfo, creating one if need be.
893 (dump_gimple_expr_loc): Move implementation to...
894 (dump_context::dump_gimple_expr_loc): ...this new member function.
895 Start a new optinfo and add the stmt to it.
896 (dump_generic_expr): Move implementation to...
897 (dump_context::dump_generic_expr): ...this new member function.
898 Add the tree to any pending optinfo, creating one if need be.
899 (dump_generic_expr_loc): Move implementation to...
900 (dump_context::dump_generic_expr_loc): ...this new member
901 function. Add the tree to any pending optinfo, creating one if
903 (dump_printf): Move implementation to...
904 (dump_context::dump_printf_va): ...this new member function. Add
905 the text to any pending optinfo, creating one if need be.
906 (dump_printf_loc): Move implementation to...
907 (dump_context::dump_printf_loc_va): ...this new member function.
908 Start a new optinfo and add the stmt to it.
909 (dump_dec): Move implementation to...
910 (dump_context::dump_dec): ...this new member function. Add the
911 value to any pending optinfo, creating one if need be.
912 (dump_context::dump_symtab_node): New member function.
913 (dump_context::get_scope_depth): New member function.
914 (dump_context::begin_scope): New member function.
915 (dump_context::end_scope): New member function.
916 (dump_context::ensure_pending_optinfo): New member function.
917 (dump_context::begin_next_optinfo): New member function.
918 (dump_context::end_any_optinfo): New member function.
919 (dump_context::s_current): New global.
920 (dump_context::s_default): New global.
921 (dump_scope_depth): Delete global.
922 (dumpfile_ensure_any_optinfo_are_flushed): New function.
923 (dump_symtab_node): New function.
924 (get_dump_scope_depth): Reimplement in terms of dump_context.
925 (dump_begin_scope): Likewise.
926 (dump_end_scope): Likewise.
927 (selftest::temp_dump_context::temp_dump_context): New ctor.
928 (selftest::temp_dump_context::~temp_dump_context): New dtor.
929 (selftest::verify_item): New function.
930 (ASSERT_IS_TEXT): New macro.
931 (ASSERT_IS_TREE): New macro.
932 (ASSERT_IS_GIMPLE): New macro.
933 (selftest::test_capture_of_dump_calls): New test.
934 (selftest::dumpfile_c_tests): Call it.
935 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
936 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
937 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
939 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
940 (dump_node, dump_bb): Move these unrelated decls.
941 (class dump_manager): Add leading comment.
942 * optinfo.cc: New file.
943 * optinfo.h: New file.
945 2018-07-19 Michael Collison <michael.collison@arm.com>
946 Richard Henderson <rth@redhat.com>
948 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
949 (subti): Handle op1 zero.
950 (subvti4, usub4ti4): New.
951 (*sub<GPI>3_compare1_imm): New.
952 (sub<GPI>3_carryinCV): New.
953 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
954 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
956 2018-07-19 Michael Collison <michael.collison@arm.com>
957 Richard Henderson <rth@redhat.com>
959 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
960 (addti3): Create simpler code if low part is already known to be 0.
961 (addvti4, uaddvti4): New.
962 (*add<GPI>3_compareC_cconly_imm): New.
963 (*add<GPI>3_compareC_cconly): New.
964 (*add<GPI>3_compareC_imm): New.
965 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
966 handle constants within this pattern..
967 (*add<GPI>3_compareV_cconly_imm): New.
968 (*add<GPI>3_compareV_cconly): New.
969 (*add<GPI>3_compareV_imm): New.
970 (add<GPI>3_compareV): New.
971 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
972 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
973 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
974 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
975 with 'comparison' operator.
976 (*add<GPI>3_compareV_cconly_imm): Ditto.
977 (*add<GPI>3_compareV_cconly): Ditto.
978 (*add<GPI>3_compareV_imm): Ditto.
979 (add<GPI>3_compareV): Ditto.
980 (add<mode>3_carryinC): Ditto.
981 (*add<mode>3_carryinC_zero): Ditto.
982 (*add<mode>3_carryinC): Ditto.
983 (add<mode>3_carryinV): Ditto.
984 (*add<mode>3_carryinV_zero): Ditto.
985 (*add<mode>3_carryinV): Ditto.
987 2018-07-19 Michael Collison <michael.collison@arm.com>
988 Richard Henderson <rth@redhat.com>
990 * config/aarch64/aarch64-modes.def (CC_V): New.
991 * config/aarch64/aarch64-protos.h
992 (aarch64_addti_scratch_regs): Declare
993 (aarch64_subvti_scratch_regs): Declare.
994 (aarch64_expand_subvti): Declare.
995 (aarch64_gen_unlikely_cbranch): Declare
996 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
997 for signed overflow using CC_Vmode.
998 (aarch64_get_condition_code_1): Handle CC_Vmode.
999 (aarch64_gen_unlikely_cbranch): New function.
1000 (aarch64_addti_scratch_regs): New function.
1001 (aarch64_subvti_scratch_regs): New function.
1002 (aarch64_expand_subvti): New function.
1004 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
1006 * config/aarch64/aarch64-option-extensions.def: New entry for profile
1008 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
1009 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
1012 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
1015 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1016 address check not strict.
1018 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
1020 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
1022 (store_pair_lanes<mode>): Likewise.
1023 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
1024 enum value 'ADDR_QUERY_LDP_STP_N'.
1025 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
1026 (aarch64_print_address_internal): Add declaration.
1027 (aarch64_print_ldpstp_address): Remove.
1028 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
1029 (aarch64_print_operand): Change printing of 'y'.
1030 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
1031 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
1032 'true' rather than '1'.
1033 * config/aarch64/constraints.md (Uml): Likewise.
1034 (Uml): Rename to Umn.
1037 2018-07-19 Richard Biener <rguenther@suse.de>
1039 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
1041 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
1042 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
1043 (vn_reference_hasher): Likewise.
1044 (struct vn_tables_s): Remove obstack and alloc-pool members.
1045 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
1046 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
1047 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
1048 (vn_reference_insert_pieces): Likewise.
1049 (alloc_vn_nary_op_noinit): Adjust.
1050 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
1051 (vn_phi_eq): Adjust.
1052 (shared_lookup_phiargs): Remove.
1053 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
1054 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
1055 (visit_reference_op_call): Likewise.
1056 (copy_nary, copy_phi, copy_reference): Remove.
1057 (process_scc): Rewind the obstack when iterating. Do not
1058 copy the elements to valid_info but just move them from one
1059 hashtable to the other.
1060 (allocate_vn_table): Adjust.
1061 (free_vn_table): Likewise.
1062 (init_scc_vn): Likewise.
1063 (free_scc_vn): Likewise.
1065 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
1068 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
1069 indirect_return as function type attribute.
1070 (ix86_attribute_table): Change indirect_return to function
1072 * doc/extend.texi: Update indirect_return attribute.
1074 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
1076 * wide-int.h (widest2_int): New.
1077 * gimple-fold.c (arith_overflowed_p): Use it.
1078 * tree.h (widest2_int_cst): New.
1079 * tree-vrp.c (wide_int_binop_overflow): Rename from
1080 vrp_int_const_binop.
1081 Rewrite to work on trees.
1082 (extract_range_from_multiplicative_op_1): Abstract code to...
1083 (wide_int_range_min_max): ...here.
1084 (wide_int_range_cross_product): ...and here.
1085 (extract_range_from_binary_expr_1): Abstract overflow code to...
1086 (wide_int_range_mult_wrapping): ...here.
1087 * tree-vrp.h (wide_int_range_cross_product): New.
1088 (wide_int_range_mult_wrapping): New.
1090 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
1091 Julia Koval <julia.koval@intel.com>
1093 * config/i386/x86-tune-costs.h (skylake_memcpy,
1094 skylake_memset): Replace rep_prefix with unrolling for size 512.
1096 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
1099 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle comparision with EQ_EXPR
1102 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
1104 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
1105 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
1106 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
1107 previously known as "PowerPC AltiVec Built-in Functions". Move
1108 some material to new subsubsections "PowerPC AltiVec Built-in
1109 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
1111 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
1112 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
1113 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
1114 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
1116 2018-07-18 Richard Biener <rguenther@suse.de>
1118 PR tree-optimization/86557
1119 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
1122 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
1124 * config/s390/s390.c (s390_function_profiler): Generate CFI.
1126 2018-07-17 Jeff Law <law@redhat.com>
1128 * config/arm/arm.c (get_label_padding): Update for recent
1129 changes to label_to_alignment.
1131 PR tree-optimization/86010
1132 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
1134 * config/mips/mips.c (vr4130_align_insns): Update for recent
1135 changes to label_to_alignment.
1137 * config/frv/frv.c (frv_label_align): Update for recent changes
1138 to label_to_alignment.
1140 * config/nios2/nios2.c (nios2_label_align): Update for recent
1141 changes which dropped ALIGN_LABELS_LOG.
1143 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
1145 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
1146 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
1148 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
1150 * config/arc/arc.c (arc_label_align): Use align_labels instead of
1151 deprecated align_labels_log.
1153 2018-07-17 Richard Biener <rguenther@suse.de>
1156 * dwarf2out.c (init_sections_and_labels): Always generate
1157 a debug_line_str_section for early LTO debug.
1158 (dwarf2out_finish): Reset debug_line_str_hash output early.
1159 Bump counter for extra dwarf5 .debug_loc labels to not conflict
1161 (dwarf2out_early_finish): Output debug_line_str.
1163 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
1165 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
1166 index register on z196 or later.
1168 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
1170 * config/s390/s390.c (s390_default_align): Set default function
1172 (s390_override_options_after_change): Call s390_default align.
1173 (s390_option_override_internal): Call s390_default_align.
1174 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
1176 2018-07-17 Jakub Jelinek <jakub@redhat.com>
1179 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
1180 to _looptemp_ clauses, other than the first two.
1182 2018-07-17 Martin Liska <mliska@suse.cz>
1184 * opts.c: Do not enable OPT_falign_* for -Os.
1186 2018-07-17 Martin Liska <mliska@suse.cz>
1188 * align.h (MAX_CODE_ALIGN): New.
1189 (MAX_CODE_ALIGN_VALUE): New.
1190 * common/config/i386/i386-common.c (ix86_handle_option):
1191 (MAX_CODE_ALIGN): Moved to align.h.
1192 * final.c (MAX_CODE_ALIGN): Likewise.
1193 * opts.c (parse_and_check_align_values):
1194 (MAX_CODE_ALIGN): Likewise.
1195 (MAX_CODE_ALIGN_VALUE): Likewise.
1197 2018-07-17 Martin Liska <mliska@suse.cz>
1199 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
1200 in order to fulfil coding style.
1201 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
1202 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1203 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
1204 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
1205 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
1206 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
1207 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
1208 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
1209 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1211 2018-07-17 Martin Liska <mliska@suse.cz>
1213 * align.h: New file.
1214 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions directly.
1215 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
1216 align_flags of label_to_alignment.
1217 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into align_flags
1219 * config/m68k/m68k.c: Do not use removed align_labels_value and
1221 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
1222 (LOOP_ALIGN): Likewise.
1223 (LABEL_ALIGN): Likewise.
1224 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
1225 Remove not used macro.
1226 (rs6000_loop_align): Change return type to align_flags.
1227 (rs6000_loop_align_max_skip): Remove.
1228 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
1229 Change return type to align_flags.
1230 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
1231 Remove not used macro.
1232 (rs6000_loop_align): Change return type to align_flags.
1233 (rs6000_loop_align_max_skip): Remove.
1234 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
1235 * config/rx/rx-protos.h (rx_align_for_label): Make it
1237 * config/rx/rx.c (rx_align_for_label): Change return type
1239 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
1241 into align_flags class.
1242 (LABEL_ALIGN): Likewise.
1243 (LOOP_ALIGN): Likewise.
1244 * config/s390/s390.c (s390_label_align): Use align_flags
1246 (s390_asm_output_function_label): Likewise.
1247 * config/sh/sh.c (sh_override_options_after_change):
1248 Use align_flags class directly without macros.
1249 (find_barrier): Likewise.
1250 (barrier_align): Likewise.
1251 (sh_loop_align): Likewise.
1252 * config/spu/spu.c (spu_option_override):
1253 Use align_flags_tuple::get_value instead of removed macros.
1254 (spu_sched_init): Likewise.
1255 * config/spu/spu.h (GTY): Likewise.
1256 * config/visium/visium.c (visium_option_override):
1257 Set "8" as default secondary alignment.
1258 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
1259 in order to guarantee secondary alignment of 8.
1260 * coretypes.h: Include align.h header file.
1261 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
1262 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
1263 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
1264 * doc/tm.texi.in: Likewise.
1265 * final.c (struct label_alignment): Remove not used structure.
1266 (LABEL_ALIGN): Change type to align_flags.
1267 (LOOP_ALIGN): Likewise.
1268 (JUMP_ALIGN): Likewise.
1269 (default_loop_align_max_skip): Remove.
1270 (default_label_align_max_skip): Likewise.
1271 (default_jump_align_max_skip): Likewise.
1272 (default_label_align_after_barrier_max_skip):
1273 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
1274 (LABEL_TO_MAX_SKIP): Remove.
1275 (label_to_alignment): Return align_flags type instead of integer.
1276 (label_to_max_skip): Remove.
1277 (align_fuzz): Use align_flags type.
1278 (compute_alignments): Use align_flags type and use align_flags::max
1279 to combine multiple alignments.
1280 (grow_label_align): Grow vec instead of C array.
1281 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
1282 (shorten_branches): Use align_flags type and use align_flags::max
1283 to combine multiple alignments.
1284 (final_scan_insn_1): Remove usage of secondary alignment that comes
1285 from label alignment, but instead use proper secondary alignment
1286 which is computed in grow_label_align.
1287 * flags.h (struct align_flags_tuple): Move to align.h.
1288 (struct align_flags): Likewise.
1289 (state_align_loops): Rename to align_loops.
1290 (state_align_jumps): Rename to align_jumps.
1291 (state_align_labels): Rename to align_labels.
1292 (state_align_functions): Rename to align_functions.
1293 (align_loops_log): Remove.
1294 (align_jumps_log): Remove.
1295 (align_labels_log): Remove.
1296 (align_functions_log): Remove.
1297 (align_loops_max_skip): Remove.
1298 (align_jumps_max_skip): Remove.
1299 (align_labels_max_skip): Remove.
1300 (align_functions_max_skip): Remove.
1301 (align_loops_value): Remove.
1302 (align_jumps_value): Remove.
1303 (align_labels_value): Remove.
1304 (align_functions_value): Remove.
1305 * output.h (label_to_alignment): Change return type to align_flags.
1306 (label_to_max_skip): Remove.
1307 * target.def: Remove loop_align_max_skip, label_align_max_skip,
1308 jump_align_max_skip macros.
1309 * targhooks.h (default_loop_align_max_skip): Remove.
1310 (default_label_align_max_skip): Likewise.
1311 (default_jump_align_max_skip): Likewise.
1312 (default_label_align_after_barrier_max_skip): Remove.
1313 * toplev.c (read_log_maxskip): Use ::normalize function.
1314 (parse_N_M): Remove not used argument and also call ::normalize.
1315 (parse_alignment_opts): Do not pass unused arguments.
1316 * varasm.c (assemble_start_function): Use directly align_functions
1317 instead of removed macros.
1318 * system.h: Do not poison removed macros.
1320 2018-07-17 Jakub Jelinek <jakub@redhat.com>
1323 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
1324 and cond temporaries don't have reference type if iterator has
1325 pointer type. For init use &for_pre_body instead of pre_p if
1326 for_pre_body is non-empty.
1328 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
1330 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
1331 double-double modes to SFmode directly directly.
1332 (trunc<mode>sf2_fprs): Delete.
1334 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
1336 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
1337 for conversions between IFmode and the decimal floating point modes.
1338 (init_float128_ieee): Use the correct names for conversions between
1339 KFmode and the decimal floating point modes.
1341 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
1343 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
1344 for the conversions between TDmode and IFmode.
1345 (init_float128_ieee): Use more correct names for the conversions
1346 between TDmode and KFmode.
1348 2018-07-16 Jakub Jelinek <jakub@redhat.com>
1350 PR tree-optimization/86526
1351 * builtins.c (expand_builtin_memcmp): Formatting fixes.
1352 (inline_expand_builtin_string_cmp): Likewise.
1353 (inline_string_cmp): Likewise. Use c_readstr instead of
1354 builtin_memcpy_read_str. Add unit_mode temporary.
1356 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
1359 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
1360 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
1362 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
1364 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
1365 Alphabetize prototypes of built-in functions, separating out
1366 built-in functions that are listed in this section but should be
1367 described elsewhere.
1369 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
1372 * expmed.c (emit_store_flag): Do not emit setcc followed by a
1373 conditional move when trapping comparison was split to a
1374 non-trapping one (and vice versa).
1376 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
1378 * config/s390/s390.c (s390_function_profiler): Generate nops
1379 instead of profiler call sequences.
1380 * config/s390/s390.opt: Add the new option.
1382 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
1384 * config/s390/s390.c (s390_function_profiler): Generate
1385 __mcount_loc section.
1386 * config/s390/s390.opt: Add the new option.
1388 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
1390 * common.opt: Add the new warning.
1391 * config/s390/s390.c (s390_function_profiler): Emit "brasl
1392 %r0,__fentry__" when -mfentry is specified.
1393 (s390_option_override_internal): Disallow -mfentry for 31-bit
1395 * config/s390/s390.opt: Add the new option.
1397 2018-07-16 Richard Biener <rguenther@suse.de>
1400 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
1401 for function-local FUNCTION_DECL and RESULT_DECL immediately.
1403 2018-07-16 Martin Liska <mliska@suse.cz>
1406 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
1409 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
1411 * config/arc/arcHS.md: Update ARCHS scheduling rules.
1413 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
1415 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
1417 * config/arc/arc-cpus.def (hs4x): New cpu.
1419 * config/arc/arc-tables.opt: Regenerate.
1420 * config/arc/arc.c (arc_sched_issue_rate): New function.
1421 (TARGET_SCHED_ISSUE_RATE): Define.
1422 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
1423 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
1425 (attr tune): Add ARCHS4x tune values.
1426 (attr tune_dspmpy): Define.
1427 (*tst): Correct instruction type.
1428 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
1429 * config/arc/arcHS4x.md: New file.
1430 * config/arc/fpu.md: Update instruction type attributes.
1431 * config/arc/t-multilib: Regenerate.
1433 2018-07-16 Tom de Vries <tdevries@suse.de>
1436 * var-tracking.c (vt_initialize): Fix pre_dec handling.
1438 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1440 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
1443 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
1445 PR tree-optimization/86514
1446 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
1447 conversion to a boolean type from a type with greater precision.
1449 2018-07-16 Tom de Vries <tdevries@suse.de>
1451 * var-tracking.c (vt_initialize): Print adjusted insn slim if
1452 dump_flags request TDF_SLIM.
1454 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
1456 * fold-const.c (int_const_binop_1): Abstract...
1457 (wide_int_binop): ...wide int code here.
1458 (poly_int_binop): ...poly int code here.
1459 Abstract the rest of int_const_binop_1 into int_const_binop.
1460 * fold-const.h (wide_int_binop): New.
1461 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
1462 Remove useless PLUS/MINUS_EXPR case.
1463 (zero_nonzero_bits_from_vr): Move wide int code...
1464 (zero_nonzero_bits_from_bounds): ...here.
1465 (extract_range_from_binary_expr_1): Move mask optimization code...
1466 (range_easy_mask_min_max): ...here.
1467 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
1468 (range_easy_mask_min_max): New.
1470 2018-07-15 Jeff Law <law@redhat.com>
1473 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
1476 2018-07-14 Jim Wilson <jimw@sifive.com>
1478 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
1480 2018-07-14 Paul Koning <ni1d@arrl.net>
1482 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
1484 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
1486 * lto-streamer-out.c (copy_function_or_variable): Dump info about
1489 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
1490 Steve Munroe <munroesj52@gmail.com>
1492 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
1493 (_mm_andnot_si128): Likewise.
1494 (_mm_or_si128): Likewise.
1495 (_mm_xor_si128): Likewise.
1497 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
1500 * builtins.c (expand_builtin_memcmp): Inline the calls first
1501 when result_eq is false.
1502 (expand_builtin_strcmp): Inline the calls first.
1503 (expand_builtin_strncmp): Likewise.
1504 (inline_string_cmp): New routine. Expand a string compare
1505 call by using a sequence of char comparison.
1506 (inline_expand_builtin_string_cmp): New routine. Inline expansion
1507 a call to str(n)cmp/memcmp.
1508 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
1510 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
1512 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
1514 * config/arm/driver-arm.c: Include arm-native.h.
1515 (host_detect_local_cpu): Use auto-generated data tables.
1516 (vendors, arm_cpu_table): Delete. Move part information to ...
1517 * config/arm/arm-cpus.in: ... here.
1518 * config/arm/parsecpu.awk (gen_native): New function.
1519 (vendor, part): New CPU fields.
1520 (END): Add support for building the native CPU detection tables.
1521 * config/arm/t-arm (arm-native.h): Add build rule.
1522 (driver-arm.o): Add dependency on arm-native.h.
1524 2018-07-13 Richard Biener <rguenther@suse.de>
1527 * match.pd (addr1 - addr2): Allow either of the operand to
1530 2018-07-13 Tom de Vries <tdevries@suse.de>
1532 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
1535 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
1537 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
1538 arrays instead of numbered variables.
1540 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
1542 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
1543 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
1545 2018-07-13 Richard Biener <rguenther@suse.de>
1548 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
1549 instead of get_context_die.
1551 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
1552 Richard Biener <rguenther@suse.de>
1555 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
1556 that the loop latch destination where phi is defined.
1558 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
1560 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
1561 (riscv_expand_epilogue): Add assertion to check interrupt mode.
1562 (riscv_set_current_function): Extract getting interrupt type to new
1564 (riscv_get_interrupt_type): New function.
1565 (riscv_merge_decl_attributes): New function, checking interrupt type is
1567 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
1569 2018-07-12 Paul Koning <ni1d@arrl.net>
1571 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
1574 2018-07-12 Paul Koning <ni1d@arrl.net>
1576 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
1577 zero reference, add doloop_end instead.
1578 * doc/md.texi (decrement_and_branch_until_zero): Remove.
1579 (Looping patterns): Remove decrement_and_branch_until_zero. Add
1580 detail for doloop_end.
1582 2018-07-12 Martin Sebor <msebor@redhat.com>
1585 * attribs.c (decl_attributes): Reject conflicting attributes before
1586 calling attribute handlers.
1588 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
1590 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
1592 (gcc::dump_manager::get_dump_file_name): likewise.
1593 (dump_begin): Likewise.
1594 * dumpfile.h (dump_begin): Update prototype.
1595 (gcc::dump_manager::get_dump_file_name,
1596 gcc::dump_manager::get_dump_file_name): Update prototype.
1598 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1600 * internal-fn.h (vectorizable_internal_fn_p): New function.
1601 * tree-vect-slp.c (compatible_calls_p): Likewise.
1602 (vect_build_slp_tree_1): Remove nops argument. Handle calls
1603 to internal functions.
1604 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
1606 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1608 * fold-const.h (inverse_conditions_p): Declare.
1609 * fold-const.c (inverse_conditions_p): New function.
1610 * match.pd: Use inverse_conditions_p. Add folds of view_converts
1611 that test the inverse condition of a conditional internal function.
1612 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
1613 * internal-fn.c (internal_fn_mask_index): Handle conditional
1615 (vectorized_internal_fn_supported_p): New function.
1616 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
1617 (any_pred_load_store): Replace with...
1618 (need_to_predicate): ...this new variable.
1619 (redundant_ssa_names): New variable.
1620 (ifcvt_can_use_mask_load_store): Move initial checks to...
1621 (ifcvt_can_predicate): ...this new function. Handle tree codes
1622 for which a conditional internal function exists.
1623 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
1624 instead of ifcvt_can_use_mask_load_store. Update after variable
1626 (predicate_load_or_store): New function, split out from
1627 predicate_mem_writes.
1628 (check_redundant_cond_expr): New function.
1629 (value_available_p): Likewise.
1630 (predicate_rhs_code): Likewise.
1631 (predicate_mem_writes): Rename to...
1632 (predicate_statements): ...this. Use predicate_load_or_store
1633 and predicate_rhs_code.
1634 (combine_blocks, tree_if_conversion): Update after above name changes.
1635 (ifcvt_local_dce): Handle redundant_ssa_names.
1636 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
1637 general conditional functions.
1638 * tree-vect-stmts.c (vectorizable_call): Likewise.
1640 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1641 Alan Hayward <alan.hayward@arm.com>
1642 David Sherwood <david.sherwood@arm.com>
1644 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
1645 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
1646 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
1647 plus and minus and convert them into IFN_COND_FMA-based sequences.
1648 (convert_mult_to_fma): Handle conditional plus and minus.
1650 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1652 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
1653 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
1654 (cond_fnms_optab): New optabs.
1655 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
1657 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
1658 * internal-fn.h (get_conditional_internal_fn): Declare.
1659 (get_unconditional_internal_fn): Likewise.
1660 * internal-fn.c (cond_ternary_direct): New macro.
1661 (expand_cond_ternary_optab_fn): Likewise.
1662 (direct_cond_ternary_optab_supported_p): Likewise.
1663 (FOR_EACH_COND_FN_PAIR): Likewise.
1664 (get_conditional_internal_fn): New function.
1665 (get_unconditional_internal_fn): Likewise.
1666 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
1667 (gimple_match_op::gimple_match_op): Add a new overload for 5
1669 (gimple_match_op::set_op): Likewise.
1670 (gimple_resimplify5): Declare.
1671 * genmatch.c (decision_tree::gen): Generate simplifications for
1673 * gimple-match-head.c (gimple_simplify): Define an overload for
1674 5 operands. Handle calls with 5 arguments in the top-level overload.
1675 (convert_conditional_op): Handle conversions from unconditional
1676 internal functions to conditional ones.
1677 (gimple_resimplify5): New function.
1678 (build_call_internal): Pass a fifth operand.
1679 (maybe_push_res_to_seq): Likewise.
1680 (try_conditional_simplification): Try converting conditional
1681 internal functions to unconditional internal functions.
1682 Handle 3-operand unconditional forms.
1683 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
1684 Define ternary equivalents of the current rules for binary conditional
1686 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
1688 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
1689 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
1690 (optab): Handle them.
1691 (SVE_COND_FP_TERNARY): New int iterator.
1692 (sve_fmla_op, sve_fmad_op): New int attributes.
1693 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
1694 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
1695 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
1697 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1699 * target.def (preferred_else_value): New target hook.
1700 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
1701 * doc/tm.texi: Regenerate.
1702 * targhooks.h (default_preferred_else_value): Declare.
1703 * targhooks.c (default_preferred_else_value): New function.
1704 * internal-fn.h (conditional_internal_fn_code): Declare.
1705 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
1706 (get_conditional_internal_fn): Use it.
1707 (conditional_internal_fn_code): New function.
1708 * gimple-match.h (gimple_match_cond): New struct.
1709 (gimple_match_op): Add a cond member function.
1710 (gimple_match_op::gimple_match_op): Update all forms to take a
1712 * genmatch.c (expr::gen_transform): Use the same condition as res_op
1713 for the suboperation, but don't specify a particular else_value.
1714 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
1715 (visit_nary_op, visit_reference_op_load): Pass
1716 gimple_match_cond::UNCOND to the gimple_match_op constructor.
1717 * gimple-match-head.c: Include tree-eh.h
1718 (convert_conditional_op): New function.
1719 (maybe_resimplify_conditional_op): Likewise.
1720 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
1721 (gimple_resimplify2): Likewise.
1722 (gimple_resimplify3): Likewise.
1723 (gimple_resimplify4): Likewise.
1724 (maybe_push_res_to_seq): Return null for conditional operations.
1725 (try_conditional_simplification): New function.
1726 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
1728 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
1730 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
1732 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
1734 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
1736 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
1738 (hash_tree): Do not hash DECL_FCONTEXT
1739 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
1740 Do not stream DECL_FCONTEXT.
1741 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
1742 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
1744 2018-07-12 Richard Biener <rguenther@suse.de>
1747 * dwarf2out.c (gen_block_die): Only output blocks when they have
1748 at least one !DECL_IGNORED_P variable.
1750 2018-07-12 Richard Biener <rguenther@suse.de>
1753 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
1754 Remove -mieee-fp handling.
1756 2018-07-12 Richard Biener <rguenther@suse.de>
1758 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
1759 left-over from last patch.
1761 2018-07-12 Jakub Jelinek <jakub@redhat.com>
1763 PR tree-optimization/86492
1764 * gimple-ssa-store-merging.c
1765 (imm_store_chain_info::coalesce_immediate_stores): Call
1766 check_no_overlap even for the merge_overlapping case. Formatting fix.
1768 2018-07-12 Richard Biener <rguenther@suse.de>
1771 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
1772 move possibly trapping operations into the conditional.
1774 2018-07-12 Richard Biener <rguenther@suse.de>
1776 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
1777 (vn_lookup_simplify_result): Remove recursion limit applied
1779 (vn_nary_build_or_lookup_1): Adjust.
1780 (try_to_simplify): Likewise.
1781 * gimple-match-head.c (gimple_resimplify1): Instead apply one
1783 (gimple_resimplify2): Likewise.
1784 (gimple_resimplify3): Likewise.
1785 (gimple_resimplify4): Likewise.
1787 2018-07-11 Jakub Jelinek <jakub@redhat.com>
1789 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
1790 Use __mmask64 type instead of __mmask8 for __M argument.
1791 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
1792 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
1794 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
1795 __mmask16 for __M argument.
1796 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
1797 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
1798 to __mmask16 instead of __mmask8.
1799 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
1800 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
1801 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
1802 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
1803 instead of __mmask16 for __U argument.
1804 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
1805 __mmask16 instead of __mmask8 for __U argument.
1806 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
1808 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
1810 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
1812 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
1814 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
1816 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
1818 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
1820 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
1822 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
1824 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
1826 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
1827 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
1828 return type as well as __M argument type and all casts from __mmask8
1830 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
1831 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
1832 return type as well as __M argument type and all casts from __mmask8
1834 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
1835 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
1836 return type as well as __M argument type and all casts from __mmask8
1838 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
1839 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
1840 return type as well as __M argument type and all casts from __mmask8
1842 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
1843 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
1846 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
1848 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
1849 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
1852 2018-07-11 Paul Koning <ni1d@arrl.net>
1854 * doc/md.texi (define_subst): Document how multiple occurrences of
1855 the same argument in the replacement pattern are handled.
1857 2018-07-11 Paul Koning <ni1d@arrl.net>
1859 * doc/extend.texi (Common Variable Attributes): Move "mode" into
1861 (Common Type Attributes): Add "mode" attribute.
1863 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
1865 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
1866 stream DECL_ORIGINAL_TYPE.
1867 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
1868 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
1869 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
1870 Do not walk original type.
1871 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
1873 (write_ts_decl_non_common_tree_pointers): Do not stream
1875 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
1876 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
1878 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
1880 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
1881 thread twice from the same starting edge.
1883 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
1885 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
1886 * gimple.c (gimple_call_nonnull_result_p): ...here...
1887 (gimple_call_nonnull_arg): ...and here.
1888 * gimple.h (gimple_call_nonnull_result_p): New.
1889 (gimple_call_nonnull_arg): New.
1891 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
1893 * config/arm/arm-cpus.in: Move information from fpu field of each
1894 cpu definition to the isa field.
1895 * config/arm/parsecpu.awk (fpu): Delete match rule.
1896 (gen_comm_data): Don't add bits from the CPU's FPU entry.
1898 2018-07-11 Richard Biener <rguenther@suse.de>
1901 * dwarf2out.c (init_sections_and_labels): Use
1902 output_asm_line_debug_info consistently.
1903 (dwarf2out_early_finish): Likewise.
1904 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
1907 2018-07-11 Richard Biener <rguenther@suse.de>
1909 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
1910 Rework father_bb setting in a way to avoid propagating constants
1911 multiple times on a loop body.
1913 2018-07-10 Mark Wielaard <mark@klomp.org>
1916 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
1918 2018-07-10 Richard Biener <rguenther@suse.de>
1920 * hash-map.h (hash_map::iterator::operator*): Return
1921 references to key and value.
1923 2018-07-10 Jakub Jelinek <jakub@redhat.com>
1926 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
1927 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
1928 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
1929 (gimplify_omp_for): For composite loops, move outer
1930 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
1931 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
1932 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
1933 TREE_LIST for both the original class iterator and the "last" helper
1934 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
1935 loop, remember has_decl_expr from outer composite loops for the
1936 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
1938 2018-07-09 Martin Sebor <msebor@redhat.com>
1942 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
1943 accessing implicitly initialized array elements.
1944 * expr.c (string_constant): Handle string initializers of
1945 character arrays within aggregates.
1946 * gimple-fold.c (fold_array_ctor_reference): Add argument.
1947 Store element offset. As a special case, handle zero size.
1948 (fold_nonarray_ctor_reference): Same.
1949 (fold_ctor_reference): Add argument. Store subobject offset.
1950 * gimple-fold.h (fold_ctor_reference): Add argument.
1952 2018-07-09 Paul Koning <ni1d@arrl.net>
1954 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
1955 (pdp11_insn_cost): New function.
1956 (pdp11_md_asm_adjust): New function.
1957 (TARGET_INVALID_WITHIN_DOLOOP): Define.
1958 (pdp11_rtx_costs): Update to match machine better.
1959 (output_addr_const_pdp11): Correct format mismatch warnings.
1960 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
1961 * config/pdp11/pdp11.md: General change to add base_cost and/or
1962 length attributes for use by new pdp11_insn_cost function.
1963 (MIN_BRANCH): Correct definition.
1965 (doloop_end): Use standard pattern name for looping pattern.
1966 (doloop_end_nocc): New.
1967 (movsf): Add another constraint alternative.
1968 (zero_extendqihi2): Add constraint alternatives for not in place
1970 (zero_extendhisi2): Remove.
1971 (shift patterns): Add CC handling variants.
1975 (define_peephole2): New peephole to recognize mov that sets CC for
1978 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1980 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
1981 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
1984 2018-07-09 Jakub Jelinek <jakub@redhat.com>
1987 * real.c (real_nextafter): Return true if result is denormal.
1989 2018-07-09 Martin Liska <mliska@suse.cz>
1991 * common.opt: Add back wrongly removed attribute.
1993 2018-07-09 Richard Biener <rguenther@suse.de>
1996 * dwarf2out.c (gen_block_die): For an early generated DIE
1997 always output high/low PC attributes.
1999 2018-07-09 Tom de Vries <tdevries@suse.de>
2001 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
2002 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
2003 onto VAR_DECL with abstract origin.
2005 2018-07-07 Jim Wilson <jimw@sifive.com>
2007 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
2009 2018-07-07 Tom de Vries <tdevries@suse.de>
2011 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
2013 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
2015 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
2017 (combine_bound): Use wide-int overflow calculation instead of
2019 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
2021 * fold-const.c (int_const_binop_2): Same.
2022 (extract_muldiv_1): Same.
2023 (fold_div_compare): Same.
2024 (fold_abs_const): Same.
2026 * poly-int.h (add): Same.
2030 * predict.c (predict_iv_comparison): Same.
2031 * profile-count.c (slow_safe_scale_64bit): Same.
2032 * simplify-rtx.c (simplify_const_binary_operation): Same.
2033 * tree-chrec.c (tree_fold_binomial): Same.
2034 * tree-data-ref.c (split_constant_offset_1): Same.
2035 * tree-if-conv.c (idx_within_array_bound): Same.
2036 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
2037 * tree-ssa-phiopt.c (minmax_replacement): Same.
2038 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
2039 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
2040 * vr-values.c (vr_values::adjust_range_with_scev): Same.
2041 * wide-int.cc (wi::add_large): Same.
2042 (wi::mul_internal): Same.
2043 (wi::sub_large): Same.
2044 (wi::divmod_internal): Same.
2045 * wide-int.h: Change overflow type to overflow_type for neg, add,
2046 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
2047 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
2048 mul_internal, divmod_internal.
2049 (overflow_type): New enum.
2050 (accumulate_overflow): New.
2052 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
2054 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
2055 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
2057 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
2059 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
2060 argument is checked for zero before entering loop, avoid checking again.
2062 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
2064 * gimplify.h (generic_expr_could_trap_p): Set as global function.
2065 * gimplify.h (generic_expr_could_trap_p): Likwise.
2066 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
2068 2018-07-06 Jakub Jelinek <jakub@redhat.com>
2070 PR tree-optimization/86401
2071 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
2072 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
2073 (fold_bit_and_mask): ... here. New helper function for match.pd.
2074 * fold-const.h (fold_bit_and_mask): Declare.
2075 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
2077 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
2080 * target.def (translate_mode_attribute): New hook.
2081 * targhooks.h (default_translate_mode_attribute): Declare.
2082 * targhooks.c (default_translate_mode_attribute): New function.
2083 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
2084 * doc/tm.texi: Regenerate.
2085 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
2086 (rs6000_translate_mode_attribute): New function.
2088 2018-07-06 Paul Koning <ni1d@arrl.net>
2090 * doc/md.texi (define_split): Document DONE and FAIL.
2091 (define_peephole2): Ditto.
2093 2018-07-05 Jeff Law <law@redhat.com>
2095 PR tree-optimization/86010
2096 * tree-ssa-dse.c (compute_trims): More aggressively trim at
2097 both the head and tail of mem* and str* calls.
2099 2018-07-05 Jim Wilson <jimw@sifive.com>
2101 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
2103 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
2105 * config/aarch64/aarch64-simd.md: correct flags text for
2106 MIN_EXPR replacement.
2108 2018-07-05 James Clarke <jrtc27@jrtc27.com>
2110 * configure: Regenerated.
2112 2018-07-05 Carl Love <cel@us.ibm.com>
2114 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
2115 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
2116 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
2117 VSX_BUILTIN_DOUBLEL_V4SF.
2119 2018-07-05 Martin Sebor <msebor@redhat.com>
2122 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
2123 than its domain to compute its the upper bound of a char array.
2125 2018-07-05 Nathan Sidwell <nathan@acm.org>
2127 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
2128 * doc/cpp.texi: Update comment.
2129 * doc/tm.texi: Rebuilt.
2130 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
2131 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
2132 * doc/extend.texi (Backwards Compatibility): Clarify it is system
2133 headers affected by extern "C".
2134 * system.h: Poison NO_IMPLICIT_EXTERN_C.
2135 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
2136 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
2137 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
2138 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
2139 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
2140 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
2141 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
2142 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
2143 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
2144 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
2145 config/sparc/sp64-elf.h, config/spu/spu.h,
2146 config/stormy16/stormy16.h, config/v850/v850.h,
2147 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
2148 define NO_IMPLICIT_EXTERN_C.
2149 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
2151 2018-07-05 Tamar Christina <tamar.christina@arm.com>
2154 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2155 instead of GET_MODE_SIZE when comparing Units.
2157 2018-07-05 Tamar Christina <tamar.christina@arm.com>
2160 * rtlanal.c (set_noop_p): Constrain on mode change,
2161 include hard-reg-set.h
2163 2018-07-05 Tamar Christina <tamar.christina@arm.com>
2165 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
2167 2018-07-05 Jakub Jelinek <jakub@redhat.com>
2170 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
2173 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
2175 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
2177 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
2180 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
2182 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
2184 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
2187 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
2189 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
2191 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
2192 neg_*_op* variables.
2194 2018-07-04 Martin Liska <mliska@suse.cz>
2196 * tree-switch-conversion.c: Define
2197 max_ratio_for_speed and max_ratio_for_size constants.
2199 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
2200 Martin Liska <mliska@suse.cz>
2205 * common.opt: Rename align options with 'str_' prefix.
2206 * common/config/i386/i386-common.c (set_malign_value): New
2208 (ix86_handle_option): Use it to set -falign-* options/
2209 * config/aarch64/aarch64-protos.h (struct tune_params): Change
2210 type from int to string.
2211 * config/aarch64/aarch64.c: Update default values from int
2213 * config/alpha/alpha.c (alpha_override_options_after_change):
2215 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
2216 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2217 max skip conditionally.
2218 * config/i386/freebsd.h (SUBALIGN_LOG): New.
2219 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2220 max skip conditionally.
2221 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2222 max skip conditionally.
2223 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
2224 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2225 max skip conditionally.
2226 * config/i386/i386.c (struct ptt): Change type from int to
2228 (ix86_default_align): Set default values.
2229 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
2230 max skip conditionally.
2231 * config/i386/iamcu.h (SUBALIGN_LOG): New.
2232 (ASM_OUTPUT_MAX_SKIP_ALIGN):
2233 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
2234 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2235 max skip conditionally.
2236 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
2237 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
2238 * config/i386/x86-64.h (SUBALIGN_LOG): New.
2239 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2240 max skip conditionally.
2241 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
2242 * config/ia64/ia64.c (ia64_option_override): Set default values
2243 for alignment options.
2244 * config/m68k/m68k.c: Handle new str_align_* options.
2245 * config/mips/mips.c (mips_set_compression_mode): Change
2247 (mips_option_override): Set default values for options.
2248 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
2250 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2252 * config/rx/rx.c (rx_option_override): Likewise.
2253 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
2254 (LABEL_ALIGN): Use align_labels_log.
2255 (LOOP_ALIGN): Use align_loops_align.
2256 * config/s390/s390.c (s390_asm_output_function_label): Use new
2258 * config/sh/sh.c (sh_override_options_after_change):
2259 Change type of constants.
2260 * config/spu/spu.c (spu_sched_init): Likewise.
2261 * config/sparc/sparc.c (sparc_option_override): Set default
2263 * config/visium/visium.c (visium_option_override): Likewise.
2264 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
2265 emit p2align format with last argument if it's not needed.
2266 * doc/invoke.texi: Document extended format of -falign-*.
2267 * final.c: Use align_labels alignment.
2268 * flags.h (struct target_flag_state): Change type to use
2270 (struct align_flags_tuple): New.
2271 (struct align_flags): Likewise.
2272 (align_loops_log): Redefine macro to use new types.
2273 (align_loops_max_skip): Redefine macro to use new types.
2274 (align_jumps_log): Redefine macro to use new types.
2275 (align_jumps_max_skip): Redefine macro to use new types.
2276 (align_labels_log): Redefine macro to use new types.
2277 (align_labels_max_skip): Redefine macro to use new types.
2278 (align_functions_log): Redefine macro to use new types.
2279 (align_loops): Redefine macro to use new types.
2280 (align_jumps): Redefine macro to use new types.
2281 (align_labels): Redefine macro to use new types.
2282 (align_functions): Redefine macro to use new types.
2283 (align_functions_max_skip): Redefine macro to use new types.
2284 (align_loops_value): New macro.
2285 (align_jumps_value): New macro.
2286 (align_labels_value): New macro.
2287 (align_functions_value): New macro.
2288 * function.c (invoke_set_current_function_hook): Propagate
2289 alignment values from flags to global variables default in
2291 * ipa-icf.c (sem_function::equals_wpa): Use
2292 cl_optimization_option_eq instead of memcmp.
2293 * lto-streamer.h (cl_optimization_stream_out): Support streaming
2295 (cl_optimization_stream_in): Likewise.
2296 * optc-save-gen.awk: Support strings in cl_optimization.
2297 * opth-gen.awk: Likewise.
2298 * opts.c (finish_options): Remove error checking of invalid
2300 (MAX_CODE_ALIGN): Remove.
2301 (MAX_CODE_ALIGN_VALUE): Likewise.
2302 (parse_and_check_align_values): New function.
2303 (check_alignment_argument): Likewise.
2304 (common_handle_option): Use check_alignment_argument.
2305 * opts.h (parse_and_check_align_values): Declare.
2306 * toplev.c (init_alignments): Remove.
2307 (read_log_maxskip): New.
2308 (parse_N_M): Likewise.
2309 (parse_alignment_opts): Likewise.
2310 (backend_init_target): Remove usage of init_alignments.
2311 * toplev.h (parse_alignment_opts): Declare.
2312 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
2314 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
2315 * tree.c (cl_option_hasher::equal): New.
2316 * varasm.c: Use new global macros.
2318 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
2320 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
2321 Use a simpler align directive also if MAXSKIP = ALIGN-1.
2322 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2323 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2324 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2325 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
2326 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
2327 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
2328 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
2329 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2330 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2331 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2332 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2334 2018-07-04 Martin Liska <mliska@suse.cz>
2335 Jonathan Wakely <jwakely@redhat.com>
2337 * coverage.c: Use correct type.
2338 * doc/invoke.texi: Language correction.
2340 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
2343 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
2344 ENDBRANCH for non-tail call which may return via indirect branch.
2345 * doc/extend.texi: Document indirect_return attribute.
2347 2018-07-03 Martin Sebor <msebor@redhat.com>
2349 PR tree-optimization/86274
2350 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
2352 (format_floating): Correct handling of infinities and NaNs.
2354 2018-07-03 Martin Sebor <msebor@redhat.com>
2356 * print-tree.c (print_real_cst): New function.
2357 (print_node_brief): Call it.
2358 (print_node): Ditto.
2360 2018-07-03 Jeff Law <law@redhat.com>
2362 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
2363 into a single pattern.
2365 * config/h8300/h8300.md (ors code_iterator): New.
2366 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
2367 a single pattern and single splitter.
2368 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
2369 (iorqi3_1, xorqi3_1): Likewise.
2370 (iorqi3, xorqi3 expanders): Similarly.
2372 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
2373 (movmd_internal) into a single pattern using the P mode iterator.
2374 (movmd splitters): Similarly.
2375 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
2376 (movsd splitters): Similarly.
2378 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
2379 ADDB, ADDW and ADDL into a single ADD attribute which selects the
2380 right table based on the size of the operand.
2381 * config/h8300/h8300.md (length_table): Corresponding changes. All
2382 references to "addb", "addw" and "addl" changed to "add".
2383 (btst patterns): Merge two variants into a single pattern.
2384 (tstqi, tsthi): Likewise.
2385 (addhi3_incdec, addsi3_incdec): Likewise.
2386 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
2387 (mulhi3, mulsi3): Likewise.
2388 (udivhi3, udivsi3): Likewise.
2389 (divhi3, divsi3): Likewise.
2390 (andorqi3, andorhi3, andorsi3): Likewise.
2392 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
2395 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
2396 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
2398 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2400 PR tree-optimization/85694
2401 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
2402 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
2404 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
2405 (<u>avg<mode>3_ceil): New patterns.
2407 2018-07-03 David Malcolm <dmalcolm@redhat.com>
2409 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
2410 scan-tree-dump directive.
2411 * gcc.dg/vect/slp-perm-2.c: Likewise.
2412 * gcc.dg/vect/slp-perm-3.c: Likewise.
2413 * gcc.dg/vect/slp-perm-5.c: Likewise.
2414 * gcc.dg/vect/slp-perm-6.c: Likewise.
2415 * gcc.dg/vect/slp-perm-7.c: Likewise.
2416 * gcc.dg/vect/slp-perm-8.c: Likewise.
2418 2018-07-03 Marek Polacek <polacek@redhat.com>
2421 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
2423 2018-07-03 Richard Biener <rguenther@suse.de>
2426 * tree-ssa-structalias.c (find_func_clobbers): Properly
2427 handle indirect calls.
2429 2018-07-03 Jeff Law <law@redhat.com>
2431 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
2432 (shifts): New code iterator.
2433 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
2434 expander. Fix HImode handling on H8/SX.
2435 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
2436 (subqi3, subhi3, subsi3 expanders): Likewise.
2437 (andqi3, andhi3, andsi3 expanders): Likewise.
2438 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
2439 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
2440 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
2441 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
2442 (zero_extendqihi2, zero_extendqisi2): Likewise.
2443 (extendqihi2, extendqisi2): Likewise.
2444 (rotlqi3, rotlhi3, rotlsi3): Likewise.
2445 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
2446 (rotlqi3_1, rotlhi3_1): Likewise.
2447 (logicalhi3_sn, logicalsi3_sn): Likewise.
2448 (logicalhi3, logicalsi3): Likewise.
2450 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2452 * tree-vect-patterns.c (vect_recog_rotate_pattern)
2453 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
2454 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
2455 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
2456 type to append_pattern_def_seq instead of creating a stmt_vec_info
2458 (build_mask_conversion): Likewise. Remove vinfo argument.
2459 (vect_add_conversion_to_patterm): Likewise, renaming to...
2460 (vect_add_conversion_to_pattern): ...this.
2461 (vect_recog_mask_conversion_pattern): Update call to
2462 build_mask_conversion. Pass the vector type to
2463 append_pattern_def_seq here too.
2464 (vect_recog_gather_scatter_pattern): Update call to
2465 vect_add_conversion_to_pattern.
2467 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2469 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
2470 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2471 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
2472 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
2473 STMT_VINFO_PATTERN_DEF_SEQ to null here.
2474 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
2475 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
2476 append_pattern_def_seq instead of new_pattern_def_seq.
2477 (vect_recog_divmod_pattern): Do both of the above.
2478 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
2481 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2483 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
2484 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
2485 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
2486 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
2487 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
2488 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
2489 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
2490 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
2491 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
2492 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
2493 parameter with a single stmt_vec_info.
2494 (vect_recog_func_ptr): Likewise.
2495 (vect_recog_gather_scatter_pattern): Likewise, folding in...
2496 (vect_try_gather_scatter_pattern): ...this.
2497 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
2498 the stmt_vec_info of the statement to be matched. Don't clear
2499 STMT_VINFO_RELATED_STMT.
2500 (vect_pattern_recog): Update call accordingly.
2502 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2504 PR tree-optimization/85694
2505 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
2506 (uavgM3_ceil): Document new optabs.
2507 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
2508 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
2510 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
2511 (savg_ceil_optab): New optabs.
2512 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
2513 (vect_vect_recog_func_ptrs): Add it.
2514 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
2515 constant directly from the associated lhs.
2517 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2519 * tree-vect-patterns.c (vect_split_statement): New function.
2520 (vect_convert_input): Use it to try to split an existing cast.
2522 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2524 * poly-int.h (print_hex): New function.
2525 * dumpfile.h (dump_dec, dump_hex): Declare.
2526 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
2527 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
2528 min_input_precision, operation_precision and operation_sign.
2529 * tree-vect-patterns.c (vect_get_range_info): New function.
2530 (vect_same_loop_or_bb_p, vect_single_imm_use)
2531 (vect_operation_fits_smaller_type): Delete.
2532 (vect_look_through_possible_promotion): Add an optional
2533 single_use_p parameter.
2534 (vect_recog_over_widening_pattern): Rewrite to use new
2535 stmt_vec_info infomration. Handle one operation at a time.
2536 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
2537 (vect_truncatable_operation_p, vect_set_operation_type)
2538 (vect_set_min_input_precision): New functions.
2539 (vect_determine_min_output_precision_1): Likewise.
2540 (vect_determine_min_output_precision): Likewise.
2541 (vect_determine_precisions_from_range): Likewise.
2542 (vect_determine_precisions_from_users): Likewise.
2543 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
2544 (vect_vect_recog_func_ptrs): Put over_widening first.
2546 (vect_pattern_recog): Call vect_determine_precisions.
2548 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2550 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
2551 statements that have been replaced by further pattern statements.
2552 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
2554 2018-07-03 Richard Biener <rguenther@suse.de>
2556 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
2557 always set *dt. Dump vectype in vectype overload.
2558 * dumpfile.h (dump_gimple_expr): New function.
2559 (dump_gimple_expr_loc): Likewise.
2560 * dumpfile.c (dump_gimple_expr): New function.
2561 (dump_gimple_expr_loc): Likewise.
2563 2018-07-02 Jeff Law <law@redhat.com>
2565 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
2566 the H8/300, H8/300H and H8/S variants into a single pattern.
2567 (movhi_h8300, movqi_h8300hs): Similarly.
2568 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
2569 (QHI mode iterator): New.
2571 * config/h8300/h8300.md: Remove trailing whitespace.
2573 2018-07-02 Jim Wilson <jimw@sifive.com>
2575 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
2576 instead of emit_insn for interrupt returns.
2577 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
2578 (riscv_sret, riscv_uret): Likewise.
2580 2018-07-02 David Malcolm <dmalcolm@redhat.com>
2582 * pretty-print.c (selftest::test_pp_format): Move save and restore
2583 of quotes to class auto_fix_quotes, and add an instance.
2584 * selftest.c: Include "intl.h".
2585 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
2586 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
2587 * selftest.h (selftest::auto_fix_quotes): New class.
2589 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
2591 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
2592 (aarch64_sve_prepare_conditional_op): Remove.
2593 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
2594 Allow aarch64_simd_reg_or_zero as select operand; remove
2595 the aarch64_sve_prepare_conditional_op call.
2596 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
2597 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
2598 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
2599 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
2600 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
2601 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
2602 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
2603 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
2604 and a splitters to match all of the *_any patterns.
2605 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
2607 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
2608 (SVE_COND_FP_BINARY_REV): Remove.
2609 (sve_int_op_rev, sve_fp_op_rev): New.
2610 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
2611 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
2612 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
2613 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
2614 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
2615 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
2616 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
2617 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
2618 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
2620 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
2621 Remove match_dup 1 from the inner unspec.
2622 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
2624 * config/aarch64/aarch64.md (movprfx): New attr.
2625 (length): Default movprfx to 8.
2626 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
2627 (*madd<SVE_I>, *msub<SVE_I): Likewise.
2628 (*<su>mul<SVE_I>3_highpart): Likewise.
2629 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
2630 (*v<ASHIFT><SVE_I>3): Likewise.
2631 (*<su><MAXMIN><SVE_I>3): Likewise.
2632 (*<su><MAXMIN><SVE_F>3): Likewise.
2633 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
2634 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
2635 (*div<SVE_F>4): Likewise.
2637 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
2639 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
2642 2018-07-02 Richard Biener <rguenther@suse.de>
2644 PR tree-optimization/86363
2645 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
2646 memset argument refers to a non-variable address.
2648 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
2650 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
2651 {PLUS,MINUS}_EXPR code to...
2652 (adjust_symbolic_bound): ...here,
2653 (combine_bound): ...here,
2654 (set_value_range_with_overflow): ...and here.
2656 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
2658 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
2660 (extract_range_from_abs_expr): ...here.
2662 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
2664 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
2665 -fno-omit-frame-pointer when not optimizing.
2667 2018-07-02 Martin Liska <mliska@suse.cz>
2670 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
2671 (propagate_nothrow): Likewise.
2673 2018-07-02 Martin Liska <mliska@suse.cz>
2676 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
2678 2018-07-02 David Malcolm <dmalcolm@redhat.com>
2680 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
2681 function in r262149, changing "loc" param from source_location to
2682 const dump_location_t &.
2683 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
2684 declaration, as above.
2686 2018-07-01 Paul Koning <ni1d@arrl.net>
2688 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
2689 -munit-asm, -mgnu-asm, -mdec-asm.
2690 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
2691 (pdp11_output_labelref): New.
2692 (pdp11_output_def): New.
2693 (pdp11_output_addr_vec_elt): New.
2694 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
2695 %# and %@ format codes.
2696 (pdp11_option_override): New.
2697 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
2698 (pdp11_output_ident): New.
2699 (pdp11_asm_named_section): New.
2700 (pdp11_asm_init_sections): New.
2701 (pdp11_file_start): New.
2702 (pdp11_file_end): New.
2703 (output_ascii): Use .ascii/.asciz for -mdec-asm.
2704 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
2705 %o, like %c but octal.
2706 (pdp11_option_override): New.
2707 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
2709 (DATA_SECTION_ASM_OP): Ditto.
2710 (READONLY_DATA_SECTION_ASM_OP): New.
2711 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
2712 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
2713 (ASM_OUTPUT_LABELREF): Ditto.
2714 (ASM_OUTPUT_DEF): Ditto.
2715 (ASM_OUTPUT_EXTERNAL): New.
2716 (ASM_OUTPUT_SOURCE_FILENAME): New.
2717 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
2718 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
2719 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
2720 %# and %@ format codes.
2721 * config/pdp11/pdp11.opt (mgnu-asm): New.
2722 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
2723 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
2724 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
2726 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
2728 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
2729 dereferencing path[] beyond its length.
2731 (debug_all_paths): New.
2732 (rewire_first_differing_edge): New.
2733 (adjust_paths_after_duplication): New.
2734 (duplicate_thread_path): Call adjust_paths_after_duplication.
2736 (thread_through_all_blocks): Add new argument to
2737 duplicate_thread_path.
2739 2018-06-30 Jim Wilson <jimw@sifive.com>
2741 * config/riscv/predicates.md (p2m1_shift_operand): New.
2742 (high_mask_shift_operand): New.
2743 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
2744 pattern using p2m1_shift_operand.
2745 (lshsi3_zero_extend_3+2): New combiner pattern using
2746 high_mask_shift_operand.
2748 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2750 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
2752 (vect_recog_rotate_pattern): ...here.
2753 (vect_convert_input): Try to insert casts of invariants in the
2755 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
2756 preheader to be empty.
2758 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2760 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
2761 vector type. If given, install it in the new statement's
2763 (vect_element_precision): New function.
2764 (vect_unpromoted_value): New struct.
2765 (vect_unpromoted_value::vect_unpromoted_value): New function.
2766 (vect_unpromoted_value::set_op): Likewise.
2767 (vect_look_through_possible_promotion): Likewise.
2768 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
2769 (vect_widened_op_tree, vect_convert_input): Likewise.
2770 (vect_convert_inputs, vect_convert_output): Likewise.
2771 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
2772 to handle the optional cast of the multiplication result and
2773 vect_widened_op_tree to detect the widened multiplication itself.
2774 Do not require the input and output of promotion casts to have
2775 the same sign, but base the signedness of the operation on the
2776 input rather than the result. If the pattern includes two
2777 promotions, check that those promotions have the same sign.
2778 Do not restrict the MULT_EXPR handling to a double-width result;
2779 handle quadruple-width results and wider. Use vect_convert_inputs
2780 to convert the inputs to the common type.
2781 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
2782 to handle the optional cast of the ABS result. Also allow a sign
2783 change or a sign extension between the ABS and MINUS.
2784 Use vect_widened_op_tree to detect the widened subtraction and use
2785 vect_convert_inputs to convert the inputs to the common type.
2786 (vect_handle_widen_op_by_const): Delete.
2787 (vect_recog_widen_op_pattern): New function.
2788 (vect_recog_widen_mult_pattern): Use it.
2789 (vect_recog_widen_shift_pattern): Likewise.
2790 (vect_recog_widen_sum_pattern): Use
2791 vect_look_through_possible_promotion to handle the promoted
2794 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2796 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
2797 the containing gimple_seq *.
2798 * gimple-iterator.h (gsi_for_stmt): Declare it.
2799 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
2800 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
2801 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
2802 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
2803 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
2805 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
2807 (vect_mark_pattern_stmts): ...here. Handle cases in which the
2808 statement being replaced is part of an existing pattern
2809 definition sequence, inserting the new pattern statements before
2811 (vect_pattern_recog_1): Don't return a bool. If the statement
2812 is already part of a pattern, instead apply pattern matching
2813 to the pattern definition statements. Don't clear the
2814 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
2815 (vect_pattern_recog): Don't break after the first match;
2816 continue processing the pattern definition statements instead.
2817 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
2819 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2821 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
2822 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2823 (vect_recog_widen_sum_pattern): Use it.
2825 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2827 * tree-vect-loop.c (vectorizable_reduction): Assert that the
2828 phi is not a pattern statement and has not been replaced by
2829 a pattern statement.
2830 * tree-vect-patterns.c (type_conversion_p): Don't check
2831 STMT_VINFO_IN_PATTERN_P.
2832 (vect_recog_vector_vector_shift_pattern): Likewise.
2833 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
2834 the pattern statement rather than the original statement; check
2835 directly for a WIDEN_MULT_EXPR here.
2836 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
2837 vect_is_simple_use to return the pattern statement rather
2838 than the original statement; use is_pattern_stmt_p to check
2839 for such a pattern statement.
2840 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
2841 to return the pattern statement rather than the original statement;
2842 don't do the same transformation here.
2843 (vect_is_simple_use): If the defining statement has been replaced
2844 by a pattern statement, return the pattern statement instead.
2845 Remove the corresponding (local) transformation from the vectype
2848 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2850 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
2851 end and default to null.
2852 * tree-vect-loop.c (vect_create_epilog_for_reduction)
2853 (vectorizable_reduction): Update calls accordingly, dropping the
2854 gimple ** argument if the passed-back statement isn't needed.
2855 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
2856 (vect_recog_rotate_pattern): Likewise.
2857 (vect_recog_mask_conversion_pattern): Likewise.
2858 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
2859 (vect_mask_constant_operand_p): Likewise.
2860 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
2861 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
2862 (get_group_load_store_type, get_load_store_type): Likewise.
2863 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
2864 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
2865 (vectorizable_conversion, vectorizable_assignment): Likewise.
2866 (vectorizable_shift, vectorizable_operation): Likewise.
2867 (vectorizable_store, vect_is_simple_cond): Likewise.
2868 (vectorizable_condition, vectorizable_comparison): Likewise.
2869 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
2870 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
2871 and move it to the end. Cope with null def_stmt_outs.
2873 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
2875 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
2877 2018-06-29 Jeff Law <law@redhat.com>
2879 * config/v850/v850.c (v850_legitimate_address_p): Handle large
2880 displacements for TARGET_V850E2V3 and newer.
2881 (TARGET_LRA_P): Remove. Defaults to LRA now.
2882 * config/v850/v850.md (sign23byte_load): Remove.
2883 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
2884 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
2886 2018-06-29 Martin Liska <mliska@suse.cz>
2889 * coverage.c (coverage_init): Mangle full path name.
2890 * doc/invoke.texi: Document the change.
2891 * gcov-io.c (mangle_path): New.
2892 * gcov-io.h (mangle_path): Likewise.
2893 * gcov.c (mangle_name): Use mangle_path for path mangling.
2895 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2897 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
2898 if starting source register is not even.
2900 2018-06-29 Martin Liska <mliska@suse.cz>
2902 PR tree-optimization/86263
2903 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
2904 Make edge redirection.
2906 2018-06-29 David Malcolm <dmalcolm@redhat.com>
2908 * dumpfile.c (dump_loc): Add indentation based on scope depth.
2909 (dump_scope_depth): New variable.
2910 (get_dump_scope_depth): New function.
2911 (dump_begin_scope): New function.
2912 (dump_end_scope): New function.
2913 * dumpfile.h (get_dump_scope_depth): New declaration.
2914 (dump_begin_scope): New declaration.
2915 (dump_end_scope): New declaration.
2916 (class auto_dump_scope): New class.
2917 (AUTO_DUMP_SCOPE): New macro.
2918 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
2921 2018-06-29 Richard Biener <rguenther@suse.de>
2923 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
2924 compute_all_dependences succeeds.
2925 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
2926 exceed --param loop-max-datarefs-for-datadeps.
2928 2018-06-29 Jakub Jelinek <jakub@redhat.com>
2930 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
2932 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
2935 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
2936 alternative 0 in preferred_for_speed attribute.
2938 2018-06-28 Paul Koning <ni1d@arrl.net>
2940 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
2941 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
2942 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
2943 * config/pdp11/pdp11.md: Correct "length" attribute calculation
2944 for shift insn patterns.
2946 2018-06-28 David Malcolm <dmalcolm@redhat.com>
2948 * cgraph.c (cgraph_node::get_body): Replace assignments to
2949 "dump_file" with calls to set_dump_file.
2950 * dumpfile.c (alt_dump_file): Make static, and group with...
2951 (alt_flags): ...this definition.
2952 (dumps_are_enabled): New variable.
2953 (refresh_dumps_are_enabled): New function.
2954 (set_dump_file): New function.
2955 (set_alt_dump_file): New function.
2956 (gcc::dump_manager::dump_start): Replace assignments to
2957 "dump_file" and "alt_dump_file" with calls to set_dump_file and
2959 (gcc::dump_manager::dump_finish): Likewise.
2960 * dumpfile.h (alt_dump_file): Delete decl.
2961 (dumps_are_enabled): New variable decl.
2962 (set_dump_file): New function decl.
2963 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
2965 * tree-nested.c (lower_nested_functions): Replace assignments to
2966 "dump_file" with calls to set_dump_file.
2968 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
2970 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
2971 goto_locus of each outgoing edge of each basic block.
2973 2018-06-28 Richard Biener <rguenther@suse.de>
2975 * dwarf2out.c (decl_scope_table): Remove.
2976 (push_decl_scope): Likewise.
2977 (pop_decl_scope): Likewise.
2978 (gen_type_die_for_member): Do not call push/pop_decl_scope.
2979 (gen_struct_or_union_type_die): Likewise.
2980 (gen_tagged_type_die): Likewise.
2981 (dwarf2out_init): Do not initialize decl_scope_table.
2982 (dwarf2out_c_finalize): Do not free it.
2984 2018-06-28 Richard Biener <rguenther@suse.de>
2986 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
2987 deciding whether to not re-use a DIE.
2989 2018-06-28 Richard Biener <rguenther@suse.de>
2991 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
2992 DW_AT_abstract_origin attribute.
2994 2018-06-28 Martin Liska <mliska@suse.cz>
2996 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
2997 Use newly introduced constants.
2998 * tree-switch-conversion.h (struct jump_table_cluster):
2999 Define max_ratio_for_size and max_ratio_for_speed.
3001 2018-06-28 Martin Liska <mliska@suse.cz>
3003 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
3004 Add new checking assert to catch invalid state.
3005 (jump_table_cluster::can_be_handled): Handle single case
3007 (jump_table_cluster::is_beneficial): Bail out for such case.
3008 (bit_test_cluster::find_bit_tests):
3009 Add new checking assert to catch invalid state.
3010 (bit_test_cluster::can_be_handled): Handle single case
3012 (bit_test_cluster::is_beneficial): Bail out for such case.
3013 (switch_decision_tree::analyze_switch_statement):
3016 2018-06-28 Martin Liska <mliska@suse.cz>
3018 * common.opt: Introduce -completion option.
3019 * gcc.c (driver_handle_option): Handle it.
3020 (driver::main): Print completions if completion
3022 * opt-suggestions.c (option_proposer::get_completions):
3024 (option_proposer::suggest_completion): Likewise.
3025 (option_proposer::find_param_completions): Likewise.
3026 (verify_autocompletions): Likewise.
3027 (test_completion_valid_options): Likewise.
3028 (test_completion_valid_params): Likewise.
3029 (in_completion_p): Likewise.
3030 (empty_completion_p): Likewise.
3031 (test_completion_partial_match): Likewise.
3032 (test_completion_garbage): Likewise.
3033 (opt_proposer_c_tests): Likewise.
3034 * opt-suggestions.h: Declare new functions.
3035 * opts.c (common_handle_option): Handle OPT__completion_.
3036 * selftest-run-tests.c (selftest::run_tests): Add
3037 opt_proposer_c_tests.
3038 * selftest.c (assert_str_startswith): New.
3039 * selftest.h (assert_str_startswith): Likewise.
3040 (opt_proposer_c_tests): New.
3041 (ASSERT_STR_STARTSWITH): Likewise.
3043 2018-06-28 Martin Liska <mliska@suse.cz>
3045 * Makefile.in: Add opt-suggestions.o.
3046 * gcc-main.c: Include opt-suggestions.h.
3047 * gcc.c (driver::driver): Likewise.
3048 (driver::~driver): Remove m_option_suggestions.
3049 (driver::build_option_suggestions): Moved to option_proposer.
3050 (driver::suggest_option): Likewise.
3051 (driver::handle_unrecognized_options): Use option_proposer.
3052 * gcc.h (class driver): Add new memver m_option_proposer.
3053 * opt-suggestions.c: New file.
3054 * opt-suggestions.h: New file.
3056 2018-06-28 Martin Liska <mliska@suse.cz>
3058 * vec.h (class auto_string_vec): New (moved from auto_argvec).
3059 (auto_string_vec::~auto_string_vec): Likewise.
3061 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
3063 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
3064 prevent_decl_creation_for_types fields up and add reset_location field.
3065 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
3066 statement if id->reset_location is true.
3067 (copy_edges_for_bb): Do not set goto_locus on the new edges if
3068 id->reset_location is true.
3069 (copy_phis_for_bb): Force input_location on the arguments if
3070 id->reset_location is true.
3071 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
3072 is set on the function to be inlined.
3074 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
3076 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
3078 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
3080 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
3081 registers for Pmode.
3082 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
3083 hard registers for the clobbered pseudo.
3085 2018-06-27 Paul Koning <ni1d@arrl.net>
3087 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
3088 mutually exclusive options.
3089 * config/pdp11/constraints.md (h): New constraint.
3090 (O): Update definition to match shift code generation.
3091 (D): New constraint.
3092 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
3094 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
3096 (output_jump): Change arguments.
3097 (pdp11_fixed_cc_regs): New function.
3098 (pdp11_cc_mode): Ditto.
3099 (pdp11_expand_shift): Ditto.
3100 (pdp11_assemble_shift): Ditto.
3101 (pdp11_small_shift): Ditto.
3102 (pdp11_branch_cost): Remove.
3103 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
3105 (pdp11_register_move_cost): Update for CC registers.
3106 (pdp11_rtx_costs): Add case for LSHIFTRT.
3107 (pdp11_output_jump): Add CCNZ mode conditional branches.
3108 (notice_update_cc_on_set): Remove.
3109 (pdp11_cc_mode): New function.
3110 (simple_memory_operand): Correct pre/post decrement case.
3111 (no_side_effect_operand): New function.
3112 (pdp11_regno_reg_class): Add CC_REGS class.
3113 (pdp11_fixed_cc_regs): New function.
3114 (pdp11_small_shift): New function.
3115 (pdp11_expand_shift): New function to expand shift insns.
3116 (pdp11_assemble_shift): New function to output shifts.
3117 (pdp11_branch_cost): Remove.
3118 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
3119 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
3120 (WCHAR_TYPE): Ditto.
3121 (PTRDIFF_TYPE): Ditto.
3122 (ADJUST_INSN_LENGTH): New macro.
3123 (FIXED_REGISTERS): Add CC registers.
3124 (CALL_USED_REGISTERS): Ditto.
3126 (REG_CLASS_NAMES): Ditto.
3127 (REG_CLASS_CONTENTS): Ditto.
3128 (SELECT_CC_MODE): Use new function.
3129 (TARGET_FLAGS_REGNUM): New macro.
3130 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
3131 (cc0_reg_rtx): Remove.
3132 (CC_STATUS_MDEP): Remove.
3133 (CC_STATUS_MDEFP_INIT): Remove.
3134 (CC_IN_FPU): Remove.
3135 (NOTICE_UPDATE_CC): Remove.
3136 (REGISTER_NAMES): Add CC registers.
3137 (BRANCH_COST): Change to constant 1.
3138 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
3140 * config/pdp11/pdp11.opt (mbcopy): Remove.
3141 (mbcopy-builtin): Remove.
3142 (mbranch-cheap): Remove.
3143 (mbranch-expensive): Remove.
3144 * config/pdp11/predicates.md (expand_shift_operand): Update to
3145 match shift code generation.
3146 (ccnz_operator): New predicate.
3147 * doc/invoke.texi (PDP-11 Options): Remove deleted options
3148 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
3149 Remove non-existent option -mabshi, -mno-abshi. Document mutually
3151 * doc/md.texi (PDP-11): Document new D and h constraints. Update
3152 description of O constraint.
3154 2018-06-27 Jeff Law <law@redhat.com>
3155 Austin Law <austinklaw@gmail.com>
3157 * config/v850/v850.md (addsi3_set_flags): New pattern.
3158 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
3159 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
3160 (zero_extendhisi2_v850_set_flags): Likewise.
3161 (zero_extendqisi2_v850_set_flags): Likewise.
3162 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
3163 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
3164 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
3166 * config/v850/v850-protos.h (notice_update_cc): Remove.
3167 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
3168 (v850_print_operand): Handle 'D' and "d".
3169 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
3170 Add handling of arithmetic/logical operations compared against zero.
3171 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
3172 Do not look at v850_compare_op, instead get mode from last argument.
3173 (v850_gen_compare): Remove
3174 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
3175 after reload for prologue insns.
3176 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
3178 (construct_save_jarl): Likewise.
3179 (TARGET_FLAGS_REGNUM): Define.
3180 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
3181 (NOTICE_UPDATE_CC): Remove.
3182 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
3183 than cc0. Conditionalize on reload_completed.
3184 (cmpsi_insn, setfcc_insn): Likewise.
3185 (tst1 splitter): Turn into define_and_split which sets the flags
3187 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
3188 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
3189 (cstoresf4, cstoredf4): Clobber the flags.
3190 (cmpsi, cmpsf, cmpdf): Remove expanders.
3191 (setf_insn): Remove pattern.
3192 (addsi3): Turn into define_and_split which clobbers the flags after
3193 reload and a suitable pattern (addsi3_clobber_flags) for use after
3195 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
3196 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
3197 (ashrsi3, ashrsi3_v850e2): Likewise.
3198 (bins): Clobber the flags.
3199 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
3200 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
3201 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
3202 (call_value_internal_short, call_value_internal_long): Likewise.
3203 (callt_save_interrupt, callt_return_interrupt): Likewise.
3204 (save_interrupt, return_interrupt): Likewise.
3205 (callt_save_all_interrupt, save_all_interrupt): Likewise.
3206 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
3207 (restore_all_interrupt, _restore_all_interrupt): Likewise.
3208 (All FP comparisons): Only allow after reload has completed.
3210 (divh, divhu): Tweak output template.
3211 (branch_z_normal, branch_z_invert): Remove
3212 (branch_nz_normal, branch_nz_invert): Likewise.
3213 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
3215 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
3216 * config/v850/v850.c (notice_update_cc): Remove.
3217 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
3218 (CC_NO_CARRY): Likewise.
3219 (NOTICE_UPDATE_CC): Define to nothing.
3220 * config/v850/v850.md: Remove block comment on cc0 handling
3221 Remove "cc" attribute from all patterns. Remove cc_status handling
3222 from all patterns. Minor formatting fixes.
3224 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3226 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
3227 (cortex-a76.cortex-a55): Likewise.
3228 * config/aarch64/aarch64-tune.md: Regenerate.
3229 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
3230 cortex-a76.cortex-a55.
3232 2018-06-27 Jeff Law <law@redhat.com>
3234 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
3235 (MULTILIB_DIRNAMES): Similarly.
3237 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
3239 * gimple.h (gimple_return_retbnd): Delete.
3240 (gimple_return_set_retbnd): Likewise.
3241 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
3242 gimple_return_set_retbnd.
3243 * gimple-pretty-print.c (dump_gimple_return): Remove call to
3244 gimple_return_retbnd and adjust.
3245 * tree-inline.h (struct copy_body_data): Remove retbnd field.
3246 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
3247 Explicitly return NULL in a couple more cases. Move assertion
3248 on debug statements and remove unreachable code.
3249 (reset_debug_binding): Do not test id->retbnd.
3250 (expand_call_inline): Do not set it.
3252 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
3254 * configure.ac: Add --disable-gcov option.
3255 * configure: Regenerate.
3256 * Makefile.in: Honour @enable_gcov@.
3257 * doc/install.texi: Document --disable-gcov.
3259 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3261 * config/arm/arm-cpus.in (cortex-a76): New entry.
3262 (cortex-a76.cortex-a55): Likewise.
3263 * config/arm/arm-tables.opt: Regenerate.
3264 * config/arm/arm-tune.md: Likewise.
3265 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
3266 * doc/invoke.texi (ARM Options): Document cortex-a76 and
3267 cortex-a76.cortex-a55.
3269 2018-06-27 Tamar Christina <tamar.christina@arm.com>
3272 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
3274 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
3276 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
3278 (EPILOGUE_USES): Likewise.
3280 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
3282 * tree-inline.c (remap_location): New function extracted from...
3283 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
3284 (copy_phis_for_bb): ...here. Call remap_location.
3285 (copy_cfg_body): Adjust call to copy_edges_for_bb.
3287 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
3289 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
3290 unaligned vsx for 16B memset.
3292 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
3295 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
3296 ieee128_float_type_node to long_double_type_node unless
3297 TARGET_LONG_DOUBLE_128 is set.
3299 2018-06-26 David Malcolm <dmalcolm@redhat.com>
3301 * cfgloop.c (get_loop_location): Convert return type from
3302 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
3303 by implicit construction from rtx_insn *, and using
3304 dump_user_location_t::from_function_decl for the fallback case.
3305 * cfgloop.h (get_loop_location): Convert return type from
3306 location_t to dump_user_location_t.
3307 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
3308 dump_printf_loc to pass in a dump_location_t rather than a
3309 location_t, via the gimple stmt.
3310 * coverage.c (get_coverage_counts): Update calls to
3311 dump_printf_loc to pass in dump_location_t rather than a
3313 * doc/optinfo.texi (Dump types): Convert example of
3314 dump_printf_loc from taking "locus" to taking "insn". Update
3315 description of the "_loc" calls to cover dump_location_t.
3316 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
3318 (dump_user_location_t::dump_user_location_t): New constructors,
3319 from gimple *stmt and rtx_insn *.
3320 (dump_user_location_t::from_function_decl): New function.
3321 (dump_loc): Make static.
3322 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
3323 const dump_location_t &.
3324 (dump_generic_expr_loc): Delete.
3325 (dump_printf_loc): Convert param "loc" from location_t to
3326 const dump_location_t &.
3327 (selftest::test_impl_location): New function.
3328 (selftest::dumpfile_c_tests): New function.
3329 * dumpfile.h: Include "profile-count.h".
3330 (class dump_user_location_t): New class.
3331 (struct dump_impl_location_t): New struct.
3332 (class dump_location_t): New class.
3333 (dump_printf_loc): Convert 2nd param from source_location to
3334 const dump_location_t &.
3335 (dump_generic_expr_loc): Delete.
3336 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
3337 const dump_location_t &.
3338 * gimple-fold.c (fold_gimple_assign): Update call to
3339 dump_printf_loc to pass in a dump_location_t rather than a
3340 location_t, via the gimple stmt.
3341 (gimple_fold_call): Likewise.
3342 * gimple-loop-interchange.cc
3343 (loop_cand::analyze_iloop_reduction_var): Update for change to
3344 check_reduction_path.
3345 (tree_loop_interchange::interchange): Update for change to
3347 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
3348 change in return-type of find_loop_location.
3349 (graphite_regenerate_ast_isl): Likewise.
3350 * graphite-optimize-isl.c (optimize_isl): Likewise.
3351 * graphite.c (graphite_transform_loops): Likewise.
3352 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
3353 pass in a dump_location_t rather than a location_t, via the
3355 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3356 * ipa.c (walk_polymorphic_call_targets): Likewise.
3357 * loop-unroll.c (report_unroll): Convert "locus" param from
3358 location_t to dump_location_t.
3359 (decide_unrolling): Update for change to get_loop_location's
3361 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
3362 location_t to dump_user_location_t.
3363 (grid_find_single_omp_among_assignments_1): Updates calls to
3364 dump_printf_loc to pass in a dump_location_t rather than a
3365 location_t, via the gimple stmt.
3366 (grid_parallel_clauses_gridifiable): Convert "tloc" from
3367 location_t to dump_location_t. Updates calls to dump_printf_loc
3368 to pass in a dump_location_t rather than a location_t, via the
3370 (grid_inner_loop_gridifiable_p): Likewise.
3371 (grid_dist_follows_simple_pattern): Likewise.
3372 (grid_gfor_follows_tiling_pattern): Likewise.
3373 (grid_target_follows_gridifiable_pattern): Likewise.
3374 (grid_attempt_target_gridification): Convert initialization
3375 of local "grid" from memset to zero-initialization; FIXME: does
3376 this require C++11? Update call to dump_printf_loc to pass in a
3377 optinfo_location rather than a location_t, via the gimple stmt.
3378 * profile.c (read_profile_edge_counts): Updates call to
3379 dump_printf_loc to pass in a dump_location_t rather than a
3381 (compute_branch_probabilities): Likewise.
3382 * selftest-run-tests.c (selftest::run_tests): Call
3384 * selftest.h (dumpfile_c_tests): New decl.
3385 * tree-loop-distribution.c (pass_loop_distribution::execute):
3386 Update for change in return type of find_loop_location.
3387 * tree-parloops.c (parallelize_loops): Likewise.
3388 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
3389 "locus" from location_t to dump_user_location_t.
3390 (canonicalize_loop_induction_variables): Likewise.
3391 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
3392 for change in return type of find_loop_location.
3393 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
3394 to dump_printf_loc to pass in a dump_location_t rather than a
3395 location_t, via the stmt.
3396 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
3398 * tree-vect-loop-manip.c (find_loop_location): Convert return
3399 type from source_location to dump_user_location_t.
3400 (vect_do_peeling): Update for above change.
3401 (vect_loop_versioning): Update for change in type of
3403 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
3404 from location_t to dump_user_location_t.
3405 (vect_estimate_min_profitable_iters): Update for change in type
3407 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
3408 location_t to dump_location_t.
3409 (vect_slp_bb): Update for change in type of vect_location.
3410 * tree-vectorizer.c (vect_location): Convert from source_location
3411 to dump_user_location_t.
3412 (try_vectorize_loop_1): Update for change in vect_location's type.
3413 (vectorize_loops): Likewise.
3414 (increase_alignment): Likewise.
3415 * tree-vectorizer.h (vect_location): Convert from source_location
3416 to dump_user_location_t.
3417 (find_loop_location): Convert return type from source_location to
3418 dump_user_location_t.
3419 (check_reduction_path): Convert 1st param from location_t to
3420 dump_user_location_t.
3421 * value-prof.c (check_counter): Update call to dump_printf_loc to
3422 pass in a dump_user_location_t rather than a location_t; update
3423 call to error_at for change in type of "locus".
3424 (check_ic_target): Update call to dump_printf_loc to
3425 pass in a dump_user_location_t rather than a location_t, via the
3428 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
3430 * config/s390/s390.h (enum processor_flags): Do not use
3431 default tune parameter when -march was specified.
3433 2018-06-26 Jakub Jelinek <jakub@redhat.com>
3436 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
3437 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
3440 2018-06-26 Richard Biener <rguenther@suse.de>
3442 PR tree-optimization/86287
3444 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
3445 (vect_analyze_loop): Initialize n_stmts.
3447 2018-06-26 Richard Biener <rguenther@suse.de>
3450 * fold-const.c (fold_convertible_p): Pointer extension
3453 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
3456 * dwarf2out.c (loc_list_has_views): Adjust comments.
3457 (dw_loc_list): Split single cross-partition range with
3460 2018-06-25 Jeff Law <law@redhat.com>
3462 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
3463 on -mbig-switch by default.
3465 * config/v850/predicates.md (const_float_1_operand): Fix match_code
3467 (const_float_0_operand): Remove unused predicate.
3468 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
3469 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
3470 (recipsf2): New expander. Original pattern now called
3472 (recipdf2, recipdf2_insn): Similarly.
3473 (rsqrtsf2, rsqrtsf2_insn): Similarly
3474 (rsqrtdf2, rsqrtdf2_insn): Similarly
3476 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
3478 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
3479 Simplify logic for FreeBSD (twice).
3481 2018-06-25 Martin Sebor <msebor@redhat.com>
3483 PR tree-optimization/86204
3484 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
3485 a strnlen result if it's less than the length of the string.
3487 2018-06-25 Martin Sebor <msebor@redhat.com>
3489 PR tree-optimization/85700
3490 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
3491 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
3492 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
3494 2018-06-25 Martin Sebor <msebor@redhat.com>
3496 * doc/extend.texi (Zero-length arrays): Update and clarify.
3498 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
3500 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
3501 added IEEE/IBM long double multilib support on PowerPC little
3502 endian Linux systems.
3503 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
3504 (MULTILIB_DEFAULTS): Likewise.
3505 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3507 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
3508 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
3509 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
3511 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
3514 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
3515 (REORDER_45): Likewise.
3517 2018-06-25 Jeff Law <law@redhat.com>
3519 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
3520 dividend to 32 bits. Adjust length.
3521 (udivmodhi4): Cleanup output template. Fix length.
3523 2018-06-25 Carl Love <cel@us.ibm.com>
3525 * config/rs6000/vsx.md: Change word selector to prefered location.
3527 2018-06-25 Richard Biener <rguenther@suse.de>
3529 PR tree-optimization/86304
3530 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
3531 epilogue-if-converted loops as well.
3533 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
3535 * lto-section-out.c (lto_begin_section): Do not print section
3536 name for noaddr and unnumbered dumps.
3538 2018-06-25 Richard Biener <rguenther@suse.de>
3540 * tree-vectorizer.h (struct vec_info_shared): New structure
3541 with parts split out from struct vec_info and loop_nest from
3542 struct _loop_vec_info.
3543 (struct vec_info): Adjust accordingly.
3544 (struct _loop_vec_info): Likewise.
3545 (LOOP_VINFO_LOOP_NEST): Adjust.
3546 (LOOP_VINFO_DATAREFS): Likewise.
3547 (LOOP_VINFO_DDRS): Likewise.
3548 (struct _bb_vec_info): Likewise.
3549 (BB_VINFO_DATAREFS): Likewise.
3550 (BB_VINFO_DDRS): Likewise.
3551 (struct _stmt_vec_info): Add dr_aux member.
3552 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
3553 (DR_MISALIGNMENT_UNINITIALIZED): New.
3554 (set_dr_misalignment): Adjust.
3555 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
3556 (vect_analyze_loop): Adjust prototype.
3557 (vect_analyze_loop_form): Likewise.
3558 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
3559 Compute dependences lazily.
3560 (vect_record_base_alignments): Use shared datarefs/ddrs.
3561 (vect_verify_datarefs_alignment): Likewise.
3562 (vect_analyze_data_refs_alignment): Likewise.
3563 (vect_analyze_data_ref_accesses): Likewise.
3564 (vect_analyze_data_refs): Likewise.
3565 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
3566 constructor parameter for shared part.
3567 (vect_analyze_loop_form): Pass in shared part and adjust.
3568 (vect_analyze_loop_2): Pass in storage for the number of
3569 stmts. Move loop nest finding to the caller. Compute
3571 (vect_analyze_loop): Pass in shared part.
3572 (vect_transform_loop): Verify shared datarefs are unchanged.
3573 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
3574 constructor parameter for shared part.
3575 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
3576 (vect_slp_bb): Verify shared datarefs are unchanged before
3578 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
3580 (new_stmt_vec_info): Initialize DR_AUX misalignment to
3581 DR_MISALIGNMENT_UNINITIALIZED.
3582 * tree-vectorizer.c (vec_info::vec_info): Add constructor
3583 parameter for shared part.
3584 (vec_info::~vec_info): Adjust.
3585 (vec_info_shared::vec_info_shared): New.
3586 (vec_info_shared::~vec_info_shared): Likewise.
3587 (vec_info_shared::save_datarefs): Likewise.
3588 (vec_info_shared::check_datarefs): Likewise.
3589 (try_vectorize_loop_1): Construct shared part live for analyses
3590 of a single loop for multiple vector sizes.
3591 * tree-parloops.c (gather_scalar_reductions): Adjust.
3593 2018-06-25 Richard Biener <rguenther@suse.de>
3595 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
3596 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
3597 (vect_analyze_data_refs): Remove similar code from here and
3598 simplify accordingly.
3600 2018-06-25 Richard Biener <rguenther@suse.de>
3602 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
3603 for reverse storage order accesses rather than asserting
3604 they cannot happen here.
3606 2018-06-25 Tom de Vries <tdevries@suse.de>
3609 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
3610 Use data16 instead of .byte for insn prefix.
3612 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
3615 * parser.c (make_char_string_pack): Pass this literal chars
3616 through cpp_interpret_string.
3617 (cp_parser_userdef_numeric_literal): Check the result of
3618 make_char_string_pack.
3620 2018-06-24 Maya Rashish <coypu@sdf.org>
3622 * ginclude/stddef.h: Simplify conditions around avoiding
3623 re-definition of __size_t.
3625 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
3627 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
3628 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
3630 2018-06-22 Maya Rashish <coypu@sdf.org>
3632 * doc/invoke.texi (mno-fancy-math-387): Update for changes
3633 made to OpenBSD and NetBSD through the years.
3635 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3637 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
3638 behavior of vec_pack (vector double, vector double) to match
3639 behavior of vec_float2 (vector double, vector double).
3641 2018-06-22 Olivier Hainque <hainque@adacore.com>
3643 * gimplify.c (gimplify_function_tree): Prevent creation
3644 of a trampoline for the address of the current function
3645 passed to entry/exit instrumentation hooks.
3647 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
3650 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
3653 2018-06-22 Martin Liska <mliska@suse.cz>
3655 PR tree-optimization/86263
3656 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
3657 Bail out if is_enabled is false.
3658 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
3660 (jump_table_cluster::is_enabled): New function.
3662 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
3664 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
3665 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
3666 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
3667 (lto_input_ts_binfo_tree_pointers): Likewise.
3668 * tree-streamer-out.c (streamer_write_tree_bitfields,
3669 write_ts_binfo_tree_pointers): Likewise.
3670 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
3672 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
3674 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
3676 2018-06-22 Martin Liska <mliska@suse.cz>
3678 * symbol-summary.h (get): Make it pure and inline move
3679 functionality from ::get function.
3680 (get): Remove and inline into ::get and ::get_create.
3681 (get_create): Move code from ::get function.
3683 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3686 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
3687 -x assembler-with-cpp.
3689 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3691 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
3692 _FILE_OFFSET_BITS=64 for C++.
3694 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
3696 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
3697 conversion insn that shows up when pr85657-3.c is compiled using
3698 IEEE 128-bit long double.
3699 (neg<mode>2_internal): Use the correct mode to check whether the
3700 mode is IBM extended.
3701 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
3702 multiply and divide external functions from being created more
3705 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3707 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
3709 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
3710 the edge can be forwarded.
3711 (cfg_layout_merge_blocks): Likewise.
3713 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3715 * except.c (finish_eh_generation): Commit edge insertions only after
3716 the EH edges have been redirected from post-landing to landing pads.
3718 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3720 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
3721 create_tmp_var_for to create the FRAME decl.
3722 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
3724 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3726 * tree-inline.c (copy_edges_for_bb): Minor tweak.
3727 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
3728 debug statement when resetting its value.
3729 (expand_call_inline): Copy the locus of the call onto the assignment
3730 of the return value, if any. Use local variable in more cases.
3732 2018-06-21 Martin Liska <mliska@suse.cz>
3734 * ipa-pure-const.c (propagate_nothrow): Use
3735 funct_state_summaries->get.
3736 (dump_malloc_lattice): Likewise.
3737 (propagate_malloc): Likewise.
3739 2018-06-21 Richard Biener <rguenther@suse.de>
3741 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
3742 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
3743 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
3745 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
3746 BLOCK_ABSTRACT_ORIGIN unconditionally.
3748 2018-06-21 David Malcolm <dmalcolm@redhat.com>
3750 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
3752 * ipa-reference.c (ipa_reference_c_finalize): Delete
3753 ipa_ref_opt_sum_summaries and set it to NULL.
3755 2018-06-21 Tom de Vries <tdevries@suse.de>
3757 PR tree-optimization/85859
3758 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
3759 test with comment from bb_no_side_effects_p.
3761 2018-06-21 Richard Biener <rguenther@suse.de>
3763 PR tree-optimization/86232
3764 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
3765 max for constant niter.
3767 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
3769 * config/aarch64/aarch64-simd.md
3770 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
3772 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
3774 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
3775 Make opernads of the unspec commutative.
3777 2018-06-21 Richard Biener <rguenther@suse.de>
3779 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
3780 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
3781 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
3782 (vect_analyze_data_ref_dependence): Re-order checks to deal with
3784 (vect_record_base_alignments): Do not record base alignment
3785 for gathers or scatters.
3786 (vect_compute_data_ref_alignment): Drop return value that is always
3787 true. Bail out early for gathers or scatters.
3788 (vect_enhance_data_refs_alignment): Bail out early for gathers
3790 (vect_find_same_alignment_drs): Likewise.
3791 (vect_analyze_data_refs_alignment): Remove dead code.
3792 (vect_slp_analyze_and_verify_node_alignment): Likewise.
3793 (vect_analyze_data_refs): For possible gathers or scatters do
3794 not create an alternate DR, just check their possible validity
3795 and mark them. Adjust DECL_NONALIASED handling to not rely
3797 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
3798 update inits of gathers or scatters.
3799 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
3800 Also copy gather/scatter flag to pattern vinfo.
3802 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
3804 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
3805 behavior of vec_packsu (vector unsigned long long, vector unsigned
3806 long long) to match behavior of vec_packs with same signature.
3808 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
3809 Thomas Schwinge <thomas@codesourcery.com>
3810 Cesar Philippidis <cesar@codesourcery.com>
3812 * gimplify.c (gimplify_scan_omp_clauses): Add support for
3813 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
3814 (gimplify_adjust_omp_clauses): Likewise.
3815 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
3816 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
3817 (gimplify_omp_target_update): Update handling of acc update and
3819 * omp-low.c (install_var_field): Remove unused parameter
3820 base_pointers_restrict.
3821 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
3822 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
3824 (omp_target_base_pointers_restrict_p): Delete.
3825 (scan_omp_target): Update call to scan_sharing_clauses.
3826 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
3828 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
3829 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
3830 (convert_local_omp_clauses): Likewise.
3831 * tree-pretty-print.c (dump_omp_clause): Likewise.
3832 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
3834 (omp_clause_code_name): Likewise.
3836 2018-06-20 Jakub Jelinek <jakub@redhat.com>
3839 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
3842 PR tree-optimization/86231
3843 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
3844 anti-range don't overwrite *vr0min before using it to compute *vr0max.
3846 2018-06-20 Tom de Vries <tdevries@suse.de>
3848 PR tree-optimization/86097
3849 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
3850 iv type if signedness of iv type is not the same as that of *nit.
3852 2018-06-20 Jakub Jelinek <jakub@redhat.com>
3854 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
3855 EDGE_EH edges, verify they are all EDGE_EH.
3857 2018-06-20 Maya Rashish <coypu@sdf.org>
3859 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
3861 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3863 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
3864 * config/aarch64/aarch64.c (xgene1_tunings): Add
3865 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
3866 (aarch64_mode_valid_for_sched_fusion_p):
3867 Allow 16-byte modes.
3868 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
3869 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
3871 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
3873 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
3874 * config/aarch64/iterators.md (VQ2): New mode iterator.
3876 2018-06-20 Martin Liska <mliska@suse.cz>
3878 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
3879 Change default ratio from 10 to 8.
3881 2018-06-20 Martin Liska <mliska@suse.cz>
3883 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
3885 (bit_test_cluster::find_bit_tests): Likewise.
3886 (switch_decision_tree::analyze_switch_statement): Find clusters.
3887 * tree-switch-conversion.h (struct jump_table_cluster): Document
3890 2018-06-20 Martin Liska <mliska@suse.cz>
3892 * tree-switch-conversion.c (switch_conversion::collect):
3893 Record m_uniq property.
3894 (switch_conversion::expand): Bail out for special conditions.
3895 (group_cluster::~group_cluster): New.
3896 (group_cluster::group_cluster): Likewise.
3897 (group_cluster::dump): Likewise.
3898 (jump_table_cluster::emit): New.
3899 (switch_decision_tree::fix_phi_operands_for_edges): New.
3900 (struct case_node): Remove struct.
3901 (jump_table_cluster::can_be_handled): New.
3902 (case_values_threshold): Moved to header.
3903 (reset_out_edges_aux): Likewise.
3904 (jump_table_cluster::is_beneficial): New.
3905 (bit_test_cluster::can_be_handled): Likewise.
3906 (add_case_node): Remove.
3907 (bit_test_cluster::is_beneficial): New.
3908 (case_bit_test::cmp): New.
3909 (bit_test_cluster::emit): New.
3910 (expand_switch_as_decision_tree_p): Remove.
3911 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
3912 (fix_phi_operands_for_edge): Likewise.
3913 (switch_decision_tree::analyze_switch_statement): New.
3914 (compute_cases_per_edge): Move ...
3915 (switch_decision_tree::compute_cases_per_edge): ... here.
3916 (try_switch_expansion): Likewise.
3917 (switch_decision_tree::try_switch_expansion): Likewise.
3918 (record_phi_operand_mapping): Likewise.
3919 (switch_decision_tree::record_phi_operand_mapping): Likewise.
3920 (emit_case_decision_tree): Likewise.
3921 (switch_decision_tree::emit): Likewise.
3922 (balance_case_nodes): Likewise.
3923 (switch_decision_tree::balance_case_nodes): Likewise.
3924 (dump_case_nodes): Likewise.
3925 (switch_decision_tree::dump_case_nodes): Likewise.
3926 (emit_jump): Likewise.
3927 (switch_decision_tree::emit_jump): Likewise.
3928 (emit_cmp_and_jump_insns): Likewise.
3929 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
3930 (emit_case_nodes): Likewise.
3931 (switch_decision_tree::emit_case_nodes): Likewise.
3932 (conditional_probability): Remove.
3933 * tree-switch-conversion.h (enum cluster_type): New.
3935 (struct cluster): Likewise.
3936 (cluster::cluster): Likewise.
3937 (struct simple_cluster): Likewise.
3938 (simple_cluster::simple_cluster): Likewise.
3939 (struct group_cluster): Likewise.
3940 (struct jump_table_cluster): Likewise.
3941 (struct bit_test_cluster): Likewise.
3942 (struct min_cluster_item): Likewise.
3943 (struct case_tree_node): Likewise.
3944 (case_tree_node::case_tree_node): Likewise.
3945 (jump_table_cluster::case_values_threshold): Likewise.
3946 (struct case_bit_test): Likewise.
3947 (struct switch_decision_tree): Likewise.
3948 (struct switch_conversion): Likewise.
3949 (switch_decision_tree::reset_out_edges_aux): Likewise.
3951 2018-06-20 Martin Liska <mliska@suse.cz>
3953 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
3954 (hoist_edge_and_branch_if_true): Likewise.
3955 (expand_switch_using_bit_tests_p): Likewise.
3956 (struct case_bit_test): Likewise.
3957 (case_bit_test_cmp): Likewise.
3958 (emit_case_bit_tests): Likewise.
3959 (switch_conversion::switch_conversion): New class.
3960 (struct switch_conv_info): Remove old struct.
3961 (collect_switch_conv_info): More to ...
3962 (switch_conversion::collect): ... this.
3963 (check_range): Likewise.
3964 (switch_conversion::check_range): Likewise.
3965 (check_all_empty_except_final): Likewise.
3966 (switch_conversion::check_all_empty_except_final): Likewise.
3967 (check_final_bb): Likewise.
3968 (switch_conversion::check_final_bb): Likewise.
3969 (create_temp_arrays): Likewise.
3970 (switch_conversion::create_temp_arrays): Likewise.
3971 (free_temp_arrays): Likewise.
3972 (gather_default_values): Likewise.
3973 (switch_conversion::gather_default_values): Likewise.
3974 (build_constructors): Likewise.
3975 (switch_conversion::build_constructors): Likewise.
3976 (constructor_contains_same_values_p): Likewise.
3977 (switch_conversion::contains_same_values_p): Likewise.
3978 (array_value_type): Likewise.
3979 (switch_conversion::array_value_type): Likewise.
3980 (build_one_array): Likewise.
3981 (switch_conversion::build_one_array): Likewise.
3982 (build_arrays): Likewise.
3983 (switch_conversion::build_arrays): Likewise.
3984 (gen_def_assigns): Likewise.
3985 (switch_conversion::gen_def_assigns): Likewise.
3986 (prune_bbs): Likewise.
3987 (switch_conversion::prune_bbs): Likewise.
3988 (fix_phi_nodes): Likewise.
3989 (switch_conversion::fix_phi_nodes): Likewise.
3990 (gen_inbound_check): Likewise.
3991 (switch_conversion::gen_inbound_check): Likewise.
3992 (process_switch): Use the newly created class.
3993 (switch_conversion::expand): New.
3994 (switch_conversion::~switch_conversion): New.
3995 * tree-switch-conversion.h: New file.
3997 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3999 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
4000 tree-vect-patterns.c.
4001 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
4002 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
4003 (vect_recog_sad_pattern): Likewise.
4004 (vect_recog_widen_sum_pattern): Likewise.
4005 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
4006 (vect_recog_widen_shift_pattern): Remove the type_in argument.
4007 (vect_recog_rotate_pattern): Likewise.
4008 (vect_recog_mult_pattern): Likewise.
4009 (vect_recog_vector_vector_shift_pattern): Likewise.
4010 (vect_recog_divmod_pattern): Likewise.
4011 (vect_recog_mixed_size_cond_pattern): Likewise.
4012 (vect_recog_bool_pattern): Likewise.
4013 (vect_recog_mask_conversion_pattern): Likewise.
4014 (vect_try_gather_scatter_pattern): Likewise.
4015 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
4016 (vect_recog_over_widening_pattern): Likewise.
4017 (vect_recog_gather_scatter_pattern): Likewise.
4018 (vect_recog_func_ptr): Move from tree-vectorizer.h
4019 (vect_vect_recog_func_ptrs): Move further down the file.
4020 (vect_recog_func): Likewise. Remove the third argument.
4021 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
4022 (vect_pattern_recog_1): Expect the pattern function to do any
4023 necessary target tests. Also expect it to provide a vector type.
4024 Remove the type_in handling.
4026 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4028 * tree-vect-patterns.c (vect_pattern_detected): New function.
4029 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
4030 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
4031 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
4032 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
4033 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
4034 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
4035 (vect_recog_mask_conversion_pattern)
4036 (vect_try_gather_scatter_pattern): Likewise.
4038 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4040 * tree-vect-patterns.c (vect_get_internal_def): New function.
4041 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
4042 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
4043 (search_type_for_mask_1): Use it.
4045 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4047 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
4048 redundant WIDEN_SUM_EXPR handling.
4049 (vect_recog_sad_pattern): Likewise.
4051 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4053 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
4054 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
4055 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
4056 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
4057 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
4059 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4061 * tree-vect-stmts.c (vectorizable_call): Make sure that we
4062 use the stmt_vec_info of the original bb statement for the
4063 new zero assignment, even if the call is part of a pattern.
4065 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4067 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
4068 that the sequence is attached to the original statement rather
4069 than the pattern statement.
4070 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
4071 PATTERN_DEF_SEQ from the original statement rather than
4072 the main pattern statement.
4073 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
4074 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
4075 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
4077 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4079 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
4080 definition statements before the early exit for statements that aren't
4082 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
4084 (vect_transform_loop): ...here. Process pattern definition
4085 statements without first checking whether the main pattern
4086 statement is live or relevant.
4088 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
4090 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
4091 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
4093 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
4095 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
4096 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
4097 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
4098 (expand_block_compare): Change select_block_compare_mode call.
4099 (expand_strncmp_align_check): Use new functions, fix comment.
4100 (emit_final_str_compare_gpr): New function.
4101 (expand_strn_compare): Refactor and clean up code.
4102 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
4104 2018-06-19 Tony Reix <tony.reix@atos.com>
4105 Damien Bergamini <damien.bergamini@atos.com>
4106 David Edelsohn <dje.gcc@gmail.com>
4108 * collect2.c (static_obj): New variable.
4109 (static_libs): New variable.
4110 (is_in_list): Uncomment declaration.
4111 (main): Track AIX libraries linked statically.
4112 (is_in_list): Uncomment definition.
4113 (scan_prog_file): Don't add AIX shared libraries initializer
4114 to constructor list if linking statically.
4116 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
4118 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
4120 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
4122 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
4124 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
4127 2018-06-19 Martin Liska <mliska@suse.cz>
4129 * config/i386/i386.c (ix86_can_inline_p): Do not use
4130 ipa_fn_summaries::get_create.
4131 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
4133 (devirtualization_time_bonus): Likewise.
4134 (ipcp_propagate_stage): Likewise.
4135 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
4136 (edge_set_predicate): Likewise.
4137 (evaluate_conditions_for_known_args): Likewise.
4138 (evaluate_properties_for_edge): Likewise.
4139 (ipa_call_summary::reset): Tranform to ...
4140 (ipa_call_summary::~ipa_call_summary): ... this.
4141 (ipa_fn_summary::reset): Transform to ...
4142 (ipa_fn_summary::~ipa_fn_summary): ... this.
4143 (ipa_fn_summary_t::remove): Rename to ...
4144 (ipa_fn_summary_t::remove_callees): ... this.
4145 (ipa_fn_summary_t::duplicate): Use placement new
4146 instead of memory copy.
4147 (ipa_call_summary_t::duplicate): Likewise.
4148 (ipa_call_summary_t::remove): Remove.
4149 (dump_ipa_call_summary): Change get_create to get.
4150 (ipa_dump_fn_summary): Dump only when summary exists.
4151 (analyze_function_body): Use symbol_summary::get instead
4153 (compute_fn_summary): Likewise.
4154 (estimate_edge_devirt_benefit): Likewise.
4155 (estimate_edge_size_and_time): Likewise.
4156 (inline_update_callee_summaries): Likewise.
4157 (remap_edge_change_prob): Likewise.
4158 (remap_edge_summaries): Likewise.
4159 (ipa_merge_fn_summary_after_inlining): Likewise.
4160 (write_ipa_call_summary): Likewise.
4161 (ipa_fn_summary_write): Likewise.
4162 (ipa_free_fn_summary): Likewise.
4163 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
4164 (struct ipa_call_summary): Likewise.
4165 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
4167 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
4168 (estimate_size_after_inlining): Likewise.
4169 (estimate_growth): Likewise.
4170 (growth_likely_positive): Likewise.
4171 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
4172 (inline_call): Likewise.
4173 * ipa-inline.c (caller_growth_limits): Likewise.
4174 (can_inline_edge_p): Likewise.
4175 (can_inline_edge_by_limits_p): Likewise.
4176 (compute_uninlined_call_time): Likewise.
4177 (compute_inlined_call_time): Likewise.
4178 (want_inline_small_function_p): Likewise.
4179 (edge_badness): Likewise.
4180 (update_caller_keys): Likewise.
4181 (update_callee_keys): Likewise.
4182 (inline_small_functions): Likewise.
4183 (inline_to_all_callers_1): Likewise.
4184 (dump_overall_stats): Likewise.
4185 (early_inline_small_functions): Likewise.
4186 (early_inliner): Likewise.
4187 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4188 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4189 * ipa-pure-const.c (malloc_candidate_p): Likewise.
4190 * ipa-split.c (execute_split_functions): Likewise.
4191 * symbol-summary.h: Likewise.
4192 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
4194 2018-06-19 Richard Biener <rguenther@suse.de>
4196 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
4197 (vectorize_loops): ... here. Fix dbgcnt handling.
4198 (try_vectorize_loop): Wrap try_vectorize_loop_1.
4200 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
4203 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
4204 ieee128 argument takes up only one (vector) register, not two (floating
4207 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
4209 * gimplify.c (gimplify_init_constructor): Really never clear for an
4210 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
4212 2018-06-19 Richard Biener <rguenther@suse.de>
4214 PR tree-optimization/86179
4215 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
4216 after failed recognition.
4218 2018-06-18 Martin Sebor <msebor@redhat.com>
4221 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
4222 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
4223 Handle integer subtraction.
4224 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
4225 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
4227 2018-06-18 David Malcolm <dmalcolm@redhat.com>
4229 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
4230 param from rtx to rtx_insn *.
4231 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
4233 (frv_ifcvt_modify_insn): Likwise.
4234 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
4235 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
4236 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
4237 as_a <rtx_insn *> cast to local "unprotected_region" once
4238 it's been established that it's not NULL or pc_rtx.
4239 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
4240 param "sethi" from rtx to rtx_insn *.
4241 (nds32_group_float_insns): Likewise for param "insn".
4242 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
4244 (vax_output_int_subtract): Likewise.
4245 * config/vax/vax.c (vax_output_int_add): Likewise for param
4247 (vax_output_int_subtract): Likewise.
4248 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
4249 (emit_pattern_after): Likewise for param "after".
4250 (emit_insn_after): Likewise.
4251 (emit_jump_insn_after): Likewise.
4252 (emit_call_insn_after): Likewise.
4253 (emit_debug_insn_after): Likewise.
4254 (emit_pattern_before): Likewise for param "before".
4255 (emit_insn_before): Likewise.
4256 (emit_jump_insn_before): Likewise.
4257 * final.c (get_insn_template): Likewise for param "insn", removing
4259 * output.h (get_insn_template): Likewise for 2nd param.
4260 * rtl.h (emit_insn_before): Likewise.
4261 (emit_jump_insn_before): Likewise.
4262 (emit_debug_insn_before_noloc): Likewise.
4263 (emit_insn_after): Likewise.
4264 (emit_jump_insn_after): Likewise.
4265 (emit_call_insn_after): Likewise.
4266 (emit_debug_insn_after): Likewise.
4267 (set_insn_deleted): Likewise for param.
4269 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
4272 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
4273 floating point modes, so that IFmode is numerically greater than
4274 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
4275 to declare the ordering. This prevents IFmode from being
4276 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
4277 machine. Include rs6000-modes.h to share the fractional values
4278 between genmodes* and the rest of the compiler.
4282 * config/rs6000/rs6000-modes.h: New file.
4283 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
4284 meaning of rs6000_long_double_size so that 126..128 selects an
4285 appropriate 128-bit floating point type.
4286 (rs6000_option_override_internal): Likewise.
4287 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
4288 (TARGET_LONG_DOUBLE_128): Change the meaning of
4289 rs6000_long_double_size so that 126..128 selects an appropriate
4290 128-bit floating point type.
4291 (LONG_DOUBLE_TYPE_SIZE): Update comment.
4292 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
4293 source and destination to match the standard usage.
4294 (truncifkf2): Likewise.
4295 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
4296 ISA 2.07 to use an explicit clobber, instead of passing in a
4298 (copysign<mode>3_soft): Likewise.
4300 2018-06-18 David Malcolm <dmalcolm@redhat.com>
4302 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
4303 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
4304 (vect_slp_analyze_instance_dependence): Likewise.
4305 (vect_enhance_data_refs_alignment): Likewise.
4306 (vect_analyze_data_refs_alignment): Likewise.
4307 (vect_slp_analyze_and_verify_instance_alignment
4308 (vect_analyze_data_ref_accesses): Likewise.
4309 (vect_prune_runtime_alias_test_list): Likewise.
4310 (vect_analyze_data_refs): Likewise.
4311 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
4312 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
4313 (vect_analyze_scalar_cycles_1): Likewise.
4314 (vect_get_loop_niters): Likewise.
4315 (vect_analyze_loop_form_1): Likewise.
4316 (vect_update_vf_for_slp): Likewise.
4317 (vect_analyze_loop_operations): Likewise.
4318 (vect_analyze_loop): Likewise.
4319 (vectorizable_induction): Likewise.
4320 (vect_transform_loop): Likewise.
4321 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
4322 * tree-vect-slp.c (vect_analyze_slp): Likewise.
4323 (vect_make_slp_decision): Likewise.
4324 (vect_detect_hybrid_slp): Likewise.
4325 (vect_slp_analyze_operations): Likewise.
4326 (vect_slp_bb): Likewise.
4327 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
4328 (vectorizable_bswap): Likewise.
4329 (vectorizable_call): Likewise.
4330 (vectorizable_simd_clone_call): Likewise.
4331 (vectorizable_conversion): Likewise.
4332 (vectorizable_assignment): Likewise.
4333 (vectorizable_shift): Likewise.
4334 (vectorizable_operation): Likewise.
4335 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
4337 2018-06-18 Martin Sebor <msebor@redhat.com>
4339 PR tree-optimization/81384
4340 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
4341 * builtins.c (expand_builtin_strnlen): New function.
4342 (expand_builtin): Call it.
4343 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
4344 * builtins.def (BUILT_IN_STRNLEN): New.
4345 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
4346 Warn for bounds in excess of maximum object size.
4347 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
4348 single-value ranges. Handle strnlen.
4349 (handle_builtin_strlen): Handle strnlen.
4350 (strlen_check_and_optimize_stmt): Same.
4351 * doc/extend.texi (Other Builtins): Document strnlen.
4353 2018-06-18 Maya Rashish <coypu@sdf.org>
4355 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
4356 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
4357 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
4359 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
4361 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
4363 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4365 * tree.c (escaped_string::escape): Replace cast to char * by
4366 const_cast<char *> (unescaped).
4368 2018-06-18 Nick Clifton <nickc@redhat.com>
4371 * tree.c (escaped_string): New class. Converts an unescaped
4372 string into its escaped equivalent.
4373 (warn_deprecated_use): Use the new class to convert the
4374 deprecation message, if present.
4375 (test_escaped_strings): New self test.
4376 (test_c_tests): Add test_escaped_strings.
4377 * doc/extend.texi (deprecated): Add a note that the
4378 deprecation message is affected by the -fmessage-length
4379 option, and that control characters will be escaped.
4380 (#pragma GCC error): Document this pragma.
4381 (#pragma GCC warning): Likewise.
4382 * doc/invoke.texi (-fmessage-length): Document this option's
4383 effect on the #warning and #error preprocessor directives and
4384 the deprecated attribute.
4386 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
4388 * tree.c (decl_value_expr_lookup): Revert latest change.
4389 (decl_value_expr_insert): Likewise.
4391 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
4393 * gimplify.c (nonlocal_vlas): Delete.
4394 (nonlocal_vla_vars): Likewise.
4395 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
4397 (gimplify_body): Do not create and destroy nonlocal_vlas.
4398 * tree-nested.c: Include diagnostic.h.
4399 (use_pointer_in_frame): Tweak.
4400 (lookup_field_for_decl): Add assertion and declare the transformation.
4401 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
4402 internal error when the reference is in a wrong context. Do not
4403 create a debug decl by default.
4404 (note_nonlocal_block_vlas): Delete.
4405 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
4406 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
4407 create a debug decl by default.
4408 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
4409 call is in a wrong context.
4410 (fixup_vla_decls): New function.
4411 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
4412 debug variables were created.
4413 * tree.c (decl_value_expr_lookup): Add checking assertion.
4414 (decl_value_expr_insert): Likewise.
4416 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
4419 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
4420 * tree-scalar-evolution.c (interpret_expr): Likewise.
4421 (expression_expensive_p): Likewise.
4422 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
4423 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
4424 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
4425 (ssa_defined_by_minus_one_stmt_p): New.
4427 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
4430 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
4431 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
4432 * dojump.c (do_jump): Likewise.
4433 * expr.c (expand_expr_real_2): Check operand type's sign.
4434 * fold-const.c (const_unop): Handle ABSU_EXPR.
4435 (fold_abs_const): Likewise.
4436 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
4437 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
4438 (strip_sign_op_1): Likesise.
4439 * match.pd: Add new pattern to generate ABSU_EXPR.
4440 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
4441 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
4442 * tree-eh.c (operation_could_trap_helper_p): Likewise.
4443 * tree-inline.c (estimate_operator_cost): Likewise.
4444 * tree-pretty-print.c (dump_generic_node): Likewise.
4445 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
4446 * tree.def (ABSU_EXPR): New.
4448 2018-06-16 Jakub Jelinek <jakub@redhat.com>
4451 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
4452 documented as preserved for backward compatibility only.
4453 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
4455 PR rtl-optimization/86108
4456 * bb-reorder.c (create_forwarder_block): Renamed to ...
4457 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
4458 jump from new landing pad to the second part.
4459 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
4462 2018-06-15 Jakub Jelinek <jakub@redhat.com>
4465 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
4466 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
4467 Only call store_expr for halves if the mode is the same.
4470 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
4471 Fix up comment formatting.
4473 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
4475 * typed-splay-tree.h (typed_splay_tree::remove): New function.
4476 (typed_splay_tree::closure,
4477 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
4478 (typed_splay_tree::typed_splay_tree,
4479 typed_splay_tree::operator =): Declared private.
4480 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
4481 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
4482 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
4483 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
4484 typed_splay_tree::splay_tree_splay,
4485 typed_splay_tree::splay_tree_foreach_helper,
4486 typed_splay_tree::splay_tree_insert,
4487 typed_splay_tree::splay_tree_remove,
4488 typed_splay_tree::splay_tree_lookup,
4489 typed_splay_tree::splay_tree_predecessor,
4490 typed_splay_tree::splay_tree_successor,
4491 typed_splay_tree::splay_tree_min,
4492 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
4493 (typed_splay_tree::root, typed_splay_tree::comp,
4494 typed_splay_tree::delete_key,
4495 typed_splay_tree::delete_value): New data members.
4496 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
4497 typed_splay_tree::remove.
4499 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
4501 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
4502 -mginv and -mno-ginv to the assembler.
4503 * config/mips/mips.opt (-mcrc): New option.
4505 * doc/invoke.text (-mcrc): Document.
4508 2018-06-15 Nick Clifton <nickc@redhat.com>
4511 * tree.c (escaped_string): New class. Converts an unescaped
4512 string into its escaped equivalent.
4513 (warn_deprecated_use): Use the new class to convert the
4514 deprecation message, if present.
4515 (test_escaped_strings): New self test.
4516 (test_c_tests): Add test_escaped_strings.
4517 * doc/extend.texi (deprecated): Add a note that the
4518 deprecation message is affected by the -fmessage-length
4519 option, and that control characters will be escaped.
4520 (#pragma GCC error): Document this pragma.
4521 (#pragma GCC warning): Likewise.
4522 * doc/invoke.texi (-fmessage-length): Document this option's
4523 effect on the #warning and #error preprocessor directives and
4524 the deprecated attribute.
4526 2018-06-15 Richard Biener <rguenther@suse.de>
4528 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
4529 here, also noting vector size used.
4530 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
4531 size used in MSG_OPTIMIZED_LOCATIONS dump.
4532 (pass_slp_vectorize::execute): Adjust.
4534 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
4537 * config/arc/arc.c (arc_return_address_register): Fix
4540 2018-06-15 Richard Biener <rguenther@suse.de>
4543 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
4544 leave useless conversion stripping to force_gimple_operand_gsi.
4545 (gimplify_build2): Likewise.
4546 (gimplify_build1): Likewise.
4548 2018-06-15 Richard Biener <rguenther@suse.de>
4551 * tree-cfg.c (move_stmt_op): unshare invariant addresses
4552 before adjusting their block.
4554 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
4556 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
4557 multilibs for *-*-rtems*.
4558 * config/riscv/t-rtems: New file.
4560 2018-06-14 Jakub Jelinek <jakub@redhat.com>
4563 * match.pd ((A +- CST1) +- CST2): Punt if last resort
4564 unsigned_type_for returns NULL.
4567 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
4568 subregs of multi-word pseudos unless the float mode has word size.
4570 2018-06-14 Richard Biener <rguenther@suse.de>
4573 * tree-vect-generic.c (build_word_mode_vector_type): Remove
4574 duplicate and harmful type_hash_canon.
4575 * tree.c (type_hash_canon): Assert we didn't find ourselves.
4577 2018-06-14 Richard Biener <rguenther@suse.de>
4580 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
4583 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
4585 * config/rtems.h (STDINT_LONG32): Define.
4587 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
4588 Prachi Godbole <prachi.godbole@imgtec.com>
4590 * config/mips/mips-cpus.def: Define P6600.
4591 * config/mips/mips-tables.opt: Regenerate.
4592 * config/mips/mips.c (mips_ucbranch_type): New enum.
4593 (mips_rtx_cost_data): Add support for P6600.
4594 (mips_issue_rate): Likewise.
4595 (mips_multipass_dfa_lookahead): Likewise.
4596 (mips_avoid_hazard): Likewise.
4597 (mips_reorg_process_insns): Likewise.
4598 (mips_classify_branch_p6600): New function.
4599 * config/mips/mips.h (TUNE_P6600): New define.
4600 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
4601 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
4602 * config/mips/mips.md: Include p6600.md.
4603 (processor): Add p6600.
4604 * config/mips/p6600.md: New file.
4605 * doc/invoke.texi: Add p6600 to supported architectures.
4607 2018-06-13 Martin Sebor <msebor@redhat.com>
4609 PR tree-optimization/86114
4610 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
4612 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
4614 2018-06-13 Richard Biener <rguenther@suse.de>
4616 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
4617 Properly set vector type of the intermediate stmt.
4618 * tree-vect-stmts.c (vectorizable_operation): The destination
4619 var always has vectype_out type.
4621 2018-06-13 Jeff Law <law@redhat.com>
4623 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
4624 integer 0 for argument to print_rtl_with_bb.
4625 (rl78_reorg): Likewise.
4627 2018-06-13 David Malcolm <dmalcolm@redhat.com>
4629 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
4630 from rtx to rtx_insn *.
4631 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
4633 (add_sched_insns_for_speculation): Likewise for local "target",
4634 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
4635 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
4636 from rtx_insn ** to rtx_code_label **.
4637 (reorg_emit_nops): Likewise.
4638 (c6x_reorg): Likewise for local "call_labels".
4639 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
4641 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
4642 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
4643 the loops over LABEL_REFS.
4644 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
4646 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
4647 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
4648 (split_branches): Strengthen local "olabel" from rtx to
4649 rtx_insn *, adding a safe_as_a cast.
4650 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
4652 (add_insn_after): Likewise for first two params.
4653 (add_insn_before): Likewise.
4654 (remove_insn): Likewise for param.
4655 (emit_pattern_before_noloc): Likewise for second and third params.
4656 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
4657 (emit_call_insn_before_noloc): Likewise.
4658 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
4660 (emit_barrier_before): Likewise.
4661 (emit_label_before): Strengthen "label" param from "rtx" to
4662 "rtx_code_label *". Strengthen "before" param from "rtx" to
4664 (emit_insn_after_1): Strengthen "after" param from "rtx" to
4666 (emit_pattern_after_noloc): Likewise.
4667 (emit_insn_after_noloc): Likewise.
4668 (emit_jump_insn_after_noloc): Likewise.
4669 (emit_call_insn_after_noloc): Likewise.
4670 (emit_debug_insn_after_noloc): Likewise.
4671 (emit_barrier_after): Likewise.
4672 (emit_label_after): Likewise for both params.
4673 (emit_pattern_after_setloc): Likewise for "after" param. Convert
4674 "loc" param from "int" to "location_t".
4675 (emit_insn_after_setloc): Likewise.
4676 (emit_jump_insn_after_setloc): Likewise.
4677 (emit_call_insn_after_setloc): Likewise.
4678 (emit_debug_insn_after_setloc): Likewise.
4679 (emit_pattern_before_setloc): Likewise for "before" param. Convert
4680 "loc" param from "int" to "location_t".
4681 (emit_pattern_before): Convert NULL_RTX to NULL.
4682 (emit_insn_before_setloc): Convert "loc" param from "int" to
4684 (emit_jump_insn_before_setloc): Likewise.
4685 (emit_call_insn_before_setloc): Likewise.
4686 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
4687 rtx_insn *. Convert "loc" param from "int" to "location_t".
4688 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
4689 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
4690 Convert 3rd param from "int" to "location_t".
4691 (emit_barrier_before, emit_barrier_after, next_real_insn):
4692 Strengthen param from rtx to rtx_insn *.
4693 (emit_label_before): Strengthen 1st param from "rtx" to
4694 "rtx_code_label *". Strengthen 2nd param from "rtx" to
4696 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
4697 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
4698 Strengthen 2nd param from "rtx" to "rtx_insn *".
4699 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
4700 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
4701 Likewise. Convert 3rd param from "int" to "location_t".
4702 (emit_label_after): Strengthen 1st param from "rtx" to
4704 (next_real_insn, remove_insn): Strengthen param from "rtx" to
4706 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
4707 from "rtx" to "rtx_insn *".
4709 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
4711 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
4712 bodies streamed in with -Q.
4713 * dumpfile.c (dump_files): Add lto-stream-out dump file.
4714 * dumpfile.h (tree_dump_index): Add lto_stream_out.
4715 * gimple-streamer-out.c: Include gimple-pretty-print.h
4716 (output_bb): Dump stmts streamed.
4717 * lto-section-out.c: Include print-tree.h
4718 (lto_begin_section): Dump sections created.
4719 (lto_output_decl_index): Dump decl encoded.
4720 * lto-streamer-out.c: Include print-tree.h
4721 (create_output_block): Dump output block created.
4722 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
4723 (output_function): Dump function output.
4724 (output_constructor): Dump constructor streamed.
4725 (write_global_stream): Output indexes encoded.
4726 (produce_asm_for_decls): Dump streams encoded.
4727 * lto-streamer.c (streamer_dump_file): New global var.
4728 * lto-streamer.h (streamer_dump_file): Declare.
4729 * passes.c (ipa_write_summaries): Initialize streamer dump.
4730 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
4733 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
4736 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
4737 offsets for register save directives. Emit a second batch of save
4738 directives, if need be, when the function accesses prior frames.
4740 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4742 * config/arc/fpu.md (fmasf4): Force operand to register.
4743 (fnmasf4): Likewise.
4745 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4747 * config/arc/arc-protos.h (arc_pad_return): Remove.
4748 * config/arc/arc.c (machine_function): Remove force_short_suffix
4750 (arc_print_operand): Adjust printing of '&'.
4751 (arc_verify_short): Remove conditional printing of short suffix.
4752 (arc_final_prescan_insn): Remove reference to size_reason.
4753 (pad_return): New function.
4754 (arc_reorg): Call pad_return.
4755 (arc_pad_return): Remove.
4756 (arc_init_machine_status): Remove reference to force_short_suffix.
4757 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
4758 (attr length): When attribute iscompact is true force to 2
4759 regardless; in the case of maybe check if we want to force the
4760 instruction to have 4 bytes length.
4761 (nopv): Change it to generate 4 byte long nop as well.
4762 (blockage): New pattern.
4763 (simple_return): Remove call to arc_pad_return.
4764 (p_return_i): Likewise.
4766 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4768 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
4770 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4772 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
4775 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4777 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
4778 for ARC700 and ARCv2.
4780 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
4783 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
4784 operands[2] instead of operands[1].
4787 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
4789 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
4790 case, check whether the outer register overlaps an unallocatable
4791 register, not just whether it fits the required class.
4793 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
4795 * poly-int.h (can_div_trunc_p): Add new overload in which all values
4797 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
4798 (memrefs_conflict_p): Likewise.
4799 (init_alias_analysis): Likewise.
4800 * cfgexpand.c (expand_debug_expr): Likewise.
4801 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
4802 * cse.c (fold_rtx): Likewise.
4803 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
4804 * expr.c (emit_block_move_hints): Likewise.
4805 (clear_storage_hints, push_block, emit_push_insn): Likewise.
4806 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
4807 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
4808 (emit_group_store): Likewise.
4809 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
4810 to read the PRE/POST_MODIFY increment.
4811 * calls.c (store_one_arg): Use strip_offset.
4812 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
4814 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
4816 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
4817 (simplify_binary_operation_1): Extend CONST_INT handling to
4819 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
4820 than a HOST_WIDE_INT.
4821 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
4823 (adjust_mems, add_stores): Update accodingly.
4824 (vt_canonicalize_addr): Track polynomial offsets.
4825 (emit_note_insn_var_location): Likewise.
4826 (vt_add_function_parameter): Likewise.
4827 (vt_initialize): Likewise.
4829 2018-06-12 Jeff Law <law@redhat.com>
4831 * config.gcc (alpha*-*-freebsd*): Remove.
4832 * config/alpha/freebsd.h: Remove.
4834 2018-06-12 David Malcolm <dmalcolm@redhat.com>
4837 * spellcheck-tree.c (levenshtein_distance): Rename to...
4838 (get_edit_distance): ...this, and update for underlying renaming.
4839 * spellcheck-tree.h (levenshtein_distance): Rename to...
4840 (get_edit_distance): ...this.
4841 * spellcheck.c (levenshtein_distance): Rename to...
4842 (get_edit_distance): ...this. Convert from Levenshtein distance
4843 to Damerau-Levenshtein distance by supporting transpositions of
4844 adjacent characters. Rename "v1" to "v_next" and "v0" to
4846 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
4847 (selftest::test_edit_distance_unit_test_oneway): ...this, and
4848 update for underlying renaming.
4849 (selftest::levenshtein_distance_unit_test): Rename to...
4850 (selftest::test_get_edit_distance_unit): ...this, and update for
4851 underlying renaming.
4852 (selftest::test_find_closest_string): Add example from PR 69968
4853 where transposition helps
4854 (selftest::test_metric_conditions): Update for renaming.
4855 (selftest::test_metric_conditions): Likewise.
4856 (selftest::spellcheck_c_tests): Likewise.
4857 * spellcheck.h (levenshtein_distance): Rename both overloads to...
4858 (get_edit_distance): ...this.
4859 (best_match::consider): Update for renaming.
4861 2018-06-12 Martin Sebor <msebor@redhat.com>
4863 PR tree-optimization/85259
4864 * builtins.c (compute_objsize): Handle constant offsets.
4865 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
4866 true iff a warning has been issued.
4867 * gimple.h (gimple_nonartificial_location): New function.
4868 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
4869 gimple_nonartificial_location and handle -Wno-system-headers.
4870 (handle_builtin_stxncpy): Same.
4872 2018-06-12 Martin Sebor <msebor@redhat.com>
4875 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
4877 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
4879 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
4880 BUILTIN_VEC_XST entries for pointer to double and long long.
4882 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
4885 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
4887 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
4890 2018-06-12 Martin Liska <mliska@suse.cz>
4892 * doc/options.texi: Document IntegerRange.
4894 2018-06-12 Martin Liska <mliska@suse.cz>
4896 * config/i386/i386.opt: Make MPX-related options as Deprecated.
4897 * opt-functions.awk: Handle Deprecated flag.
4898 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
4900 (read_cmdline_option): Report warning for a deprecated option.
4901 * opts.h (struct cl_option): Add new field cl_deprecated.
4902 (CL_ERR_DEPRECATED): New.
4904 2018-06-12 Martin Liska <mliska@suse.cz>
4906 * doc/options.texi: Document Deprecated option flag.
4908 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4910 * config/arc/arc-arch.h (arc_extras): New enum.
4911 (arc_cpu_t):Add field extra.
4912 (arc_cpu_types): Consider the extras.
4913 * config/arc/arc-cpus.def: Add extras info.
4914 * config/arc/arc-opts.h (processor_type): Consider extra field.
4915 * config/arc/arc.c (arc_override_options): Handle extra field.
4917 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4919 * config/arc/arc-arch.h: Update ARC_OPTX macro.
4920 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
4922 * config/arc/arc.c (arc_init): Update pic warning.
4923 (irq_range): Update irq range parsing warnings.
4924 (arc_override_options): Update various warning messages.
4925 (arc_handle_aux_attribute): Likewise.
4927 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
4929 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
4931 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4933 * doc/sourcebuild.texi: Document usage of line number 0 in verify
4934 compiler messages directives.
4936 2018-06-12 Matthew Fortune <mfortune@gmail.com>
4938 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
4939 * config/mips/mips-tables.opt: Regenerate.
4940 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
4942 * doc/invoke.texi: Document -march=i6500.
4944 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
4946 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
4947 (i6400_gpmul): Add cpu_unit.
4948 (i6400_gpdiv): Likewise.
4949 (i6400_msa_add_d): Update reservations.
4950 (i6400_msa_int_add) Likewise.
4951 (i6400_msa_short_logic3) Likewise.
4952 (i6400_msa_short_logic2) Likewise.
4953 (i6400_msa_short_logic) Likewise.
4954 (i6400_msa_move) Likewise.
4955 (i6400_msa_cmp) Likewise.
4956 (i6400_msa_short_float2) Likewise.
4957 (i6400_msa_div_d) Likewise.
4958 (i6400_msa_long_logic1) Likewise.
4959 (i6400_msa_long_logic2) Likewise.
4960 (i6400_msa_mult) Likewise.
4961 (i6400_msa_long_float2) Likewise.
4962 (i6400_msa_long_float4) Likewise.
4963 (i6400_msa_long_float5) Likewise.
4964 (i6400_msa_long_float8) Likewise.
4965 (i6400_fpu_fadd): Include frint type.
4966 (i6400_fpu_store): New define_insn_reservation.
4967 (i6400_fpu_load): Likewise.
4968 (i6400_fpu_move): Likewise.
4969 (i6400_fpu_fcmp): Likewise.
4970 (i6400_fpu_fmadd): Likewise.
4971 (i6400_int_mult): Include imul3nc type and update reservation.
4972 (i6400_int_div): Include idiv3 type and update reservation.
4973 (i6400_int_load): Update to check type not move_type.
4974 (i6400_int_store): Likewise.
4975 (i6400_int_prefetch): Set zero latency.
4977 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
4979 * gcc.c: Document new %@{...} sequence.
4980 (LINK_COMMAND_SPEC): Use it for the -L switches.
4981 (cpp_unique_options): Use it for the -I switches.
4982 (at_file_argbuf): New global variable.
4983 (in_at_file): Likewise.
4984 (alloc_args): Create at_file_argbuf.
4985 (clear_args): Truncate at_file_argbuf.
4986 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
4987 (open_at_file): New function.
4988 (close_at_file): Likewise.
4989 (create_at_file): Delete.
4990 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
4993 (validate_switches_from_spec): Deal with %@{...} sequence.
4994 (validate_switches): Likewise.
4995 (driver::finalize): Call clear_args.
4997 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
4999 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
5001 2018-06-11 Martin Sebor <msebor@redhat.com>
5003 * doc/invoke.texi (-Wall): List -Wc++17-compat.
5004 (Wno-class-memaccess): Add @opindex.
5005 (Wno-templates, Wno-multiple-inheritance): Same.
5006 (Wno-virtual-inheritance, Wno-namespaces): Same.
5007 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
5008 (Wno-format-overflow, Wno-format-truncation): Same.
5009 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
5010 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
5011 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
5012 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
5013 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
5014 (Wno-misspelled-isr): Same.
5016 2018-06-11 Martin Sebor <msebor@redhat.com>
5018 * PR tree-optimization/86083
5019 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
5021 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
5023 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
5025 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
5028 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
5029 on the correct operand.
5030 (*movdi_internal64): Ditto.
5032 2018-06-11 Martin Liska <mliska@suse.cz>
5034 PR tree-optimization/86089
5035 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
5037 2018-06-11 Julia Koval <julia.koval@intel.com>
5039 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
5040 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
5041 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
5043 2018-06-11 Olivier Hainque <hainque@adacore.com>
5045 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
5046 for Ada with strict dwarf2.
5048 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
5051 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
5054 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
5056 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
5058 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
5060 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
5063 2018-06-08 Martin Liska <mliska@suse.cz>
5065 * tree-cfg.h (debug_function): Fix argument type to match
5068 2018-06-08 Martin Liska <mliska@suse.cz>
5070 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
5071 Remove usage of MPX-related (and removed) fields.
5072 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
5074 2018-06-08 David Malcolm <dmalcolm@redhat.com>
5076 * cfg.c (debug): Use TDF_NONE rather than 0.
5077 * cfghooks.c (debug): Likewise.
5078 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
5079 (struct dump_option_value_info): Convert to...
5080 (struct kv_pair): ...this template type.
5081 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
5083 (optinfo_verbosity_options): Likewise.
5084 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
5086 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
5087 than int for "optgroup_flags" param.
5088 (dump_generic_expr_loc): Use dump_flags_t rather than int for
5090 (dump_dec): Likewise.
5091 (dump_finish): Use TDF_NONE rather than 0.
5092 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
5093 rather than int for "optgroup_flags" param. Use TDF_NONE rather
5094 than 0. Update for change to option_ptr.
5095 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
5096 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
5097 0. Update for changes to optinfo_verbosity_options and
5099 (opt_info_switch_p): Convert optgroup_flags from int to
5101 (dump_basic_block): Use dump_flags_t rather than int
5102 for "dump_kind" param.
5103 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
5104 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
5105 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
5106 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
5107 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
5108 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
5109 TDF_NONE): Convert from macros to...
5110 (enum dump_flag): ...this new enum.
5111 (dump_flags_t): Update to use enum.
5112 (operator|, operator&, operator~, operator|=, operator&=):
5113 Implement for dump_flags_t.
5114 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
5115 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
5116 Convert from macros to...
5117 (enum optgroup_flag): ...this new enum.
5118 (optgroup_flags_t): New typedef.
5119 (operator|, operator|=): Implement for optgroup_flags_t.
5120 (struct dump_file_info): Convert field "alt_flags" to
5121 dump_flags_t. Convert field "optgroup_flags" to
5123 (dump_basic_block): Use dump_flags_t rather than int for param.
5124 (dump_generic_expr_loc): Likewise.
5125 (dump_dec): Likewise.
5126 (dump_register): Convert param "optgroup_flags" to
5128 (opt_info_enable_passes): Likewise.
5129 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
5131 * gimple-pretty-print.c (debug): Likewise.
5132 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
5133 (merged_store_group::apply_stores): Likewise.
5134 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
5135 * gimple.c (verify_gimple_pp): Likewise.
5136 * graphite-poly.c (print_pbb_body): Likewise.
5137 * passes.c (pass_manager::register_one_dump_file): Convert
5138 local "optgroup_flags" to optgroup_flags_t.
5139 * print-tree.c (print_node): Use TDF_NONE rather than 0.
5141 (debug_body): Likewise.
5142 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
5143 to optgroup_flags_t.
5144 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
5146 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
5147 (convert_mult_to_fma): Likewise.
5148 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
5149 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
5150 * tree-vect-data-refs.c (dump_lower_bound): Convert param
5151 "dump_kind" to dump_flags_t.
5153 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
5155 * config/rs6000/rs6000.c (min, max): Delete.
5157 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
5159 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
5160 -mabi=spe and -mabi=no-spe.
5162 2018-06-08 Martin Liska <mliska@suse.cz>
5164 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
5165 where we expect an existing summary.
5167 2018-06-08 Martin Liska <mliska@suse.cz>
5169 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
5170 * ipa-inline.h (estimate_edge_growth): Likewise.
5172 2018-06-08 Martin Liska <mliska@suse.cz>
5174 * cgraph.c (function_version_hasher::hash): Use
5175 cgraph_node::get_uid ().
5176 (function_version_hasher::equal):
5177 * cgraph.h (cgraph_node::get_uid): New method.
5178 * ipa-inline.c (update_caller_keys): Use
5179 cgraph_node::get_uid ().
5180 (update_callee_keys): Likewise.
5181 * ipa-utils.c (searchc): Likewise.
5182 (ipa_reduced_postorder): Likewise.
5183 * lto-cgraph.c (input_node): Likewise.
5184 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
5185 * symbol-summary.h (symtab_insertion): Likewise.
5186 (symtab_removal): Likewise.
5187 (symtab_duplication): Likewise.
5188 * tree-pretty-print.c (dump_function_header): Likewise.
5189 * tree-sra.c (convert_callers_for_node): Likewise.
5191 2018-06-08 Martin Liska <mliska@suse.cz>
5193 * cgraph.c (symbol_table::create_edge): Always assign a new
5195 (symbol_table::free_edge): Do not recycle numbers.
5196 * cgraph.h (cgraph_edge::get): New method.
5197 * symbol-summary.h (symtab_removal): Use it.
5198 (symtab_duplication): Likewise.
5199 (call_summary::hashable_uid): Remove.
5201 2018-06-08 Martin Liska <mliska@suse.cz>
5203 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
5204 (initialize_growth_caches): Remove.
5205 (free_growth_caches): Likewise.
5206 (do_estimate_edge_time): Use edge_growth_cache.
5207 (do_estimate_edge_size): Likewise.
5208 (do_estimate_edge_hints): Likewise.
5209 * ipa-inline.c (reset_edge_caches): Likewise.
5210 (recursive_inlining): Likewise.
5211 (inline_small_functions): Likewise.
5212 * ipa-inline.h (initialize_growth_caches): Remove.
5213 (estimate_edge_size): Likewise.
5214 (estimate_edge_time): Likewise.
5215 (estimate_edge_hints): Likewise.
5216 (reset_edge_growth_cache): Likewise.
5217 * symbol-summary.h (call_summary::remove): New method.
5219 2018-06-08 Martin Liska <mliska@suse.cz>
5221 * ipa-cp.c (class edge_clone_summary): New summary.
5222 (grow_edge_clone_vectors): Remove.
5223 (ipcp_edge_duplication_hook): Remove.
5224 (class edge_clone_summary_t): New call_summary class.
5225 (ipcp_edge_removal_hook): Remove.
5226 (edge_clone_summary_t::duplicate): New function.
5227 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
5228 (create_specialized_node): Likewise.
5229 (ipcp_driver): Initialize edge_clone_summaries and do not
5232 2018-06-08 Martin Liska <mliska@suse.cz>
5234 * symbol-summary.h (get): New function.
5235 (call_summary::m_initialize_when_cloning): New class member.
5237 2018-06-08 Martin Liska <mliska@suse.cz>
5239 * cgraph.c (cgraph_node::remove): Do not recycle uid.
5240 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
5241 (symbol_table::allocate_cgraph_symbol): Do not set uid.
5242 * passes.c (uid_hash_t): Record removed_nodes by their uids.
5243 (remove_cgraph_node_from_order): Use the removed_nodes set.
5244 (do_per_function_toporder): Likwise.
5245 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
5246 instead of summary_uid.
5247 (symtab_removal): Likewise.
5248 (symtab_duplication): Likewise.
5250 2018-06-08 Martin Liska <mliska@suse.cz>
5252 * ipa-cp.c (ipcp_store_bits_results): Use
5253 ipcp_transformation_sum.
5254 (ipcp_store_vr_results): Likewise.
5255 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
5257 (ipcp_transformation_initialize): ... this.
5258 (ipa_set_node_agg_value_chain):
5259 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
5260 (write_ipcp_transformation_info): Likewise.
5261 (read_ipcp_transformation_info): Likewise.
5262 (ipcp_update_bits): Likewise.
5263 (ipcp_update_vr): Likewise.
5264 (ipcp_transform_function): Likewise.
5265 * ipa-prop.h: Rename ipcp_transformation_summary to
5266 ipcp_transformation.
5267 (class ipcp_transformation_t): New function summary.
5268 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
5269 (ipa_get_agg_replacements_for_node): Likewise.
5271 2018-06-08 Martin Liska <mliska@suse.cz>
5273 * ipa-pure-const.c (struct funct_state_d): Do it class instead
5275 (class funct_state_summary_t): New function_summary class.
5276 (has_function_state): Remove.
5277 (get_function_state): Likewise.
5278 (set_function_state): Likewise.
5279 (add_new_function): Likewise.
5280 (funct_state_summary_t::insert): New function.
5281 (duplicate_node_data): Remove.
5282 (remove_node_data): Remove.
5283 (funct_state_summary_t::duplicate): New function.
5284 (register_hooks): Create new funct_state_summaries.
5285 (pure_const_generate_summary): Use it.
5286 (pure_const_write_summary): Likewise.
5287 (pure_const_read_summary): Likewise.
5288 (propagate_pure_const): Likewise.
5289 (propagate_nothrow): Likewise.
5290 (dump_malloc_lattice): Likewise.
5291 (propagate_malloc): Likewise.
5292 (execute): Do not register hooks, just remove summary
5294 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
5297 2018-06-08 Martin Liska <mliska@suse.cz>
5299 * ipa-reference.c (remove_node_data): Remove.
5300 (duplicate_node_data): Likewise.
5301 (class ipa_ref_var_info_summary_t): New class.
5302 (class ipa_ref_opt_summary_t): Likewise.
5303 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
5304 (get_reference_optimization_summary): Use
5305 ipa_ref_opt_sum_summaries.
5306 (set_reference_vars_info): Remove.
5307 (set_reference_optimization_summary): Likewise.
5308 (ipa_init): Create summaries.
5309 (init_function_info): Use function summary.
5310 (ipa_ref_opt_summary_t::duplicate): New function.
5311 (ipa_ref_opt_summary_t::remove): New function.
5312 (get_read_write_all_from_node): Fix GNU coding style.
5313 (propagate): Use function summary.
5314 (write_node_summary_p): Fix GNU coding style.
5315 (stream_out_bitmap): Likewise.
5316 (ipa_reference_read_optimization_summary): Use function summary.
5317 (ipa_reference_c_finalize): Do not release hooks.
5319 2018-06-08 Martin Liska <mliska@suse.cz>
5321 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
5322 (analyze_function_body): Extract multiple calls of get_create.
5323 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
5324 * ipa-inline.c (recursive_inlining): Use ::get method.
5325 * ipa-inline.h (estimate_edge_growth): Likewise.
5327 2018-06-08 Martin Liska <mliska@suse.cz>
5329 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
5331 (hsa_function_summary::hsa_function_summary): Use the new enum
5333 (hsa_gpu_implementation_p): Use hsa_summaries::get.
5334 * hsa-gen.c (hsa_get_host_function): Likewise.
5335 (get_brig_function_name): Likewise.
5336 * ipa-hsa.c (process_hsa_functions): Likewise.
5337 (ipa_hsa_write_summary): Likewise.
5338 * symbol-summary.h (symtab_duplication): Use ::get function/
5339 (get): New function.
5341 2018-06-08 Martin Liska <mliska@suse.cz>
5343 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
5345 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
5346 (hsa_register_kernel): Likewise.
5347 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
5348 * hsa-gen.c (hsa_get_host_function): Likewise.
5349 (get_brig_function_name): Likewise.
5350 (generate_hsa): Likewise.
5351 (pass_gen_hsail::execute): Likewise.
5352 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
5353 (devirtualization_time_bonus): Likewise.
5354 (ipcp_propagate_stage): Likewise.
5355 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
5356 (edge_set_predicate): Likewise.
5357 (evaluate_conditions_for_known_args): Likewise.
5358 (evaluate_properties_for_edge): Likewise.
5359 (ipa_fn_summary::reset): Likewise.
5360 (ipa_fn_summary_t::duplicate): Likewise.
5361 (dump_ipa_call_summary): Likewise.
5362 (ipa_dump_fn_summary): Likewise.
5363 (analyze_function_body): Likewise.
5364 (compute_fn_summary): Likewise.
5365 (estimate_edge_devirt_benefit): Likewise.
5366 (estimate_edge_size_and_time): Likewise.
5367 (estimate_calls_size_and_time): Likewise.
5368 (estimate_node_size_and_time): Likewise.
5369 (inline_update_callee_summaries): Likewise.
5370 (remap_edge_change_prob): Likewise.
5371 (remap_edge_summaries): Likewise.
5372 (ipa_merge_fn_summary_after_inlining): Likewise.
5373 (ipa_update_overall_fn_summary): Likewise.
5374 (read_ipa_call_summary): Likewise.
5375 (inline_read_section): Likewise.
5376 (write_ipa_call_summary): Likewise.
5377 (ipa_fn_summary_write): Likewise.
5378 (ipa_free_fn_summary): Likewise.
5379 * ipa-hsa.c (process_hsa_functions): Likewise.
5380 (ipa_hsa_write_summary): Likewise.
5381 (ipa_hsa_read_section): Likewise.
5382 * ipa-icf.c (sem_function::merge): Likewise.
5383 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
5384 (do_estimate_edge_time): Likewise.
5385 (estimate_size_after_inlining): Likewise.
5386 (estimate_growth): Likewise.
5387 (growth_likely_positive): Likewise.
5388 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
5389 (inline_call): Likewise.
5390 * ipa-inline.c (caller_growth_limits): Likewise.
5391 (can_inline_edge_p): Likewise.
5392 (can_inline_edge_by_limits_p): Likewise.
5393 (compute_uninlined_call_time): Likewise.
5394 (compute_inlined_call_time): Likewise.
5395 (want_inline_small_function_p): Likewise.
5396 (edge_badness): Likewise.
5397 (update_caller_keys): Likewise.
5398 (update_callee_keys): Likewise.
5399 (recursive_inlining): Likewise.
5400 (inline_small_functions): Likewise.
5401 (inline_to_all_callers_1): Likewise.
5402 (dump_overall_stats): Likewise.
5403 (early_inline_small_functions): Likewise.
5404 (early_inliner): Likewise.
5405 * ipa-inline.h (estimate_edge_growth): Likewise.
5406 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
5407 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5408 * ipa-prop.h (IPA_NODE_REF): Likewise.
5409 (IPA_EDGE_REF): Likewise.
5410 * ipa-pure-const.c (malloc_candidate_p): Likewise.
5411 (propagate_malloc): Likewise.
5412 * ipa-split.c (execute_split_functions): Likewise.
5413 * symbol-summary.h: Rename get to get_create.
5415 (get_create): Likewise.
5416 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
5418 2018-06-08 Martin Liska <mliska@suse.cz>
5420 * symbol-summary.h (release): Move definition out of class
5422 (symtab_removal): Likewise.
5423 (symtab_duplication): Likewise.
5425 2018-06-08 Martin Liska <mliska@suse.cz>
5427 * symbol-summary.h (function_summary): Move constructor
5428 implementation out of class declaration.
5429 (release): Likewise.
5430 (symtab_insertion): Likewise.
5431 (symtab_removal): Likewise.
5432 (symtab_duplication): Likewise.
5435 2018-06-08 Martin Liska <mliska@suse.cz>
5437 * Makefile.in: Remove support for MPX (macros, related functions,
5438 fields in cgraph_node, ...).
5439 * builtin-types.def (BT_BND): Likewise.
5440 (BT_FN_BND_CONST_PTR): Likewise.
5441 (BT_FN_CONST_PTR_BND): Likewise.
5442 (BT_FN_VOID_PTR_BND): Likewise.
5443 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
5444 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
5445 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
5446 (expand_builtin_mempcpy_with_bounds): Likewise.
5447 (expand_builtin_memset_with_bounds): Likewise.
5448 (expand_builtin_memset_args): Likewise.
5449 (std_expand_builtin_va_start): Likewise.
5450 (expand_builtin): Likewise.
5451 (expand_builtin_with_bounds): Likewise.
5452 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
5453 (DEF_LIB_BUILTIN_CHKP): Likewise.
5454 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
5455 (DEF_CHKP_BUILTIN): Likewise.
5456 (BUILT_IN_MEMCPY): Likewise.
5457 (BUILT_IN_MEMMOVE): Likewise.
5458 (BUILT_IN_MEMPCPY): Likewise.
5459 (BUILT_IN_MEMSET): Likewise.
5460 (BUILT_IN_STPCPY): Likewise.
5461 (BUILT_IN_STRCAT): Likewise.
5462 (BUILT_IN_STRCHR): Likewise.
5463 (BUILT_IN_STRCPY): Likewise.
5464 (BUILT_IN_STRLEN): Likewise.
5465 (BUILT_IN_MEMCPY_CHK): Likewise.
5466 (BUILT_IN_MEMMOVE_CHK): Likewise.
5467 (BUILT_IN_MEMPCPY_CHK): Likewise.
5468 (BUILT_IN_MEMSET_CHK): Likewise.
5469 (BUILT_IN_STPCPY_CHK): Likewise.
5470 (BUILT_IN_STRCAT_CHK): Likewise.
5471 (BUILT_IN_STRCPY_CHK): Likewise.
5472 * calls.c (store_bounds): Likewise.
5473 (emit_call_1): Likewise.
5474 (special_function_p): Likewise.
5475 (maybe_warn_nonstring_arg): Likewise.
5476 (initialize_argument_information): Likewise.
5477 (finalize_must_preallocate): Likewise.
5478 (compute_argument_addresses): Likewise.
5479 (expand_call): Likewise.
5480 * cfgexpand.c (expand_call_stmt): Likewise.
5481 (expand_return): Likewise.
5482 (expand_gimple_stmt_1): Likewise.
5483 (pass_expand::execute): Likewise.
5484 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
5485 (cgraph_node::remove): Likewise.
5486 (cgraph_node::dump): Likewise.
5487 (cgraph_node::verify_node): Likewise.
5488 * cgraph.h (chkp_function_instrumented_p): Likewise.
5489 (symtab_node::get_alias_target): Likewise.
5490 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
5491 (cgraph_local_p): Likewise.
5492 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
5493 (cgraph_edge::rebuild_references): Likewise.
5494 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
5495 (walk_polymorphic_call_targets): Likewise.
5496 (cgraph_node::expand_thunk): Likewise.
5497 (symbol_table::output_weakrefs): Likewise.
5498 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
5499 (ix86_handle_option): Likewise.
5500 * config/i386/constraints.md: Likewise.
5501 * config/i386/i386-builtin-types.def (BND): Likewise.
5505 * config/i386/i386-builtin.def (BDESC_END): Likewise.
5506 (BDESC_FIRST): Likewise.
5508 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
5509 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
5510 * config/i386/i386.c (enum reg_class): Likewise.
5511 (ix86_target_string): Likewise.
5512 (ix86_option_override_internal): Likewise.
5513 (ix86_conditional_register_usage): Likewise.
5514 (ix86_valid_target_attribute_inner_p): Likewise.
5515 (ix86_set_indirect_branch_type): Likewise.
5516 (ix86_set_current_function): Likewise.
5517 (ix86_function_arg_regno_p): Likewise.
5518 (init_cumulative_args): Likewise.
5519 (ix86_function_arg_advance): Likewise.
5520 (ix86_function_arg): Likewise.
5521 (ix86_pass_by_reference): Likewise.
5522 (ix86_function_value_regno_p): Likewise.
5523 (ix86_function_value_1): Likewise.
5524 (ix86_function_value_bounds): Likewise.
5525 (ix86_return_in_memory): Likewise.
5526 (ix86_setup_incoming_vararg_bounds): Likewise.
5527 (ix86_va_start): Likewise.
5528 (indirect_thunk_need_prefix): Likewise.
5529 (print_reg): Likewise.
5530 (ix86_print_operand): Likewise.
5531 (ix86_expand_call): Likewise.
5532 (ix86_output_function_return): Likewise.
5533 (reg_encoded_number): Likewise.
5534 (BDESC_VERIFYS): Likewise.
5535 (ix86_init_mpx_builtins): Likewise.
5536 (ix86_init_builtins): Likewise.
5537 (ix86_emit_cmove): Likewise.
5538 (ix86_emit_move_max): Likewise.
5539 (ix86_expand_builtin): Likewise.
5540 (ix86_builtin_mpx_function): Likewise.
5541 (ix86_get_arg_address_for_bt): Likewise.
5542 (ix86_load_bounds): Likewise.
5543 (ix86_store_bounds): Likewise.
5544 (ix86_load_returned_bounds): Likewise.
5545 (ix86_store_returned_bounds): Likewise.
5546 (ix86_class_likely_spilled_p): Likewise.
5547 (ix86_hard_regno_mode_ok): Likewise.
5548 (x86_order_regs_for_local_alloc): Likewise.
5549 (ix86_mitigate_rop): Likewise.
5550 (ix86_bnd_prefixed_insn_p): Likewise.
5551 (ix86_mpx_bound_mode): Likewise.
5552 (ix86_make_bounds_constant): Likewise.
5553 (ix86_initialize_bounds): Likewise.
5554 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
5555 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
5556 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
5557 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
5558 (TARGET_CHKP_BOUND_MODE): Likewise.
5559 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
5560 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
5561 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
5562 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
5563 * config/i386/i386.h (TARGET_MPX): Likewise.
5564 (TARGET_MPX_P): Likewise.
5565 (VALID_BND_REG_MODE): Likewise.
5566 (FIRST_BND_REG): Likewise.
5567 (LAST_BND_REG): Likewise.
5568 (enum reg_class): Likewise.
5569 (BND_REG_P): Likewise.
5570 (BND_REGNO_P): Likewise.
5571 (BNDmode): Likewise.
5572 (ADJUST_INSN_LENGTH): Likewise.
5573 * config/i386/i386.md: Likewise.
5574 * config/i386/i386.opt: Likewise.
5575 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
5576 (defined): Likewise.
5577 (LINK_MPX): Likewise.
5578 (MPX_SPEC): Likewise.
5579 (LIBMPX_SPEC): Likewise.
5580 (LIBMPXWRAPPERS_SPEC): Likewise.
5581 (CHKP_SPEC): Likewise.
5582 * config/i386/predicates.md: Likewise.
5583 * dbxout.c (dbxout_type): Likewise.
5584 * doc/extend.texi: Likewise.
5585 * doc/invoke.texi: Likewise.
5586 * doc/md.texi: Likewise.
5587 * doc/tm.texi: Likewise.
5588 * doc/tm.texi.in: Likewise.
5589 * dwarf2out.c (is_base_type): Likewise.
5590 (gen_formal_types_die): Likewise.
5591 (gen_subprogram_die): Likewise.
5592 (gen_type_die_with_usage): Likewise.
5593 (gen_decl_die): Likewise.
5594 (dwarf2out_late_global_decl): Likewise.
5595 * expr.c (expand_assignment): Likewise.
5596 (emit_storent_insn): Likewise.
5597 (store_expr_with_bounds): Likewise.
5598 (store_expr): Likewise.
5599 (expand_expr_real_1): Likewise.
5600 * expr.h (store_expr_with_bounds): Likewise.
5601 * function.c (use_register_for_decl): Likewise.
5602 (struct bounds_parm_data): Likewise.
5603 (assign_parms_augmented_arg_list): Likewise.
5604 (assign_parm_find_entry_rtl): Likewise.
5605 (assign_parm_is_stack_parm): Likewise.
5606 (assign_parm_load_bounds): Likewise.
5607 (assign_bounds): Likewise.
5608 (assign_parms): Likewise.
5609 (expand_function_start): Likewise.
5610 * gcc.c (CHKP_SPEC): Likewise.
5611 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
5612 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
5613 (wrestrict_dom_walker::check_call): Likewise.
5614 * gimple.c (gimple_build_call_from_tree): Likewise.
5615 * gimple.h (enum gf_mask): Likewise.
5616 (gimple_call_with_bounds_p): Likewise.
5617 (gimple_call_set_with_bounds): Likewise.
5618 * gimplify.c (gimplify_init_constructor): Likewise.
5619 * ipa-cp.c (initialize_node_lattices): Likewise.
5620 (propagate_constants_across_call): Likewise.
5621 (find_more_scalar_values_for_callers_subset): Likewise.
5622 * ipa-hsa.c (process_hsa_functions): Likewise.
5623 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
5624 * ipa-icf.c (sem_function::merge): Likewise.
5625 * ipa-inline.c (early_inliner): Likewise.
5626 * ipa-pure-const.c (warn_function_noreturn): Likewise.
5627 (warn_function_cold): Likewise.
5628 (propagate_pure_const): Likewise.
5629 * ipa-ref.h (enum GTY): Likewise.
5630 * ipa-split.c (find_retbnd): Likewise.
5631 (consider_split): Likewise.
5632 (split_function): Likewise.
5633 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
5634 * ipa.c (walk_polymorphic_call_targets): Likewise.
5635 (symbol_table::remove_unreachable_nodes): Likewise.
5636 (process_references): Likewise.
5637 (cgraph_build_static_cdtor_1): Likewise.
5638 * lto-cgraph.c (lto_output_node): Likewise.
5639 (output_refs): Likewise.
5640 (compute_ltrans_boundary): Likewise.
5641 (input_overwrite_node): Likewise.
5642 (input_node): Likewise.
5643 (input_cgraph_1): Likewise.
5644 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
5645 * passes.c (pass_manager::execute_early_local_passes): Likewise.
5646 (class pass_chkp_instrumentation_passes): Likewise.
5647 (make_pass_chkp_instrumentation_passes): Likewise.
5648 * passes.def: Likewise.
5649 * rtl.h (struct GTY): Likewise.
5650 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
5651 * stor-layout.c (layout_type): Likewise.
5652 * symtab.c: Likewise.
5653 * target.def: Likewise.
5654 * targhooks.c (default_chkp_bound_type): Likewise.
5655 (default_chkp_bound_mode): Likewise.
5656 (default_builtin_chkp_function): Likewise.
5657 (default_chkp_function_value_bounds): Likewise.
5658 (default_chkp_make_bounds_constant): Likewise.
5659 (default_chkp_initialize_bounds): Likewise.
5660 * targhooks.h (default_chkp_bound_type): Likewise.
5661 (default_chkp_bound_mode): Likewise.
5662 (default_builtin_chkp_function): Likewise.
5663 (default_chkp_function_value_bounds): Likewise.
5664 (default_chkp_make_bounds_constant): Likewise.
5665 (default_chkp_initialize_bounds): Likewise.
5666 * toplev.c (compile_file): Likewise.
5667 (process_options): Likewise.
5668 * tree-core.h (DEF_BUILTIN): Likewise.
5669 (DEF_BUILTIN_CHKP): Likewise.
5670 * tree-inline.c (declare_return_variable): Likewise.
5671 (remap_gimple_stmt): Likewise.
5672 (copy_bb): Likewise.
5673 (initialize_inlined_parameters): Likewise.
5674 (expand_call_inline): Likewise.
5675 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
5676 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
5677 (make_pass_ipa_chkp_produce_thunks): Likewise.
5678 (make_pass_chkp): Likewise.
5679 (make_pass_chkp_opt): Likewise.
5680 (make_pass_chkp_instrumentation_passes): Likewise.
5681 * tree-pretty-print.c (dump_generic_node): Likewise.
5682 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
5683 * tree-ssa-dce.c (propagate_necessity): Likewise.
5684 (eliminate_unnecessary_stmts): Likewise.
5685 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
5686 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
5687 * tree-ssa-sccvn.h: Likewise.
5688 * tree-ssa-strlen.c (get_string_length): Likewise.
5689 (valid_builtin_call): Likewise.
5690 (adjust_last_stmt): Likewise.
5691 (handle_builtin_strchr): Likewise.
5692 (handle_builtin_strcpy): Likewise.
5693 (handle_builtin_stxncpy): Likewise.
5694 (handle_builtin_memcpy): Likewise.
5695 (handle_builtin_strcat): Likewise.
5696 (strlen_check_and_optimize_stmt): Likewise.
5697 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
5698 * tree-streamer-in.c: Likewise.
5699 * tree-streamer.c (record_common_node): Likewise.
5700 * tree.c (tree_code_size): Likewise.
5701 (wide_int_to_tree_1): Likewise.
5702 (type_contains_placeholder_1): Likewise.
5703 (build_common_tree_nodes): Likewise.
5704 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
5705 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
5706 (POINTER_BOUNDS_P): Likewise.
5707 (BOUNDED_TYPE_P): Likewise.
5708 (BOUNDED_P): Likewise.
5709 (CALL_WITH_BOUNDS_P): Likewise.
5710 (pointer_bounds_type_node): Likewise.
5711 * value-prof.c (gimple_ic): Likewise.
5712 * var-tracking.c (vt_add_function_parameters): Likewise.
5713 * varasm.c (make_decl_rtl): Likewise.
5714 (assemble_start_function): Likewise.
5715 (output_constant): Likewise.
5716 (maybe_assemble_visibility): Likewise.
5717 * varpool.c (ctor_for_folding): Likewise.
5718 * chkp-builtins.def: Remove.
5719 * ipa-chkp.c: Remove.
5720 * ipa-chkp.h: Remove.
5721 * rtl-chkp.c: Remove.
5722 * rtl-chkp.h: Remove.
5723 * tree-chkp-opt.c: Remove.
5724 * tree-chkp.c: Remove.
5725 * tree-chkp.h: Remove.
5727 2018-06-07 Carl Love <cel@us.ibm.com>
5729 * config/rs6000/vsx.md (vextract_fp_from_shorth,
5730 vextract_fp_from_shortl): Add BE support.
5732 2018-06-07 Paul Koning <ni1d@arrl.net>
5734 * compare-elim.c (try_merge_compare): Don't merge compare if
5735 address contains a side effect.
5736 (try_eliminate_compare): Likewise.
5738 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
5740 * config.gcc: Support "tremont".
5741 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
5742 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5744 * config/i386/i386.c (m_TREMONT): Define.
5745 (processor_target_table): Add "tremont".
5746 (PTA_TREMONT): Define.
5747 (ix86_lea_outperforms): Add TARGET_TREMONT.
5748 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
5749 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
5750 and M_INTEL_GOLDMONT_PLUS.
5751 (fold_builtin_cpu): Add "tremont".
5752 (ix86_add_stmt_cost): Add TARGET_TREMONT.
5753 (ix86_option_override_internal): Add "tremont".
5754 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
5755 (processor_type): Add PROCESSOR_TREMONT.
5756 * config/i386/x86-tune.def: Add m_TREMONT.
5757 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
5759 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5761 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
5762 symbol defined for msp430i* devices to be lower case.
5764 2018-06-07 Richard Biener <rguenther@suse.de>
5766 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
5767 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
5768 Properly wrap signed arithmetic if overflow wraps.
5770 2018-06-07 Jakub Jelinek <jakub@redhat.com>
5772 PR tree-optimization/69615
5773 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
5774 of a cast from a same precision integral SSA_NAME in a bb dominated
5775 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
5776 cast to utype if rhs2 has already a compatible type.
5778 2018-06-07 Richard Biener <rguenther@suse.de>
5780 PR tree-optimization/85935
5781 * graphite-scop-detection.c (find_params_in_bb): Analyze
5782 condition operands with respect to the correct loop. Assert
5783 the analysis doesn't fail.
5785 2018-06-04 Carl Love <cel@us.ibm.com>
5787 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
5788 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
5789 as it is slightly cheaper.
5790 (first_match_or_eos_index_<mode>):
5791 Calculate index using natural element order.
5792 (first_match_index_<mode>):
5793 Calculate index using natural element order.
5794 (first_match_or_eos_index_<mode>):
5795 Calculate index using natural order.
5796 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
5797 for BE and LE modes.
5798 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
5799 P9V_BUILTIN_VCLZLSBB_V16QI.
5800 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
5803 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
5805 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
5806 indentation and line wrap for many prototypes. Add missing
5807 @smallexample directives around block of prototypes for vec_xl and
5810 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
5812 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
5813 track if we pass or return IEEE 128-bit floating point.
5814 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
5815 C++ mangling that is compatible with GCC 8.1.
5816 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
5817 (init_cumulative_args): Note if we pass or return IEEE 128-bit
5818 floating point types.
5819 (rs6000_function_arg_advance_1): Likewise.
5820 (rs6000_mangle_type): Optionally generate mangled names that match
5821 what GCC 8.1 generated for IEEE 128-bit floating point types.
5822 (rs6000_globalize_decl_name): If we have an external function that
5823 passes or returns IEEE 128-bit types, generate a weak reference
5824 from the mangled name used in GCC 8.1 to the current mangled
5826 (rs6000_init_builtins): Make __ibm128 use the long double type if
5827 long double is IBM extended double. Make __float128 use the long
5828 double type if long double is IEEE 128-bit.
5831 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
5832 macro for __ibm128 built-in functions.
5833 (PACK_IF): Add __ibm128 pack/unpack functions.
5834 (UNPACK_IF): Likewise.
5835 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
5836 enable long double built-in functions if long double is IEEE
5837 128-bit floating point.
5838 (rs6000_invalid_builtin): Update long double built-in function
5840 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
5841 functions, adjust the built-in function to use the long double
5842 built-in function if __ibm128 and long double are the same type.
5843 * doc/extend.texi (PowerPC builtins): Update documention for
5844 __builtin_{,un}pack_longdouble. Add documentation for
5845 __builtin_{,un}pack_ibm128.
5847 2018-06-06 Jim Wilson <jimw@sifive.com>
5849 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
5850 (struct machine_function): New field interrupt_mode.
5851 (riscv_handle_type_attribute): New function. Add forward declaration.
5852 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
5853 (riscv_expand_epilogue): Check interrupt_mode field.
5854 (riscv_set_current_function): Check interrupt attribute args and
5855 set interrupt_mode field.
5856 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
5857 (riscv_sret, riscv_uret): New.
5858 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
5859 new arguments to interrupt attribute.
5861 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
5864 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
5865 Don't handle -mcpu=power8 if -mpower9-vector is also used.
5867 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
5869 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
5870 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
5871 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5872 several redundant entries.
5874 2018-06-06 David Malcolm <dmalcolm@redhat.com>
5876 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
5877 type from "rtx" to "rtx_insn *".
5878 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
5879 for local "call_insn", removing cast.
5880 (ix86_expand_call): Likewise, introducing a "call_insn" local.
5882 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
5884 PR tree-optimization/86066
5885 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
5886 for BIT_INSERT_EXPR stores.
5888 2018-06-06 Richard Biener <rguenther@suse.de>
5890 PR tree-optimization/86062
5891 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
5892 component refs ontop
5893 of to be offsetted base.
5895 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5897 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
5898 to be static and remove check on interrupt attribute name.
5900 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
5902 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
5903 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
5905 2018-06-05 Steve Ellcey <sellcey@cavium.com>
5908 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
5910 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
5911 Remove second argument, change how error is called.
5912 (aarch64_layout_arg): Remove second argument from
5913 aarch64_err_no_fpadvsimd call.
5914 (aarch64_init_cumulative_args): Ditto.
5915 (aarch64_gimplify_va_arg_expr): Ditto.
5916 * config/aarch64/aarch64.md (mov<mode>): Ditto.
5918 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
5920 * config/i386/i386.md (simple_return_indirect_internal): New expander.
5921 (*simple_return_indirect_internal<mode>): Rename from
5922 simple_return_indirect_internal. Use W mode iterator.
5923 (rstorssp): New expander.
5924 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
5925 (clrssbsy): New expander.
5926 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
5928 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
5930 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
5932 (cmse_check_pointed_object): Likewise.
5934 2018-06-05 Martin Liska <mliska@suse.cz>
5936 PR gcov-profile/47618
5937 * doc/invoke.texi: Document how -fprofile-dir format
5940 2018-06-05 Richard Biener <rguenther@suse.de>
5942 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
5943 removal pretend DOM info isn't available so we do not update
5944 it and only remove edges, not dominated blocks. Actually free
5945 DOM info in case we removed something. Remove unreachable blocks.
5946 (mfb_keep_latches): Work with either DOM info or marked backedges.
5947 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
5948 first. Mark backedges if DOM info isn't available.
5949 (Re-)compute DOM info after cleanup_control_flow_pre.
5951 2018-06-05 Richard Biener <rguenther@suse.de>
5953 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
5954 (locus_discrim_hasher::hash): Adjust.
5955 (locus_discrim_hasher::equal): Likewise.
5956 (next_discriminator_for_locus): Work on line directly.
5957 (same_line_p): Pass in expanded locus1 as well.
5958 (assign_discriminators): Avoid redundant location expansions.
5960 2018-06-05 Richard Biener <rguenther@suse.de>
5962 PR tree-optimization/86046
5963 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
5964 if required after clearing TREE_ADDRESSABLE.
5966 2018-06-05 Richard Biener <rguenther@suse.de>
5968 PR tree-optimization/86047
5969 * tree-ssa-loop.c (for_each_index): Glob handling of all
5970 decls and constants and really handle all of them.
5972 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5975 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
5976 qualifier_void_pointer and qualifier_const_void_pointer.
5977 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
5978 (arm_init_builtins): Handle the above.
5979 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
5980 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
5981 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
5984 2018-06-05 Martin Liska <mliska@suse.cz>
5986 * auto-profile.c (read_autofdo_file): Do not use
5987 gcov_ctr_summary struct.
5988 (afdo_callsite_hot_enough_for_early_inline): Likewise.
5989 * coverage.c (struct counts_entry): Likewise.
5990 (read_counts_file): Read just single summary entry.
5991 (get_coverage_counts): Use gcov_summary struct.
5992 * coverage.h (get_coverage_counts): Likewise.
5993 * gcov-dump.c (dump_working_sets): Likewise.
5994 (tag_summary): Dump just single summary.
5995 * gcov-io.c (gcov_write_summary): Write just histogram
5997 (gcov_read_summary): Read just single summary.
5998 (compute_working_sets): Use gcov_summary struct.
5999 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
6000 of GCOV_COUNTERS_SUMMABLE.
6001 (GCOV_COUNTERS_SUMMABLE): Remove.
6002 (GCOV_FIRST_VALUE_COUNTER): Replace with
6003 GCOV_COUNTER_V_INTERVAL.
6004 (struct gcov_ctr_summary): Remove.
6005 (struct gcov_summary): Directly use fields of former
6007 (compute_working_sets): Use gcov_summary struct.
6008 * gcov.c (read_count_file): Do not use ctrs fields.
6009 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
6011 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
6013 * profile.c: Likewise.
6014 * profile.h: Likewise.
6016 2018-06-05 Martin Liska <mliska@suse.cz>
6018 PR gcov-profile/84846
6019 * gcov.c (output_lines): Print working directory only
6020 in intermediate format.
6022 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
6024 * config/s390/s390-builtin-types.def: Add void function type.
6025 * config/s390/s390-builtins.def: Use the function type for the
6028 2018-06-04 Jim Wilson <jimw@sifive.com>
6030 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
6032 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
6033 and maybe_eh_return. Change regno to unsigned int. Use new args to
6034 handle EH_RETURN_DATA_REGNO registers properly.
6035 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
6036 (riscv_expand_epilogue): Update comment. Change argument name and
6037 type. Update code to use new name and type. Pass new args to
6038 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
6040 * config/riscv/riscv.md (NORMAL_RETURN): New.
6041 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
6042 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
6043 (eh_return): Call gen_eh_return_internal and emit barrier.
6044 (eh_return_internal): Call riscv_expand_epilogue.
6046 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
6048 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
6049 bit_insertion field and declare can_be_merged_into method.
6050 (merged_store_group::can_be_merged_into): New method.
6051 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
6052 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
6053 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
6055 2018-06-04 Richard Biener <rguenther@suse.de>
6057 PR tree-optimization/85955
6058 * builtins.c (fold_builtin_sincos): Convert pointers to
6059 destination to appropriate type before dereferencing.
6061 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
6063 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
6065 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
6067 * expr.c (expand_expr_real_1): Force the operand into memory if
6068 its TYPE_MODE is BLKmode and if there is no integer mode for
6069 the number of bits being extracted.
6071 2018-06-04 Jakub Jelinek <jakub@redhat.com>
6075 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
6076 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
6078 2018-06-04 Richard Biener <rguenther@suse.de>
6080 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
6081 (cleanup_tree_cfg_noloop): ... single caller. Do
6082 start_recording_case_labels later.
6084 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
6086 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
6087 to _IMMINTRIN_H_INCLUDED.
6088 * config/i386/pconfigintrin.h: Ditto.
6089 * config/i386/waitpkgintrin.h: Ditto.
6090 * config/i386/immintrin.h: Add includes for sgxintrin.h,
6091 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
6092 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
6093 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
6094 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
6095 waitpkgintrin.h and cldemoteintrin.h.
6097 2018-06-04 Richard Biener <rguenther@suse.de>
6099 PR tree-optimization/86038
6100 * tracer.c (find_best_successor): Check probability for
6101 being initialized, bail out if not.
6103 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
6106 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
6107 of bits to ignore when comparing architectures.
6109 2018-06-04 Jakub Jelinek <jakub@redhat.com>
6111 PR tree-optimization/69615
6112 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
6113 maximum or minimum of the type, try to merge it also as if
6114 range1 is + [-, x - 1] or + [x + 1, -].
6117 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
6119 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
6121 PR tree-optimization/86034
6122 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
6123 the unsigned bitfield type in a bit insertion sequence if it does not
6124 have a larger precision than the bitfield size.
6125 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
6127 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
6129 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
6131 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
6133 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
6134 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
6135 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
6136 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
6138 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
6140 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
6141 Disable -fdelete-null-pointer-checks for ELF toolchain.
6143 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
6144 Kito Cheng <kito.cheng@gmail.com>
6146 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
6147 (nds32le-*-*, nds32be-*-*): Integrate checking process.
6148 (nds32*-*-*): Add glibc and uclibc conditions.
6149 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
6150 (TARGET_EXCEPT_UNWIND_INFO): Define.
6151 * config/nds32/elf.h: New file.
6152 * config/nds32/linux.h: New file.
6153 * config/nds32/nds32-elf.opt: New file.
6154 * config/nds32/nds32-linux.opt: New file.
6155 * config/nds32/nds32-fp-as-gp.c
6156 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
6157 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
6159 (nds32_asm_file_end): Ditto.
6160 (nds32_print_operand): Ditto.
6161 (nds32_insert_attributes): Ditto.
6162 (nds32_init_libfuncs): New function.
6163 (TARGET_HAVE_TLS): Define.
6164 (TARGET_INIT_LIBFUNCS): Define.
6165 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
6167 (TARGET_ELF): Apply different mcmodel setting.
6168 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
6169 been migrated into elf.h and linux.h files.
6170 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
6171 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
6172 (mcmodel): The content has been migrated into nds32-elf.opt and
6173 nds32-linux.opt files.
6174 * config/nds32/t-elf: New file.
6175 * config/nds32/t-linux: New file.
6177 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
6178 Shiva Chen <shiva0217@gmail.com>
6180 * config/nds32/constants.md (unspec_volatile_element): Add
6181 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
6182 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
6184 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
6185 (make_pass_nds32_fp_as_gp): Declare.
6186 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
6188 (nds32_asm_function_end_prologue): Remove unused asm output.
6189 (nds32_asm_function_begin_epilogue): Remove unused asm output.
6190 (nds32_asm_file_start): Output necessary fp_as_gp information.
6191 (nds32_option_override): Adjust register usage.
6192 (nds32_expand_prologue): Consider fp_as_gp situation.
6193 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
6194 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
6197 (simple_return): Ditto.
6198 (omit_fp_begin): Output special directive for fp_as_gp.
6199 (omit_fp_end): Output special directive for fp_as_gp.
6200 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
6201 mforbid-fp-as-gp): New options.
6203 2018-06-01 Mark Wielaard <mark@klomp.org>
6205 * dwarf2out.c (dwarf2out_finish): Remove generation of
6206 DW_AT_loclists_base.
6208 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
6210 * gimple-ssa-store-merging.c: Include gimple-fold.h.
6211 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
6212 (struct merged_store_group): Add bit_insertion field.
6213 (dump_char_array): Use standard hexadecimal format.
6214 (merged_store_group::merged_store_group): Set bit_insertion to false.
6215 (merged_store_group::apply_stores): Use optimal buffer size. Deal
6216 with BIT_INSERT_EXPR stores. Move up code updating the mask and
6217 also print the mask in the dump file.
6218 (pass_store_merging::gate): Minor tweak.
6219 (imm_store_chain_info::coalesce_immediate): Fix wrong association
6220 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
6221 stores with INTEGER_CST stores.
6222 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
6223 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
6224 and use it throughout. Generate bit insertion sequences if need be.
6225 (pass_store_merging::process_store): Remove redundant condition.
6226 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
6228 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
6230 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
6231 the 128-bit floating point types. Fix function comment.
6233 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6235 * config/aarch64/aarch64-simd.md
6236 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
6238 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
6241 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
6243 PR tree-optimization/85989
6244 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
6246 (backprop::intersect_uses): Check it when deciding whether this
6247 is a backedge reference.
6248 (backprop::process_block): Add each phi to m_visited_phis
6249 after visiting it, then clear it at the end.
6251 2018-06-01 Richard Biener <rguenther@suse.de>
6253 * tree-vectorizer.h (vect_dr_stmt): New function.
6254 (vect_get_load_cost): Adjust.
6255 (vect_get_store_cost): Likewise.
6256 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
6257 Use vect_dr_stmt instead of DR_SMTT.
6258 (vect_record_base_alignments): Likewise.
6259 (vect_calculate_target_alignment): Likewise.
6260 (vect_compute_data_ref_alignment): Likewise and make static.
6261 (vect_update_misalignment_for_peel): Likewise.
6262 (vect_verify_datarefs_alignment): Likewise.
6263 (vector_alignment_reachable_p): Likewise.
6264 (vect_get_data_access_cost): Likewise. Pass down
6265 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
6266 (vect_get_peeling_costs_all_drs): Likewise.
6267 (vect_peeling_hash_get_lowest_cost): Likewise.
6268 (vect_enhance_data_refs_alignment): Likewise.
6269 (vect_find_same_alignment_drs): Likewise.
6270 (vect_analyze_data_refs_alignment): Likewise.
6271 (vect_analyze_group_access_1): Likewise.
6272 (vect_analyze_group_access): Likewise.
6273 (vect_analyze_data_ref_access): Likewise.
6274 (vect_analyze_data_ref_accesses): Likewise.
6275 (vect_vfa_segment_size): Likewise.
6276 (vect_small_gap_p): Likewise.
6277 (vectorizable_with_step_bound_p): Likewise.
6278 (vect_prune_runtime_alias_test_list): Likewise.
6279 (vect_analyze_data_refs): Likewise.
6280 (vect_supportable_dr_alignment): Likewise.
6281 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
6282 (vect_gen_prolog_loop_niters): Likewise.
6283 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6284 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
6286 (vect_recog_mask_conversion_pattern): Likewise.
6287 (vect_try_gather_scatter_pattern): Likewise.
6288 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
6289 to vect_get_store_cost.
6290 (vect_get_store_cost): Get stmt_info instead of DR.
6291 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
6292 (vect_get_load_cost): Get stmt_info instead of DR.
6294 2018-06-01 Richard Biener <rguenther@suse.de>
6297 * gimple-fold.c (var_decl_component_p): Also allow offsetted
6298 vars wrapped in MEM_REFs.
6300 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
6302 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
6303 Fix subreg tests so that we only return a choice between
6304 GENERAL_REGS and FP_REGS if the original classes included both.
6306 2018-06-01 Richard Biener <rguenther@suse.de>
6309 * tree-ssa-structalias.c (get_function_part_constraint):
6310 Handle NULL fi->decl.
6311 (find_func_aliases_for_call): Properly handle indirect
6312 fi from direct call.
6313 (find_func_clobbers): Likewise.
6314 (ipa_pta_execute): Likewise.
6315 (create_variable_info_for): For functions that are ifunc_resolver
6316 resolve to a varinfo that contains the result of the resolver call.
6317 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
6320 2018-05-31 Michael Collison <michael.collison@arm.com>
6322 * config/aarch64/aarch64.md:
6323 (*fix_to_zero_extenddfdi2): New pattern.
6324 * gcc.target/aarch64/fix_extend1.c: New testcase.
6326 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
6330 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
6331 and BUILT_IN_STRNCMP_EQ.
6332 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
6333 BUILT_IN_STRNCMP_EQ.
6334 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
6335 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
6336 (gimple_fold_builtin): Likewise.
6337 * tree-ssa-strlen.c (compute_string_length): New function.
6338 (determine_min_obsize): New function.
6339 (handle_builtin_string_cmp): New function to handle calls to
6340 string compare functions.
6341 (strlen_optimize_stmt): Add handling to builtin string compare
6343 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
6344 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
6345 * tree.c (build_common_builtin_nodes): Add new defines of
6346 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
6348 2018-05-31 Jakub Jelinek <jakub@redhat.com>
6351 * bb-reorder.c (pass_partition_blocks::gate): Return false for
6352 functions with naked attribute.
6354 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
6356 * config/i386/sse.md (avx_vec_concat<mode>):
6357 Substitute concat_tg_mode mode attribute with xtg_mode.
6358 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
6359 (concat_tg_mode): Remove mode attribute.
6361 2018-05-31 Martin Sebor <msebor@redhat.com>
6364 * calls.c (alloc_max_size): Correct a logic error/typo.
6365 Treat excessive arguments as infinite. Warn for invalid arguments.
6366 * doc/invoke.texi (-Walloc-size-larger-than): Update.
6368 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
6371 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
6372 and movx for Haswell.
6374 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
6375 Cesar Philippidis <cesar@codesourcery.com>
6378 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
6379 when emitting error on private/firstprivate reductions.
6380 * omp-low.c (lower_omp_target): Avoid reference-type processing
6381 on pointers for firstprivate clause.
6383 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
6385 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
6388 * config/aarch64/aarch64-simd.md
6389 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
6390 (aarch64_ld1_x3_<mode>): Likewise
6391 (aarch64_st1x2<VALLDIF:mode>): Likewise
6392 (aarch64_st1_x2_<mode>): Likewise
6393 (aarch64_st1x3<VALLDIF:mode>): Likewise
6394 (aarch64_st1_x3_<mode>): Likewise
6395 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
6396 (vld1_s8_x3): Likewise.
6397 (vld1_u16_x3): Likewise.
6398 (vld1_s16_x3): Likewise.
6399 (vld1_u32_x3): Likewise.
6400 (vld1_s32_x3): Likewise.
6401 (vld1_u64_x3): Likewise.
6402 (vld1_s64_x3): Likewise.
6403 (vld1_f16_x3): Likewise.
6404 (vld1_f32_x3): Likewise.
6405 (vld1_f64_x3): Likewise.
6406 (vld1_p8_x3): Likewise.
6407 (vld1_p16_x3): Likewise.
6408 (vld1_p64_x3): Likewise.
6409 (vld1q_u8_x3): Likewise.
6410 (vld1q_s8_x3): Likewise.
6411 (vld1q_u16_x3): Likewise.
6412 (vld1q_s16_x3): Likewise.
6413 (vld1q_u32_x3): Likewise.
6414 (vld1q_s32_x3): Likewise.
6415 (vld1q_u64_x3): Likewise.
6416 (vld1q_s64_x3): Likewise.
6417 (vld1q_f16_x3): Likewise.
6418 (vld1q_f32_x3): Likewise.
6419 (vld1q_f64_x3): Likewise.
6420 (vld1q_p8_x3): Likewise.
6421 (vld1q_p16_x3): Likewise.
6422 (vld1q_p64_x3): Likewise.
6423 (vst1_s64_x2): Likewise.
6424 (vst1_u64_x2): Likewise.
6425 (vst1_f64_x2): Likewise.
6426 (vst1_s8_x2): Likewise.
6427 (vst1_p8_x2): Likewise.
6428 (vst1_s16_x2): Likewise.
6429 (vst1_p16_x2): Likewise.
6430 (vst1_s32_x2): Likewise.
6431 (vst1_u8_x2): Likewise.
6432 (vst1_u16_x2): Likewise.
6433 (vst1_u32_x2): Likewise.
6434 (vst1_f16_x2): Likewise.
6435 (vst1_f32_x2): Likewise.
6436 (vst1_p64_x2): Likewise.
6437 (vst1q_s8_x2): Likewise.
6438 (vst1q_p8_x2): Likewise.
6439 (vst1q_s16_x2): Likewise.
6440 (vst1q_p16_x2): Likewise.
6441 (vst1q_s32_x2): Likewise.
6442 (vst1q_s64_x2): Likewise.
6443 (vst1q_u8_x2): Likewise.
6444 (vst1q_u16_x2): Likewise.
6445 (vst1q_u32_x2): Likewise.
6446 (vst1q_u64_x2): Likewise.
6447 (vst1q_f16_x2): Likewise.
6448 (vst1q_f32_x2): Likewise.
6449 (vst1q_f64_x2): Likewise.
6450 (vst1q_p64_x2): Likewise.
6451 (vst1_s64_x3): Likewise.
6452 (vst1_u64_x3): Likewise.
6453 (vst1_f64_x3): Likewise.
6454 (vst1_s8_x3): Likewise.
6455 (vst1_p8_x3): Likewise.
6456 (vst1_s16_x3): Likewise.
6457 (vst1_p16_x3): Likewise.
6458 (vst1_s32_x3): Likewise.
6459 (vst1_u8_x3): Likewise.
6460 (vst1_u16_x3): Likewise.
6461 (vst1_u32_x3): Likewise.
6462 (vst1_f16_x3): Likewise.
6463 (vst1_f32_x3): Likewise.
6464 (vst1_p64_x3): Likewise.
6465 (vst1q_s8_x3): Likewise.
6466 (vst1q_p8_x3): Likewise.
6467 (vst1q_s16_x3): Likewise.
6468 (vst1q_p16_x3): Likewise.
6469 (vst1q_s32_x3): Likewise.
6470 (vst1q_s64_x3): Likewise.
6471 (vst1q_u8_x3): Likewise.
6472 (vst1q_u16_x3): Likewise.
6473 (vst1q_u32_x3): Likewise.
6474 (vst1q_u64_x3): Likewise.
6475 (vst1q_f16_x3): Likewise.
6476 (vst1q_f32_x3): Likewise.
6477 (vst1q_f64_x3): Likewise.
6478 (vst1q_p64_x3): Likewise.
6480 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6482 * config/msp430/msp430.c (msp430_output_labelref): Prepend
6483 user_label_prefix to name.
6485 * tree-core.h: Update comment about the format of NAME string
6486 passed to handler in attribute_spec.
6488 * config/msp430/msp430.md: Remove erroneous subreg expression from
6489 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
6490 zero_extend{q,h}isi2.
6492 2018-05-30 Borislav Petkov <bp@suse.de>
6494 * doc/extend.texi: Document some architecture specific
6495 constraints and sort entries.
6497 2018-05-30 Martin Sebor <msebor@redhat.com>
6500 * builtins.c (expand_builtin_stpcpy_1): New function.
6501 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
6502 only if the former succeeds.
6504 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
6506 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
6509 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
6511 * doc/invoke.texi (-flinker-output): Document
6513 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
6515 * passes.c (ipa_write_summaries): Only modify statements if body
6517 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
6518 incrementally linking.
6519 (ipa_passes): Likewise.
6520 * lto-cgraph.c (lto_output_node): When incrementally linking do not
6521 pass down resolution info.
6522 * common.opt (flag_incremental_link): Update info.
6523 * gcc.c (plugin specs): Turn flinker-output=* to
6524 -plugin-opt=-linker-output-known
6525 * toplev.c (compile_file): Also cut compilation when doing incremental
6527 * flag-types. (enum lto_partition_model): Add
6528 LTO_LINKER_OUTPUT_NOLTOREL.
6529 (invoke.texi): Add -flinker-output docs.
6530 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
6531 link same way as WPA; do not stream in dead initializers.
6533 * dwarf2out.c (dwarf2out_die_ref_for_decl,
6534 darf2out_register_external_decl): Support incremental link.
6536 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
6538 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
6540 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
6542 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
6543 it down to simple_object_copy_lto_debug_sections.
6544 (run_gcc): Determine incremental LTO link time and configure
6545 lto1 into non-wpa mode, disable renaming of debug sections.
6547 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
6549 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
6550 descriptions of various incorrectly documented functions.
6552 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
6555 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
6556 address check not strict.
6558 2018-05-30 Richard Biener <rguenther@suse.de>
6560 PR tree-optimization/85964
6561 * tracer.c (better_p): Drop initialized count check, we only
6562 call the function with initialized counts now.
6563 (find_best_successor): Do find a best edge if one
6564 has uninitialized count.
6565 (find_best_predecessor): Likewise. Do BB frequency check only
6566 if count is initialized.
6568 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
6570 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
6571 (aarch64_ldrstr_offset_compare): New.
6572 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
6573 load/store orderings.
6574 (aarch64_gen_adjusted_ldpstp): Likewise.
6576 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
6578 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
6579 Check for subset of GENERAL_REGS and FP_REGS.
6580 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
6583 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
6585 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
6586 and wi::to_poly_offset. Add the current offset and then check
6587 whether the sum fits, rather than using an unchecked addition of
6588 a checked term. Check for a shwi rather than a uhwi.
6589 * expr.c (get_bit_range): Use tree_to_poly_uint64.
6590 (store_constructor): Use poly_int_tree_p.
6591 (expand_expr_real_1): Likewise.
6592 * function.c (assign_temp): Likewise.
6593 * fold-const.c (const_binop): Use poly_int_tree_p and
6595 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
6597 * ipa-icf-gimple.c (func_checker::compare_operand): Use
6598 to_poly_offset for MEM offsets.
6599 * ipa-icf.c (sem_variable::equals): Likewise.
6600 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
6601 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
6602 wi::to_poly_offset for BIT_FIELD_REF offsets.
6603 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
6605 * var-tracking.c (emit_note_insn_var_location): Use
6606 tree_to_poly_uint64.
6608 2018-05-29 Jim Wilson <jimw@sifive.com>
6610 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
6612 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
6615 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
6616 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
6618 (sse4_1_round<mode>2): Use nonimmediate_operand
6619 for operand 1 predicate.
6621 2018-05-29 Martin Sebor <msebor@redhat.com>
6622 Richard Biener <rguenther@suse.de>
6625 * calls.c (get_size_range): Call determine_value_range instead
6626 of get_value_range..
6627 * tree-vrp.h (determine_value_range): Declared new function.
6628 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
6630 2018-05-29 Richard Biener <rguenther@suse.de>
6632 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
6633 sure to use non-pattern stmts for get_earlier_stmt arguments.
6634 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
6635 called on pattern stmts.
6636 (get_later_stmt): Likewise.
6638 2018-05-29 Martin Liska <mliska@suse.cz>
6640 PR gcov-profile/85759
6641 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
6644 2018-05-29 Jakub Jelinek <jakub@redhat.com>
6646 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
6648 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
6652 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
6653 VEC_PACK_FLOAT_EXPR): New tree codes.
6654 * tree-pretty-print.c (op_code_prio): Handle
6655 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
6656 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
6657 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
6658 * tree-inline.c (estimate_operator_cost): Likewise.
6659 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
6660 * fold-const.c (const_binop): Likewise.
6661 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
6662 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
6663 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6664 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
6665 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
6666 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
6667 * expr.c (expand_expr_real_2): Likewise.
6668 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
6669 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
6670 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
6672 * optabs.c (expand_widen_pattern_expr): For
6673 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
6674 sign from result type rather than operand's type.
6675 (expand_binop_directly): For vec_packu_float_optab and
6676 vec_packs_float_optab allow result type to be different from operand's
6678 * optabs-tree.c (optab_for_tree_code): Handle
6679 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
6680 VEC_PACK_FLOAT_EXPR. Formatting fixes.
6681 * tree-vect-generic.c (expand_vector_operations_1): Handle
6682 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
6683 VEC_PACK_FLOAT_EXPR.
6684 * tree-vect-stmts.c (supportable_widening_operation): Handle
6686 (supportable_narrowing_operation): Handle FLOAT_EXPR.
6687 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
6688 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
6689 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
6690 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
6692 (vec_pack<floatprefix>_float_<mode>): New expander.
6693 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
6695 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
6696 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
6697 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
6698 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
6699 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
6701 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
6702 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
6703 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
6704 VEC_PACK_FLOAT_EXPR): Document.
6706 2018-05-29 Richard Biener <rguenther@suse.de>
6708 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
6710 (stmt_vec_info_vec): Make pointer.
6711 (init_stmt_vec_info_vec): Remove.
6712 (free_stmt_vec_info_vec): Likewise.
6713 (set_stmt_vec_info_vec): New function.
6714 (free_stmt_vec_infos): Likewise.
6715 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
6716 (set_vinfo_for_stmt): Likewise.
6717 (get_earlier_stmt): Likewise.
6718 (get_later_stmt): Likewise.
6719 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
6720 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
6721 (vec_info::~vec_info): Free stmt_vec_infos.
6722 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
6723 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
6724 (pass_slp_vectorize::execute): Likewise.
6725 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
6726 (free_stmt_vec_info_vec): Likewise.
6727 (set_stmt_vec_info_vec): New function.
6728 (free_stmt_vec_infos): Likewise.
6729 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
6730 the global stmt_vec_info_vec.
6731 * tree-parloops.c (gather_scalar_reductions): Use
6732 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
6735 2018-05-29 Richard Biener <rguenther@suse.de>
6737 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
6739 2018-05-29 Martin Liska <mliska@suse.cz>
6740 David Malcolm <dmalcolm@redhat.com>
6742 * vec.c (test_reverse): New.
6743 (vec_c_tests): Add new test.
6744 * vec.h (vl_ptr>::reverse): New function.
6746 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
6748 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
6750 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
6753 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
6755 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
6757 2018-05-28 Richard Biener <rguenther@suse.de>
6759 PR tree-optimization/85933
6760 * tree-vect-data-refs.c (vect_record_base_alignments): Only
6761 look at stmts marked as vectorizable.
6763 2018-05-28 Richard Biener <rguenther@suse.de>
6765 PR tree-optimization/85934
6766 * tree-vect-generic.c (expand_vector_operations_1): Hoist
6767 vector boolean check before scalar optimization.
6769 2018-05-28 Jakub Jelinek <jakub@redhat.com>
6771 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
6774 2018-05-28 Mark Wielaard <mark@klomp.org>
6776 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
6777 if it is an expression containing a minus sign.
6779 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
6781 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
6783 2018-05-27 Paul Koning <ni1d@arrl.net>
6785 * config/pdp11/pdp11.md (truncsihi2): Remove.
6787 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
6788 Chung-Ju Wu <jasonwucj@gmail.com>
6790 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
6792 (unaligned_store_dw): Ditto.
6793 * config/nds32/nds32-memory-manipulation.c
6794 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
6795 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
6796 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
6797 (emit_setmem_word_loop): Rename to ...
6798 (emit_setmem_doubleword_loop): ... this.
6799 (nds32_gen_dup_4_byte_to_word_value): New function.
6800 (nds32_gen_dup_8_byte_to_double_word_value): New function.
6801 (nds32_expand_setmem_loop): Refine implementation.
6802 (nds32_expand_setmem_loop_v3m): Ditto.
6803 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
6806 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
6808 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
6810 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
6812 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
6813 (nds32_init_machine_status): Initialize machine->attr_naked_p and
6814 machine->attr_no_prologue_p.
6815 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
6816 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
6817 (nds32_expand_epilogue): Consider attr_naked_p.
6818 (nds32_expand_epilogue_v3pop): Likewise.
6819 (nds32_can_use_return_insn): Likewise.
6820 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
6821 attr_no_prologue_p fields.
6822 * config/nds32/nds32.opt (mret-in-naked-func): New option.
6824 2018-05-27 Jakub Jelinek <jakub@redhat.com>
6827 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
6829 * config/i386/sse.md
6830 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
6832 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
6834 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
6836 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
6838 (*<floatsuffix>floatv2div2sf2): Rename to ...
6839 (*float<floatunssuffix>v2div2sf2): ... this.
6840 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
6841 (float<floatunssuffix>v2div2sf2_mask): ... this.
6842 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
6843 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
6844 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
6846 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
6848 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
6850 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
6852 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
6854 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
6856 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
6857 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
6858 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
6859 gen_ufix_truncv8dfv8si2.
6860 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
6861 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
6862 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
6863 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
6864 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
6865 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
6866 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
6867 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
6869 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
6873 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
6875 2018-05-25 Jim Wilson <jimw@sifive.com>
6877 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
6878 * config/riscv/riscv.c (struct machine_function): Add
6879 interrupt_handler_p and attribute_checked_p fields.
6880 (riscv_attribute_table): Add interrupt.
6881 (riscv_interrupt_type_p): New.
6882 (riscv_save_reg_p): Save extra regs for interrupt handler.
6883 (riscv_use_save_libcall): Return false for interrupt handler.
6884 (riscv_first_stack_step): Add forward declaration.
6885 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
6886 for interrupt handler with large frame. Use it for saved reg list.
6887 (riscv_expand_prologue): Move flag_stack_usage_info support to
6888 eliminate duplication.
6889 (riscv_expand_epilogue): Generate mret for interrupt handler.
6890 (riscv_epilogue_uses): New.
6891 (riscv_can_use_return_insn): Return false for interrupt handler.
6892 (riscv_function_ok_for_sibcall): Likewise.
6893 (riscv_set_current_function): Add interrupt handler support.
6894 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
6895 * config/riscv/riscv.md (UNSPECV_MRET): New.
6897 (riscv_frflags, riscv_fsflags): Use tab after opcode.
6899 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
6901 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
6903 PR tree-optimization/85712
6904 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
6905 this candidate has already been replaced in-situ by a copy.
6907 2018-05-25 Jason Merrill <jason@redhat.com>
6909 PR c++/80485 - inline function non-zero address.
6910 * symtab.c (nonzero_address): Check DECL_COMDAT.
6912 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
6915 * config/alpha/alpha.md (ashlsi3): New insn pattern.
6916 (*ashlsi_se): Rename from *ashldi_se. Define as sign
6917 extension of SImode operation. Use const123_operand predicate.
6918 (*saddsi_1): Remove.
6919 (*saddl_se_1): Ditto.
6921 (*ssubl_se_1): Ditto.
6922 * config/alpha/predicates.md (const123_operand): New predicate.
6923 * config/alpha/constraints.md (P): Use IN_RANGE.
6925 2018-05-25 Richard Biener <rguenther@suse.de>
6927 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
6929 (ref_maybe_used_by_stmt_p): Likewise.
6930 (stmt_may_clobber_ref_p): Likewise.
6931 (stmt_may_clobber_ref_p_1): Likewise.
6932 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
6934 (ref_maybe_used_by_stmt_p): Likewise.
6935 (stmt_may_clobber_ref_p): Likewise.
6936 (stmt_may_clobber_ref_p_1): Likewise.
6937 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
6938 the alias oracle to disambiguate DRs with stmts DR analysis
6940 (vect_analyze_data_refs): Do not give up on not analyzable
6941 DRs for BB vectorization. Remove code truncating the dataref
6944 2018-05-25 Jakub Jelinek <jakub@redhat.com>
6947 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
6948 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
6949 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
6951 2018-05-25 Richard Biener <rguenther@suse.de>
6953 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
6954 function, combining stmt data ref gathering and fatal analysis
6956 (vect_analyze_data_refs): Remove now redudnant code and simplify.
6957 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
6958 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
6959 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
6960 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
6962 2018-05-25 Bin Cheng <bin.cheng@arm.com>
6964 PR tree-optimization/85720
6965 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
6966 SCC if all partitions are builtins.
6967 (version_loop_by_alias_check): New parameter. Generate cancelable
6968 runtime alias check if all partitions are builtins.
6969 (distribute_loop): Update call to above function.
6971 2018-05-25 Bin Cheng <bin.cheng@arm.com>
6973 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
6974 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
6975 (parm_default_def_partition_arg): Ditto.
6976 (set_parm_default_def_partition): Ditto.
6977 (get_parm_default_def_partitions): Ditto and make it static.
6978 (get_undefined_value_partitions): Ditto and make it static.
6979 (remove_ssa_form): Refactor call to init_var_map here.
6980 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
6981 computation for loop region.
6982 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
6983 (register_default_def): Delete.
6984 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
6985 (parm_default_def_partition_arg): Ditto.
6986 (set_parm_default_def_partition): Ditto.
6987 (get_parm_default_def_partitions): Ditto and make it static.
6988 (get_undefined_value_partitions): Ditto and make it static.
6989 (coalesce_with_default, coalesce_with_default): Update comment.
6990 (create_coalesce_list_for_region): New func factored out from
6991 create_outofssa_var_map.
6992 (populate_coalesce_list_for_outofssa): New func factored out from
6993 create_outofssa_var_map and coalesce_ssa_name.
6994 (create_outofssa_var_map): Delete.
6995 (coalesce_ssa_name): Refactor to support live range computation.
6996 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
6997 (get_parm_default_def_partitions): Delete.
6998 (get_undefined_value_partitions): Ditto.
6999 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
7000 computation for loop region.
7001 (new_tree_live_info, loe_visit_block): Ditto.
7002 (live_worklist, set_var_live_on_entry): Ditto.
7003 (calculate_live_on_exit, verify_live_on_entry): Ditto.
7004 * tree-ssa-live.h (struct _var_map): New fields.
7005 (init_var_map): Change decl.
7006 (region_contains_p): New.
7008 2018-05-25 Bin Cheng <bin.cheng@arm.com>
7010 * tree-ssa-live.h (live_merge_and_clear): Delete.
7012 2018-05-25 Richard Biener <rguenther@suse.de>
7015 * tree-dump.c (dequeue_and_dump): Remove access to removed
7016 operand 2 of a SWITCH_EXPR.
7018 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7020 * doc/sourcebuild.texi (vect_double_cond_arith): Include
7021 multiplication and division.
7022 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
7023 (cond_udiv@var{m}, cond_umod@var{m}): Document.
7024 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
7025 (cond_udiv_optab, cond_umod_optab): New optabs.
7026 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
7027 (IFN_COND_RDIV): New internal functions.
7028 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
7029 TRUNC_MOD_EXPR and RDIV_EXPR.
7030 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
7031 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
7033 (SVE_INT_BINARY): Include mult.
7034 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
7035 (optab, sve_int_op): Handle mult.
7036 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
7038 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
7039 for SVE_INT_BINARY_SD.
7041 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7043 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
7044 (optab, sve_int_op): Handle div and udiv.
7045 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
7046 for SVE_INT_BINARY_SD.
7047 (*<optab><mode>3): New insn for the same.
7049 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7051 * tree-vect-patterns.c: Include predict.h.
7052 (vect_recog_divmod_pattern): Restrict check for division support
7053 to when optimizing for size.
7055 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7057 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
7058 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
7059 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
7060 (gimple_match_op::set_op): Likewise.
7061 (gimple_resimplify4): Declare.
7062 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
7063 (expr::gen_transform): Likewise.
7064 (decision_tree::gen): Generate a simplification routine for 4 operands.
7065 * gimple-match-head.c (gimple_simplify): Add an overload for
7066 4 operands. In the top-level function, handle up to 4 call
7067 arguments and call gimple_resimplify4.
7068 (gimple_resimplify4): New function.
7069 (build_call_internal): Pass a fourth operand.
7070 (maybe_push_to_seq): Likewise.
7071 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
7072 Fold VEC_COND_EXPRs of an operation and a default value into
7073 an IFN_COND_* function if possible.
7074 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
7076 (SVE_COND_FP_BINARY): Include them.
7077 (optab, sve_fp_op): Handle them.
7078 (SVE_INT_BINARY_REV): New code iterator.
7079 (SVE_COND_FP_BINARY_REV): New int iterator.
7080 (commutative): New int attribute.
7081 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
7083 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
7085 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
7086 (*cond_<optab><mode>): New patterns for reversed operands.
7088 2018-05-25 Richard Biener <rguenther@suse.de>
7090 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
7091 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
7092 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
7093 (STMT_VINFO_GROUPED_ACCESS): Adjust.
7094 * tree-vect-data-refs.c (everywhere): Adjust users.
7095 * tree-vect-loop.c (everywhere): Likewise.
7096 * tree-vect-slp.c (everywhere): Likewise.
7097 * tree-vect-stmts.c (everywhere): Likewise.
7098 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
7100 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7102 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
7104 (gcc_cv_as_section_exclude): ... this.
7105 Try Solaris as #exclude syntax.
7106 * configure: Regenerate.
7107 * config.in: Regenerate.
7108 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
7110 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
7111 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
7113 * varasm.c (default_elf_asm_named_section): Don't check if
7114 HAVE_GAS_SECTION_EXCLUDE is defined.
7116 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7118 * doc/md.texi: Update the documentation of the cond_* optabs
7119 to mention the new final operand. Fix GET_MODE_NUNITS call.
7120 Describe the scalar case too.
7121 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
7122 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
7124 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
7125 (get_conditional_internal_fn): Update comment.
7126 * tree-vect-loop.c (vectorizable_reduction): Pass the original
7127 accumulator value as a final argument to conditional functions.
7128 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
7129 a define_expand and add an "else" operand. Assert for now that
7130 the else operand is equal to operand 2. Use SVE_INT_BINARY and
7131 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
7132 (*cond_<optab><mode>): New patterns.
7133 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
7134 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7135 (UNSPEC_COND_EOR): Delete.
7136 (optab): Remove associated mappings.
7137 (SVE_INT_BINARY): New code iterator.
7138 (sve_int_op): Remove int attribute and add "minus" to the code
7140 (SVE_COND_INT_OP): Delete.
7141 (SVE_COND_FP_OP): Rename to...
7142 (SVE_COND_FP_BINARY): ...this.
7144 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7146 * optabs.c (can_reuse_operands_p): New function.
7147 (maybe_legitimize_operands): Try to reuse the results for
7150 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
7152 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
7153 Add {q} suffix to insn mnemonic.
7155 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7157 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
7158 (msp430_warn_func_return): New.
7160 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
7162 * fold-const.c (tree_nonzero_bits): New function.
7163 * fold-const.h (tree_nonzero_bits): Likewise.
7164 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
7165 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
7167 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
7171 * varasm.c (assemble_alias): Check ifunc_resolver only on
7174 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
7177 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
7178 when memory input operand is handled.
7180 2018-05-24 Luis Machado <luis.machado@linaro.org>
7182 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
7184 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
7186 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
7188 * match.pd: Delay FMA folds until after vectorization.
7190 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
7193 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
7194 address check not strict.
7196 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
7198 * gimple-match.h (gimple_match_op): New class.
7199 (mprts_hook): Replace parameters with a gimple_match_op *.
7200 (maybe_build_generic_op): Likewise.
7201 (gimple_simplified_result_is_gimple_val): Replace parameters with
7202 a const gimple_match_op *.
7203 (gimple_simplify): Replace code_helper * and tree * parameters with
7204 a gimple_match_op * parameter.
7205 (gimple_resimplify1): Replace code_helper *, tree and tree *
7206 parameters with a gimple_match_op * parameter.
7207 (gimple_resimplify2): Likewise.
7208 (gimple_resimplify3): Likewise.
7209 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
7210 parameters with a gimple_match_op * parameter.
7211 * gimple-match-head.c (gimple_simplify): Change prototypes of
7212 auto-generated functions to take a gimple_match_op * instead of
7213 separate code_helper * and tree * parameters. Make the same
7214 change in the top-level overload and update calls to the
7215 gimple_resimplify routines. Update calls to the auto-generated
7216 functions and to maybe_push_res_to_seq in the publicly-facing
7217 operation-specific gimple_simplify overloads.
7218 (gimple_match_op::MAX_NUM_OPS): Define.
7219 (gimple_resimplify1): Replace rcode and ops with a single res_op
7220 parameter. Update call to gimple_simplify.
7221 (gimple_resimplify2): Likewise.
7222 (gimple_resimplify3): Likewise.
7223 (mprts_hook): Replace parameters with a gimple_match_op *.
7224 (maybe_build_generic_op): Likewise.
7225 (build_call_internal): Replace type, nargs and ops with
7226 a gimple_match_op *.
7227 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
7228 with a single gimple_match_op *. Update calls to mprts_hook,
7229 build_call_internal and gimple_simplified_result_is_gimple_val.
7230 Factor out code that is common to the tree_code and combined_fn cases.
7231 * genmatch.c (expr::gen_transform): Replace tem_code and
7232 tem_ops with a gimple_match_op called tem_op. Update calls
7233 to the gimple_resimplify functions and maybe_push_res_to_seq.
7234 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
7235 res_ops. Update call to the gimple_resimplify functions.
7236 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
7237 (decision_tree::gen): Make the functions take a gimple_match_op *
7238 called res_op instead of separate res_code and res_ops parameters.
7239 Update call accordingly.
7240 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
7241 and ops with a single res_op parameter. Update calls to
7242 maybe_build_generic_op and maybe_push_res_to_seq.
7243 (fold_stmt_1): Update calls to gimple_simplify and
7244 replace_stmt_with_simplification.
7245 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
7246 and gimple_simplified_result_is_gimple_val.
7247 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
7249 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
7250 with a gimple_match_op *.
7251 (vn_nary_build_or_lookup): Likewise. Update call to
7252 vn_nary_build_or_lookup_1.
7253 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
7254 gimple_match_op *. Update calls to the gimple_resimplify routines
7255 and to gimple_simplified_result_is_gimple_val.
7256 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
7257 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
7258 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
7259 (visit_nary_op): Likewise.
7260 (visit_reference_op_load): Likewise.
7262 2018-05-23 Luis Machado <luis.machado@linaro.org>
7264 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
7265 modifier for printing the step amount.
7267 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
7270 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
7273 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
7275 * doc/sourcebuild.texi (Endianness): New subsubsection.
7277 2018-05-23 Luis Machado <luis.machado@linaro.org>
7279 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7280 <prefetch_dynamic_strides>: New const bool field.
7281 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7282 prefetch_dynamic_strides.
7283 (exynosm1_prefetch_tune): Likewise.
7284 (thunderxt88_prefetch_tune): Likewise.
7285 (thunderx_prefetch_tune): Likewise.
7286 (thunderx2t99_prefetch_tune): Likewise.
7287 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
7289 (aarch64_override_options_internal): Update to set
7290 PARAM_PREFETCH_DYNAMIC_STRIDES.
7291 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7292 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7293 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7294 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7295 prefetch-dynamic-strides setting.
7297 2018-05-23 Luis Machado <luis.machado@linaro.org>
7299 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7300 <minimum_stride>: New const int field.
7301 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7302 minimum_stride field defaulting to -1.
7303 (exynosm1_prefetch_tune): Likewise.
7304 (thunderxt88_prefetch_tune): Likewise.
7305 (thunderx_prefetch_tune): Likewise.
7306 (thunderx2t99_prefetch_tune): Likewise.
7307 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
7308 <default_opt_level>: Set to 3.
7309 (aarch64_override_options_internal): Update to set
7310 PARAM_PREFETCH_MINIMUM_STRIDE.
7311 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7312 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7313 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7314 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7315 stride is constant and is below the minimum stride threshold.
7317 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7319 * config/arm/arm-cpus.in (mode26): Delete.
7320 (armv4): Delete mode26 reference.
7321 * config/arm/arm.c (arm_configure_build_target): Delete use of
7324 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
7326 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
7328 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
7329 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
7331 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
7332 (floatunsdidf2): Ditto.
7334 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
7336 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
7337 (fixuns_trunc<mode>si2_avx512f): Ditto.
7338 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
7339 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
7340 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
7342 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
7344 PR rtl-optimization/79985
7345 * df-scan.c (df_insn_refs_collect): Remove special case for
7346 global registers and asm statements.
7348 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
7350 * extend.texi (Global Register Variables): Rewrite the bullet list.
7351 Note that the register is available for allocation. Note that access
7352 via inline asm must use constraints. Add note about async-signal
7353 handlers. Remove paragraph about automagic register selection.
7355 2018-05-23 Richard Biener <rguenther@suse.de>
7357 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
7358 of fixed offset from memset VN.
7360 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
7362 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
7364 (alloc_cand_and_find_basis): Initialize first_interp field.
7365 (slsr_process_mul): Modify first_interp field.
7366 (slsr_process_add): Likewise.
7367 (slsr_process_cast): Modify first_interp field for each new
7369 (slsr_process_copy): Likewise.
7370 (dump_candidate): Dump first_interp field.
7371 (replace_mult_candidate): Process all interpretations, not just
7373 (replace_rhs_if_not_dup): Likewise.
7374 (replace_one_candidate): Likewise.
7376 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
7378 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
7380 (aarch64_needs_frame_chain): New function.
7381 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
7383 2018-05-23 Sudakshina Das <sudi.das@arm.com>
7386 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
7387 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
7388 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
7389 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
7390 as true for strict-align.
7391 (aarch64_can_inline_p): Perform checks even when callee has no
7392 attributes to check for strict alignment.
7393 * doc/extend.texi (AArch64 Function Attributes): Document
7395 * doc/invoke.texi: (AArch64 Options): Likewise.
7397 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
7399 PR tree-optimization/85853
7400 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
7401 the handling of the root of the node to...
7402 (vect_slp_analyze_node_operations_1): ...this new function,
7403 and run the whole thing with the child nodes' def types
7404 set according to their SLP node's def type.
7406 2018-05-23 Richard Biener <rguenther@suse.de>
7409 * tree-data-ref.c (create_runtime_alias_checks): Defer
7410 and ignore overflow warnings.
7412 2018-05-23 Yury Gribov <tetra2005@gmail.com>
7414 PR tree-optimization/85822
7415 * tree-vrp.c (is_masked_range_test): Fix handling of negative
7418 2018-05-23 Richard Biener <rguenther@suse.de>
7420 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
7421 memset constants via native_interpret_expr.
7423 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
7426 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
7428 (cgraph_node::create_alias): Likewise.
7429 (cgraph_node::get_availability): Check ifunc_resolver instead
7430 of looking up ifunc attribute.
7431 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
7432 * varasm.c (do_assemble_alias): Likewise.
7433 (assemble_alias): Likewise.
7434 (default_binds_local_p_3): Likewise.
7435 * cgraph.h (cgraph_node): Add ifunc_resolver.
7436 (cgraph_node::only_called_directly_or_aliased_p): Return false
7438 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
7440 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
7441 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
7442 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
7443 instead of looking up ifunc attribute.
7445 2018-05-22 Luis Machado <luis.machado@linaro.org>
7447 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
7449 2018-05-22 Martin Sebor <msebor@redhat.com>
7452 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
7453 only when expasion succeeds.
7454 (expand_builtin_strcmp): Same.
7455 (expand_builtin_strncmp): Same.
7457 2018-05-22 Martin Sebor <msebor@redhat.com>
7459 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
7461 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
7462 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7464 * config/aarch64/aarch64-ldpstp.md: Replace uses of
7465 aarch64_mem_pair_operand with memory_operand and delete operand swapping
7467 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
7468 Add check for legitimate_address.
7469 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
7470 (aarch64_swap_ldrstr_operands): New.
7471 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
7474 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
7475 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7477 * config/aarch64/aarch64.md: New patterns to generate stp
7479 (store_pair_sw, store_pair_dw): New patterns to generate stp for
7480 single words and double words.
7481 (load_pair_sw, load_pair_dw): Likewise.
7482 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
7484 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
7486 * config/aarch64/aarch64-ldpstp.md: Modify peephole
7487 for different mode ldpstp and add peephole for merged zero stores.
7489 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
7491 (aarch64_gen_store_pair): Rename calls to match new patterns.
7492 (aarch64_gen_load_pair): Rename calls to match new patterns.
7493 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
7494 (load_pair<DREG:mode><DREG2:mode>): ... This.
7495 (store_pair<mode>): Rename to...
7496 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
7497 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
7499 (V_INT_EQUIV): Handle SImode.
7500 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
7503 2018-05-22 Martin Sebor <msebor@redhat.com>
7506 * calls.c (maybe_warn_nonstring_arg): Use string length to set
7507 or ajust the presumed bound on an operation to avoid unnecessary
7510 2018-05-22 Martin Sebor <msebor@redhat.com>
7512 PR tree-optimization/85826
7513 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
7514 assuming that a DECL necesarily has a constant size.
7516 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
7519 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
7521 2018-05-22 Richard Biener <rguenther@suse.de>
7523 PR tree-optimization/85834
7524 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
7525 non-constant and non-zero memset arguments.
7527 2018-05-22 Martin Liska <mliska@suse.cz>
7530 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
7532 2018-05-22 Richard Biener <rguenther@suse.de>
7534 PR tree-optimization/85863
7535 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
7536 comparisons when vectype is specified.
7537 (vectorizable_condition): Do not specify vectype for
7538 vect_is_simple_cond when SLP vectorizing.
7540 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
7543 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
7544 define __ibm128 as long double.
7545 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
7546 as a distinct type when IEEE 128-bit support is enabled.
7547 (init_float128_ieee): Fix up conversions between IFmode and IEEE
7548 128-bit types to use the correct functions.
7549 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
7550 convert between 128-bit floating point types that have different
7551 modes but the same representation, instead of using gen_lowpart to
7553 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
7555 (IFKF_reg): New attributes to give the register constraints for
7557 (extend<mode>tf2_internal): New insns to mark an explicit
7558 conversion between 128-bit floating point types that have a
7559 different mode but share the same representation.
7561 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
7563 PR tree-optimization/85814
7564 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
7565 a null return from get_strinfo when unsharing the next
7566 strinfo in the chain.
7568 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
7571 * varasm.c (weak_finish): Clean up weak_decls.
7573 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7575 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
7576 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
7577 UNSPEC_UADALP values.
7578 * config/aarch64/iterators.md (ABAL): New int iterator.
7581 (sur): Add mappings for the above.
7582 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
7584 (aarch64_<sur>abal<mode>_4): Likewise.
7585 (aarch64_<sur>adalp<mode>_3): Likewise.
7586 (<sur>sadv16qi): New define_expand.
7588 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
7590 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
7591 (*movdf_internal): Ditto.
7592 (*rcpsf2_sse): Ditto.
7593 (*rsqrtsf2_sse): Ditto.
7594 (*sqrt<mode>2_sse): Ditto.
7596 2018-05-21 Tamar Christina <tamar.christina@arm.com>
7598 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
7600 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
7601 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
7602 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
7603 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
7605 * config/aarch64/arm_neon.h: Likewise.
7606 * config/aarch64/iterators.md (VQ_I): New.
7608 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
7610 * config.gcc: Add arc/t-multilib-linux to tmake_file for
7612 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
7615 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
7617 * config/nds32/constraints.md (S): New constraint.
7618 * config/nds32/nds32.md (call_internal): Use constraint S.
7619 (call_value_internal): Likewise.
7620 (sibcall_internal): Likewise.
7621 (sibcall_value_internal): Likewise.
7623 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
7624 Chung-Ju Wu <jasonwucj@gmail.com>
7626 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
7629 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
7630 Chung-Ju Wu <jasonwucj@gmail.com>
7632 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
7633 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
7634 (nds32_rtx_costs_impl): Simplify.
7635 (nds32_address_cost_impl): Simplify.
7636 (nds32_init_rtx_costs): New function.
7637 (nds32_rtx_costs_speed_prefer): Likewise.
7638 (nds32_rtx_costs_size_prefer): Likewise.
7639 (nds32_address_cost_speed_prefer): Likewise.
7640 (nds32_address_cost_speed_fwprop): Likewise.
7641 (nds32_address_cost_size_prefer): Likewise.
7642 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
7643 * config/nds32/nds32.c (nds32_option_override): Use
7644 nds32_init_rtx_costs function.
7646 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
7648 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
7649 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
7650 (TARGET_PIPELINE_N8): Likewise.
7651 (TARGET_PIPELINE_N10): Likewise.
7652 (TARGET_PIPELINE_N13): Likewise.
7653 (TARGET_PIPELINE_GRAYWOLF): Likewise.
7655 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
7657 * config/nds32/nds32-fpu.md: Update copyright year.
7659 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7661 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
7663 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7665 * config/nds32/nds32.c
7666 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
7667 * config/nds32/nds32.opt (minline-asm-r15): New option.
7669 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7671 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
7673 * config/nds32/nds32.md (abssi2): New pattern.
7675 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
7677 * config/i386/i386.md (rex64namesuffix): New mode attribute.
7678 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
7679 Merge insn pattern from sse_cvtsi2ss<round_name> and
7680 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
7681 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
7682 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
7683 using SWI48 mode iterator.
7684 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
7685 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
7686 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
7687 pattern from sse_cvttss2si<round_saeonly_name>
7688 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
7689 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
7690 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
7691 using SWI48 mode iterator.
7692 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
7693 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
7694 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
7695 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
7696 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
7697 using SWI48 mode iterator.
7698 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
7699 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
7700 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
7701 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
7702 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
7703 SWI48 mode iterator.
7704 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
7705 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
7706 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
7707 pattern from sse_cvttsd2si<round_saeonly_name>
7708 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
7710 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7712 * config/nds32/nds32-md-auxiliary.c
7713 (nds32_valid_smw_lwm_base_p): Refine.
7714 (nds32_output_smw_single_word): Refine.
7715 (nds32_output_smw_double_word): New.
7716 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
7718 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7720 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
7721 (nds32_output_stack_pop): Refine.
7722 (nds32_expand_unaligned_load): Refine.
7723 (nds32_expand_unaligned_store): Refine.
7725 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7726 Chung-Ju Wu <jasonwucj@gmail.com>
7728 * config/nds32/constants.md: Add TP_REGNUM constant.
7729 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
7730 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
7732 * config/nds32/nds32-doubleword.md: Consider flag_pic.
7733 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
7734 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
7735 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
7736 and PIC code generation.
7737 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
7739 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
7741 * config/nds32/nds32.md: Support TLS and PIC.
7742 * config/nds32/nds32.c: Support TLS and PIC.
7743 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
7744 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
7747 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7749 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
7750 mode with E_ prefix.
7752 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7753 Chung-Ju Wu <jasonwucj@gmail.com>
7755 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
7756 * config/nds32/nds32-md-auxiliary.c
7757 (symbolic_reference_mentioned_p): New.
7758 (nds32_legitimize_ict_address): New.
7759 (nds32_expand_ict_move): New.
7760 (nds32_indirect_call_referenced_p): New.
7761 (nds32_symbol_binds_local_p): Delete.
7762 (nds32_long_call_p): Modify.
7763 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
7764 * config/nds32/nds32-protos.h
7765 (symbolic_reference_mentioned_p): Declare.
7766 (nds32_legitimize_ict_address): Declare.
7767 (nds32_expand_ict_move): Declare.
7768 (nds32_indirect_call_referenced_p): Declare.
7769 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
7770 (nds32_relax_group): Use nds32_ict_const_p as condition.
7771 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
7772 (nds32_asm_file_start): Output ict_model directive in asm code.
7773 (nds32_legitimate_address_p): Consider indirect call.
7774 (nds32_print_operand): Consider indirect call.
7775 (nds32_print_operand_address): Consider indirect call.
7776 (nds32_insert_attributes): Handle "indirect_call" attribute.
7777 (TARGET_LEGITIMATE_ADDRESS_P): Define.
7778 (TARGET_LEGITIMATE_CONSTANT_P): Define.
7779 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
7780 (TARGET_DELEGITIMIZE_ADDRESS): Define.
7781 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
7782 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
7783 (TARGET_ICT_MODEL_SMALL): Define.
7784 (TARGET_ICT_MODEL_LARGE): Define.
7785 * config/nds32/nds32.md (movsi): Consider ict model.
7786 (call, call_value): Consider ict model.
7787 (sibcall, sibcall_value): Consider ict model.
7788 * config/nds32/nds32.opt (mict-model): New option.
7789 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
7792 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
7793 Monk Chiang <sh.chiang04@gmail.com>
7794 Jim Wilson <jimw@sifive.com>
7796 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
7797 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
7798 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
7799 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
7800 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
7801 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
7802 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
7803 compute save_libcall_adjustment properly.
7804 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
7805 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
7806 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
7807 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
7808 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
7809 (ABI_SPEC): Handle mabi=ilp32e.
7810 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
7811 (RVE): Add RVE mask.
7812 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
7813 <-march>: Add rv32e as an example.
7815 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
7818 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
7819 (intra_create_variable_infos): Handle C++ constructors.
7821 2018-05-18 Martin Liska <mliska@suse.cz>
7823 * passes.def: Remove a redundant pass.
7825 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
7828 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
7830 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7832 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
7834 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
7836 (armv2, armv2a, armv3, armv3m): Delete architectures.
7837 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
7838 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
7839 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
7841 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
7842 (*mulsidi3adddi): Likewise.
7843 (mulsidi3): Likewise.
7844 (*mulsidi3_nov6): Likewise.
7845 (umulsidi3): Likewise.
7846 (umulsidi3_nov6): Likewise.
7847 (umaddsidi4): Likewise.
7848 (*umulsidi3adddi): Likewise.
7849 (smulsi3_highpart): Likewise.
7850 (*smulsi3_highpart_nov6): Likewise.
7851 (umulsi3_highpart): Likewise.
7852 (*umulsi3_highpart_nov6): Likewise.
7853 * config/arm/arm.h (arm_arch3m): Delete.
7854 * config/arm/arm.c (arm_arch3m): Delete.
7855 (arm_option_override_internal): Update armv3-related comment.
7856 (arm_configure_build_target): Delete use of isa_bit_mode32.
7857 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
7858 (arm_rtx_costs_internal): Delete check of arm_arch3m.
7859 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
7861 (mulusa3): Likewise.
7862 * config/arm/arm-protos.h (arm_arch3m): Delete.
7863 * config/arm/arm-tables.opt: Regenerate.
7864 * config/arm/arm-tune.md: Likewise.
7865 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
7866 deleted architectures.
7868 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7870 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
7871 (armv5t, armv5te): New features.
7872 (ARMv5, ARMv5e): Delete fgroups.
7873 (ARMv5t, ARMv5te): Adjust for above changes.
7875 (armv5, armv5e): Delete arches.
7876 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
7878 (*call_reg_arm): Likewise.
7879 (*call_value_reg_armv5): Likewise.
7880 (*call_value_reg_arm): Likewise.
7881 (*call_symbol): Likewise.
7882 (*call_value_symbol): Likewise.
7883 (*sibcall_insn): Likewise.
7884 (*sibcall_value_insn): Likewise.
7886 (prefetch): Likewise.
7887 (define_split and define_peephole2 dependent on arm_arch5):
7889 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
7891 (TARGET_ARM_QBIT): Likewise.
7892 (TARGET_DSP_MULTIPLY): Likewise.
7893 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
7894 (arm_arch5, arm_arch5e): Delete.
7895 (arm_arch5t, arm_arch5te): Declare.
7896 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
7897 (arm_arch5t): Declare.
7898 (arm_option_reconfigure_globals): Update for the above.
7899 (arm_options_perform_arch_sanity_checks): Update comment, replace
7900 use of arm_arch5 with arm_arch5t.
7901 (use_return_insn): Likewise.
7902 (arm_emit_call_insn): Likewise.
7903 (output_return_instruction): Likewise.
7904 (arm_final_prescan_insn): Likewise.
7905 (arm_coproc_builtin_available): Likewise.
7906 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
7907 arm_arch5e with arm_arch5t and arm_arch5te.
7908 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
7909 (arm_arch5t, arm_arch5te): Declare.
7910 * config/arm/arm-tables.opt: Regenerate.
7911 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
7912 * config/arm/t-multilib: Likewise.
7913 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
7914 instead of arm_arch5.
7915 (*call_reg_thumb1): Likewise.
7916 (*call_value_reg_thumb1_v5): Likewise.
7917 (*call_value_reg_thumb1): Likewise.
7918 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
7920 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
7922 2018-05-18 Martin Liska <mliska@suse.cz>
7924 PR gcov-profile/84846
7925 * doc/gcov.texi: Document -t option of gcov tool.
7927 2018-05-18 Martin Liska <mliska@suse.cz>
7929 PR gcov-profile/84846
7930 * gcov.c (print_usage): Add new -t option.
7931 (process_args): Handle the option.
7932 (generate_results): Use stdout as output when requested by
7935 2018-05-18 Martin Liska <mliska@suse.cz>
7937 PR gcov-profile/84846
7938 * coverage.c (coverage_init): Write PWD to .gcno file.
7939 * doc/gcov.texi: Document how working directory is printed.
7940 * gcov-dump.c (dump_gcov_file): Print PWD.
7941 * gcov.c (output_intermediate_file): Likewise.
7942 (read_graph_file): Read PWD string.
7943 (output_lines): Print PWD.
7945 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7948 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
7949 for retval and return false if all args to phi are zero.
7951 2018-05-18 Richard Biener <rguenther@suse.de>
7953 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
7955 (evrp_dom_walker::before_dom_children): Call it.
7957 2018-05-18 Richard Biener <rguenther@suse.de>
7959 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
7960 results when processing array refs with variable index.
7962 2018-05-18 Toon Moene <toon@moene.org>
7964 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
7965 directly after that of -floop-interchange. Indicate that both
7966 options are enabled by default when specifying -O3.
7968 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7970 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
7971 iterator. Delete separate integer-mode vec_set<mode> expander.
7972 (aarch64_simd_vec_setv2di): Delete.
7973 (vec_setv2di): Delete.
7974 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
7975 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
7976 the "w, r" alternative.
7978 2018-05-18 Martin Liska <mliska@suse.cz>
7980 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
7981 * tree-pass.h (make_pass_lower_switch_O0): New function.
7982 * tree-switch-conversion.c (node_has_low_bound): Remove.
7983 (node_has_high_bound): Likewise.
7984 (node_is_bounded): Likewise.
7985 (class pass_lower_switch): Make it a template type and create
7987 (pass_lower_switch::execute): Add template argument.
7988 (make_pass_lower_switch): New function.
7989 (make_pass_lower_switch_O0): New function.
7990 (do_jump_if_equal): Remove.
7991 (emit_case_nodes): Simplify to just handle all 3 cases and leave
7992 all the hard work to tree optimization passes.
7994 2018-05-18 Martin Liska <mliska@suse.cz>
7996 * dbgcnt.c (limit_low): Renamed from limit.
7997 (limit_high): New variable.
7998 (dbg_cnt_is_enabled): Check for upper limit.
7999 (dbg_cnt): Adjust dumping.
8000 (dbg_cnt_set_limit_by_index): Add new argument for high
8002 (dbg_cnt_set_limit_by_name): Likewise.
8003 (dbg_cnt_process_single_pair): Parse new format.
8004 (dbg_cnt_process_opt): Use strtok.
8005 (dbg_cnt_list_all_counters): Remove 'value' and add
8007 * doc/invoke.texi: Document changes.
8009 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
8011 * doc/sourcebuild.texi (scalar_all_fma): Document.
8012 * tree.def (FMA_EXPR): Delete.
8013 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
8014 * internal-fn.c (ternary_direct): New macro.
8015 (expand_ternary_optab_fn): Likewise.
8016 (direct_ternary_optab_supported_p): Likewise.
8017 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
8018 * builtins.c (fold_builtin_fma): Delete.
8019 (fold_builtin_3): Don't call it.
8020 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
8021 * expr.c (expand_expr_real_2): Likewise.
8022 * fold-const.c (operand_equal_p): Likewise.
8023 (fold_ternary_loc): Likewise.
8024 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
8025 * gimple.c (DEFTREECODE): Likewise.
8026 * gimplify.c (gimplify_expr): Likewise.
8027 * optabs-tree.c (optab_for_tree_code): Likewise.
8028 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
8029 * tree-eh.c (operation_could_trap_p): Likewise.
8030 (stmt_could_throw_1_p): Likewise.
8031 * tree-inline.c (estimate_operator_cost): Likewise.
8032 * tree-pretty-print.c (dump_generic_node): Likewise.
8033 (op_code_prio): Likewise.
8034 * tree-ssa-loop-im.c (stmt_cost): Likewise.
8035 * tree-ssa-operands.c (get_expr_operands): Likewise.
8036 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
8037 * fold-const-call.h (fold_fma): Delete.
8038 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
8039 CFN_FNMA and CFN_FNMS.
8041 * genmatch.c (combined_fn): New enum.
8042 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
8043 (commutative_op): New function.
8044 (commutate): Use it. Handle more than 2 operands.
8045 (dt_operand::gen_gimple_expr): Use commutative_op.
8046 (parser::parse_expr): Allow :c to be used with non-binary
8047 operators if the commutative operand is known.
8048 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
8049 CFN_FMS, CFN_FNMA and CFN_FNMS.
8050 (backprop::process_assign_use): Remove FMA_EXPR handling.
8051 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
8052 (gen_hsa_fma): New function.
8053 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
8054 IFN_FNMA and IFN_FNMS.
8055 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
8056 * gimple-fold.h (follow_all_ssa_edges): Declare.
8057 * gimple-fold.c (follow_all_ssa_edges): New function.
8058 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
8059 gimple_build interface and use follow_all_ssa_edges to fold the result.
8060 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
8061 instead of checking for optabs directly.
8062 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
8063 rather than FMA_EXPRs.
8064 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
8065 call to IFN_FMA instead of an FMA_EXPR.
8067 2018-05-17 Jim Wilson <jimw@sifive.com>
8069 * expr.c (do_tablejump): When converting index to Pmode, if we have a
8070 sign extended promoted subreg, and the range does not have the sign bit
8071 set, then do a sign extend.
8073 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
8074 test, check for sign extended subreg and/or constant operands, and
8075 do a sign extend in that case.
8077 2018-05-17 Steve Ellcey <sellcey@cavium.com>
8079 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
8080 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
8082 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
8083 Change logics_shift_reg to logics_shift_imm.
8084 (thunderx2t99_fp_loadpair_basic): Delete.
8085 (thunderx2t99_fp_storepair_basic): Delete.
8086 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
8087 (thunderx2t99_asimd_polynomial): Delete.
8088 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
8089 and neon_fp_mul_d_scalar_q.
8090 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
8091 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
8092 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
8093 (thunderx2t99_asimd_lut): Add missing tbl types.
8094 (thunderx2t99_asimd_ext): Delete.
8095 (thunderx2t99_asimd_load1_1_mult): Delete.
8096 (thunderx2t99_asimd_load1_2_mult): Delete.
8097 (thunderx2t99_asimd_load1_ldp): New.
8098 (thunderx2t99_asimd_load1): New.
8099 (thunderx2t99_asimd_load2): Add missing *load2* types.
8100 (thunderx2t99_asimd_load3): New.
8101 (thunderx2t99_asimd_load4): New.
8102 (thunderx2t99_asimd_store1_1_mult): Delete.
8103 (thunderx2t99_asimd_store1_2_mult): Delete.
8104 (thunderx2t99_asimd_store2_mult): Delete.
8105 (thunderx2t99_asimd_store2_onelane): Delete.
8106 (thunderx2t99_asimd_store_stp): New.
8107 (thunderx2t99_asimd_store1): New.
8108 (thunderx2t99_asimd_store2): New.
8109 (thunderx2t99_asimd_store3): New.
8110 (thunderx2t99_asimd_store4): New.
8112 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
8114 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
8115 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
8117 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
8118 Segher Boessenkool <segher@kernel.crashing.org>
8121 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
8124 2018-05-17 Richard Biener <rguenther@suse.de>
8126 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
8127 for pruning loop and prune defs feeding only already visited PHIs.
8129 2018-05-17 Richard Biener <rguenther@suse.de>
8131 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
8133 2018-05-17 Bin Cheng <bin.cheng@arm.com>
8134 Richard Biener <rguenther@suse.de>
8136 PR tree-optimization/85793
8137 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
8138 for VMAT_ELEMENTWISE.
8140 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
8142 * internal-fn.h (lookup_internal_fn): Declare
8143 * internal-fn.c (lookup_internal_fn): New function.
8144 * gimple.c (gimple_build_call_from_tree): Handle calls to
8146 * gimple-pretty-print.c (dump_gimple_call): Print "." before
8147 internal function names.
8148 * tree-pretty-print.c (dump_generic_node): Likewise.
8149 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
8151 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
8153 * gimple-fold.h (gimple_build): Make the function forms take
8154 combined_fn rather than built_in_function.
8155 (gimple_simplify): Likewise.
8156 * gimple-match-head.c (gimple_simplify): Likewise.
8157 * gimple-fold.c (gimple_build): Likewise.
8158 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
8159 rather than gimple_build_call_internal.
8160 (get_initial_defs_for_reduction): Likewise.
8161 (vect_create_epilog_for_reduction): Likewise.
8162 (vectorizable_live_operation): Likewise.
8164 2018-05-17 Martin Liska <mliska@suse.cz>
8166 * gimple-ssa-sprintf.c (format_directive): Do not use
8167 space in between 'G_' and '('.
8169 2018-05-17 Jakub Jelinek <jakub@redhat.com>
8172 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
8173 even if the mask is not all ones.
8176 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
8178 (ix86_gimple_fold_builtin): Likewise.
8181 * config/i386/i386.c: Include tree-vector-builder.h.
8182 (ix86_vector_shift_count): New function.
8183 (ix86_fold_builtin): Fold shift builtins by scalar count.
8184 (ix86_gimple_fold_builtin): Likewise.
8186 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
8187 _mm512_setzero): New intrinsics.
8189 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
8190 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8192 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
8193 code generation for cases where splatting a value is not useful.
8194 * simplify-rtx.c (simplify_ternary_operation): Simplify
8195 vec_merge across a vec_duplicate and a paradoxical subreg forming
8196 a vector mode to a vec_concat.
8198 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
8200 * config.gcc: Support "goldmont-plus".
8201 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
8203 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8204 PROCESSOR_GOLDMONT_PLUS.
8205 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
8206 (processor_target_table): Add "goldmont-plus".
8207 (PTA_GOLDMONT_PLUS): Define.
8208 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
8209 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
8210 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
8211 (fold_builtin_cpu): Add "goldmont-plus".
8212 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
8213 (ix86_option_override_internal): Add "goldmont-plus".
8214 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
8215 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
8216 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
8217 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
8219 2018-05-17 Richard Biener <rguenther@suse.de>
8221 PR tree-optimization/85757
8222 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
8223 remove defs that only feed that PHI from further processing.
8225 2018-05-16 Jim Wilson <jimw@sifive.com>
8227 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
8229 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
8231 2018-05-16 Mark Wielaard <mark@klomp.org>
8233 * dwarf2out.c (count_index_strings): New function.
8234 (output_indirect_strings): Call count_index_strings and generate
8235 header for dwarf_version >= 5.
8237 2018-05-16 Mark Wielaard <mark@klomp.org>
8239 * dwarf2out.c (dwarf_FORM): New function.
8240 (set_indirect_string): Use dwarf_FORM.
8241 (reset_indirect_string): Likewise.
8242 (size_of_die): Likewise.
8243 (value_format): Likewise.
8244 (output_die): Likewise.
8245 (add_skeleton_AT_string): Likewise.
8246 (output_macinfo_op): Likewise.
8247 (index_string): Likewise.
8248 (output_index_string_offset): Likewise.
8249 (output_index_string): Likewise.
8250 (count_index_strings): Likewise.
8252 2018-05-16 Carl Love <cel@us.ibm.com>
8254 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
8255 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
8257 2018-05-16 Martin Jambor <mjambor@suse.cz>
8259 * ipa-prop.c (ipa_free_all_edge_args): Remove.
8260 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
8262 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
8264 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
8265 (fnma<mode>4): Likewise.
8266 (fms<mode>4): Likewise.
8267 (fnms<mode>4): Likewise.
8268 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
8269 (aarch64_fnma<mode>4): Likewise.
8270 (aarch64_fms<mode>4): Likewise.
8271 (aarch64_fnms<mode>4): Likewise.
8272 (aarch64_fnmadd<mode>4): Likewise.
8274 2018-05-16 Jason Merrill <jason@redhat.com>
8276 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
8278 2018-05-16 Richard Biener <rguenther@suse.de>
8280 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
8281 (dump_stmt_cost): Declare.
8282 (add_stmt_cost): Dump cost we add.
8283 (add_stmt_costs): New function.
8284 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
8286 (vect_analyze_stmt): Adjust prototype.
8287 (vectorizable_condition): Likewise.
8288 (vectorizable_live_operation): Likewise.
8289 (vectorizable_reduction): Likewise.
8290 (vectorizable_induction): Likewise.
8291 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
8292 cost vector to pass to vectorizable_ and record afterwards.
8293 (vect_model_reduction_cost): Take cost vector argument and adjust.
8294 (vect_model_induction_cost): Likewise.
8295 (vectorizable_reduction): Likewise.
8296 (vectorizable_induction): Likewise.
8297 (vectorizable_live_operation): Likewise.
8298 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
8299 SLP_TREE_NUMBER_OF_VEC_STMTS.
8300 (vect_analyze_slp_cost_1): Remove.
8301 (vect_analyze_slp_cost): Likewise.
8302 (vect_slp_analyze_node_operations): Take visited args and
8303 a target cost vector. Avoid processing already visited stmt sets.
8304 (vect_slp_analyze_operations): Use a local cost vector to gather
8305 costs and register those of non-discarded instances.
8306 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
8307 (vect_schedule_slp_instance): Remove copying of
8308 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
8310 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
8311 adding cost. Record cost entry location.
8312 (vect_prologue_cost_for_slp_op): Function to compute cost of
8313 a constant or invariant generated for SLP vect in the prologue,
8314 split out from vect_analyze_slp_cost_1.
8315 (vect_model_simple_cost): Make static. Adjust for SLP costing.
8316 (vect_model_promotion_demotion_cost): Likewise.
8317 (vect_model_store_cost): Likewise, make static.
8318 (vect_model_load_cost): Likewise.
8319 (vectorizable_bswap): Add cost vector arg and adjust.
8320 (vectorizable_call): Likewise.
8321 (vectorizable_simd_clone_call): Likewise.
8322 (vectorizable_conversion): Likewise.
8323 (vectorizable_assignment): Likewise.
8324 (vectorizable_shift): Likewise.
8325 (vectorizable_operation): Likewise.
8326 (vectorizable_store): Likewise.
8327 (vectorizable_load): Likewise.
8328 (vectorizable_condition): Likewise.
8329 (vectorizable_comparison): Likewise.
8330 (can_vectorize_live_stmts): Likewise.
8331 (vect_analyze_stmt): Likewise.
8332 (vect_transform_stmt): Adjust calls to vectorizable_*.
8333 * tree-vectorizer.c: Include gimple-pretty-print.h.
8334 (dump_stmt_cost): New function.
8336 2018-05-16 Richard Biener <rguenther@suse.de>
8338 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
8339 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
8340 * tree-ssa-dse.c: Include tree-ssa-loop.h.
8341 (check_name): New callback.
8342 (dse_classify_store): Track cycles via a visited bitmap of PHI
8343 defs and simplify handling of in-loop and across loop dead stores
8344 and properly fail for loop-variant refs. Handle byte-tracking with
8345 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
8348 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
8350 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
8351 (vect_get_mask_type_for_stmt): Likewise.
8352 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
8354 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
8355 to determine the statement's vector type and the vector type that
8356 should be used for calculating nunits. Deal with cases in which
8357 the type has to be deferred.
8358 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
8359 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
8360 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
8361 (vect_determine_vf_for_stmt): New functions, split out from...
8362 (vect_determine_vectorization_factor): ...here.
8363 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
8364 (vect_get_mask_type_for_stmt): New functions, split out from
8365 vect_determine_vectorization_factor.
8367 2018-05-16 Richard Biener <rguenther@suse.de>
8369 * tree-cfg.c (verify_gimple_assign_ternary): Properly
8370 verify the [VEC_]COND_EXPR embedded comparison.
8372 2018-05-15 Martin Sebor <msebor@redhat.com>
8374 PR tree-optimization/85753
8375 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
8376 RECORD_TYPE in addition to ARRAY_TYPE.
8378 2018-05-15 Martin Sebor <msebor@redhat.com>
8381 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
8383 2018-05-15 Richard Biener <rguenther@suse.de>
8385 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
8386 add by_clobber_p one. Change algorithm to collect all defs
8387 representing uses we need to walk and try reducing them to
8388 a single one before failing.
8389 (dse_dom_walker::dse_optimize_stmt): Adjust.
8391 2018-05-13 Mark Wielaard <mark@klomp.org>
8393 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
8394 (size_of_loc_descr): Likewise.
8395 (output_loc_operands): Likewise.
8396 (output_loc_operands_raw): Likewise.
8397 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
8398 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
8399 (hash_loc_operands): Likewise.
8400 (compare_loc_operands): Likewise.
8402 2018-05-14 Mark Wielaard <mark@klomp.org>
8404 * dwarf2out.c (count_index_addrs): New function.
8405 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
8407 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8409 PR tree-optimization/83648
8410 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
8411 return value as malloc candidate.
8413 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8416 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
8417 param as true in call to suggest_attribute.
8419 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
8421 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
8422 -mreadonly-in-sdata.
8424 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8426 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
8428 (aarch64_crypto_aesd_fused): Likewise.
8430 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
8432 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
8433 (movsi_aarch64): Likewise.
8434 (load_pairsi): Likewise.
8435 (load_pairdi): Likewise.
8436 (store_pairsi): Likewise.
8437 (store_pairdi): Likewise.
8438 (load_pairsf): Likewise.
8439 (load_pairdf): Likewise.
8440 (store_pairsf): Likewise.
8441 (store_pairdf): Likewise.
8442 (zero_extend): Likewise.
8443 (trunc): Swap alternatives.
8444 (fcvt_target): Add '?' to prefer w over r.
8446 2018-05-14 Jakub Jelinek <jakub@redhat.com>
8449 * config/i386/i386.md: Disallow non-commutative arithmetics in
8450 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
8451 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
8452 in the peephole2 before it.
8454 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
8456 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
8457 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
8458 (ix86_handle_option): Handle -mcldemote.
8459 * config.gcc: New header.
8460 * config/i386/cldemoteintrin.h: New file.
8461 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
8462 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
8464 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8465 OPTION_MASK_ISA_CLDEMOTE.
8466 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
8467 (ix86_valid_target_attribute_inner_p): Ditto.
8468 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
8469 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
8470 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
8471 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
8472 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
8474 * config/i386/i386.opt: Add -mcldemote.
8475 * config/i386/x86intrin.h: New header.
8476 * doc/invoke.texi: Add -mcldemote.
8478 2018-05-14 Richard Biener <rguenther@suse.de>
8480 * doc/match-and-simplify.texi: Adjust :s documentation.
8482 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
8484 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
8485 intended memcpy size.
8486 (REORDER_45): Likewise.
8488 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
8490 * sort.cc: New file.
8491 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
8492 * vec.c (qsort_chk): Use gcc_qsort.
8493 * Makefile.in (OBJS-libcommon): Add sort.o.
8494 (build/sort.o): New target. Use it...
8495 (BUILD_RTL): ... here, and...
8496 (build/gencfn-macros): ... here, and...
8497 (build/genmatch): ... here.
8499 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
8500 Chung-Ju Wu <jasonwucj@gmail.com>
8502 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
8503 * config/nds32/nds32-graywolf.md: New file.
8504 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
8505 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
8507 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
8508 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
8509 * config/nds32/nds32.md (pipeline_model): Add graywolf.
8510 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
8511 * config/nds32/pipelines.md: Include n15 settings.
8513 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
8514 Chung-Ju Wu <jasonwucj@gmail.com>
8516 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
8517 * config/nds32/nds32-n13.md: New file.
8518 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
8519 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
8521 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
8522 * config/nds32/nds32.md (pipeline_model): Add n13.
8523 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
8524 * config/nds32/pipelines.md: Include n13 settings.
8526 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
8527 Chung-Ju Wu <jasonwucj@gmail.com>
8529 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
8530 * config/nds32/nds32-n10.md: New file.
8531 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
8532 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
8534 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
8535 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
8536 * config/nds32/nds32.md (pipeline_model): Add n10.
8537 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
8538 * config/nds32/pipelines.md: Include n10 settings.
8540 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
8541 Kito Cheng <kito.cheng@gmail.com>
8542 Chung-Ju Wu <jasonwucj@gmail.com>
8544 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
8545 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8546 Add enum values for DSP extension instructions.
8547 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
8549 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
8550 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
8552 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
8553 * config/nds32/nds32-dspext.md: New file for DSP implementation.
8554 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
8555 * config/nds32/nds32-intrinsic.md: Likewise.
8556 * config/nds32/nds32_intrinsic.h: Likewise.
8557 * config/nds32/nds32-md-auxiliary.c: Likewise.
8558 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
8559 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
8560 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
8561 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
8562 * config/nds32/nds32-protos.h: New declarations for DSP extension.
8563 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
8564 TYPE_DMAC in switch statement.
8565 * config/nds32/nds32.c: New checking and implementation for DSP
8566 extension instructions.
8567 * config/nds32/nds32.h: Likewise.
8568 * config/nds32/nds32.md: Likewise.
8569 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
8570 * config/nds32/predicates.md: Implement new predicates for DSP
8573 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8575 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
8576 Reformat alternatives and attributes so it is easier to identify
8577 which constraints/attributes go with which instruction.
8578 (mov<mode>_hardfloat32, FMOVE64): Likewise.
8579 (mov<mode>_softfloat32, FMOVE64): Likewise.
8580 (mov<mode>_hardfloat64, FMOVE64): Likewise.
8581 (mov<mode>_softfloat64, FMOVE64): Likewise.
8583 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
8585 * doc/extend.texi (PowerPC Built-in Functions): Rename this
8587 (Basic PowerPC Built-in Functions): The new name of the
8588 subsection previously known as "PowerPC Built-in Functions".
8589 (Basic PowerPC Built-in Functions Available on all Configurations):
8591 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
8592 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
8593 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
8594 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
8596 2018-05-11 Martin Jambor <mjambor@suse.cz>
8599 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
8602 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
8605 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
8607 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
8609 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
8610 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
8611 (ix86_handle_option): Handle -mwaitpkg.
8612 * config.gcc: New header.
8613 * config/i386/cpuid.h (bit_WAITPKG): New bit.
8614 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
8615 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
8617 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8618 OPTION_MASK_ISA_WAITPKG.
8619 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
8620 (ix86_option_override_internal): Add PTA_WAITPKG.
8621 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
8622 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
8623 IX86_BUILTIN_TPAUSE.
8624 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
8625 __builtin_ia32_umwait and __builtin_ia32_tpause.
8626 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
8627 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
8628 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
8629 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
8630 UNSPECV_TPAUSE): New.
8631 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
8632 * config/i386/i386.opt: Add -mwaitpkg.
8633 * config/i386/waitpkgintrin.h: New file.
8634 * config/i386/x86intrin.h: New header.
8635 * doc/invoke.texi: Add -mwaitpkg.
8637 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
8640 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
8642 (cortex-m0): Use armv6s-m isa.
8643 (cortex-m0plus): Likewise.
8644 (cortex-m1): Likewise.
8645 (cortex-m0.small-multiply): Likewise.
8646 (cortex-m0plus.small-multiply): Likewise.
8647 (cortex-m1.small-multiply): Likewise.
8649 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
8650 Jakub Jelinek <jakub@redhat.com>
8652 PR tree-optimization/85692
8653 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
8654 source permute as well.
8656 2018-05-11 Martin Liska <mliska@suse.cz>
8659 * doc/extend.texi: Document LLVM style format for no_sanitize
8662 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8664 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
8665 mode_supports_vsx_dform_quad to mode_supports_dq_form.
8666 (mode_supports_vsx_dform_quad): Likewise.
8667 (mode_supports_vmx_dform): Move these functions to be next to the
8668 other mode_supports functions.
8669 (mode_supports_dq_form): Likewise.
8670 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
8671 mode_supports_dq_form.
8672 (reg_offset_addressing_ok_p): Likewise.
8673 (offsettable_ok_by_alignment): Likewise.
8674 (rs6000_legitimate_offset_address_p): Likewise.
8675 (legitimate_lo_sum_address_p): Likewise.
8676 (rs6000_legitimize_address): Likewise.
8677 (rs6000_legitimize_reload_address): Likewise.
8678 (rs6000_secondary_reload_inner): Likewise.
8679 (rs6000_preferred_reload_class): Likewise.
8680 (rs6000_output_move_128bit): Likewise.
8682 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
8684 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
8685 Generate SImode target register for null target.
8686 <case IX86_BUILTIN_XGETBV>: Ditto.
8687 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
8688 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
8690 2018-05-10 Carl Love <cel@us.ibm.com>
8692 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
8693 dcbtt and dcbtstt if operands[2] is 0.
8695 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
8698 * config/i386/sse.md (usadv64qi): New expander.
8700 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
8702 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
8703 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
8704 -maltivec=be support.
8705 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
8706 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
8707 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
8708 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
8709 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
8710 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
8711 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
8712 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
8713 altivec_vsumsws): Adjust.
8714 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
8715 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
8716 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
8718 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
8719 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
8720 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
8721 (altivec_lve<VI_char>x): Delete expand.
8722 (*altivec_lve<VI_char>x_internal): Rename to...
8723 (altivec_lve<VI_char>x): ... this.
8724 (altivec_lvxl_<mode>): Delete expand.
8725 (*altivec_lvxl_<mode>_internal): Rename to ...
8726 (altivec_lvxl_<mode>): ... this.
8727 (altivec_stvxl_<mode>): Delete expand.
8728 (*altivec_stvxl_<mode>_internal): Rename to ...
8729 (altivec_stvxl_<mode>): ... this.
8730 (altivec_stve<VI_char>x): Delete expand.
8731 (*altivec_stve<VI_char>x_internal): Rename to ...
8732 (altivec_stve<VI_char>x): ... this.
8733 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
8734 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
8735 reduc_plus_scal_<mode>): Adjust.
8736 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
8738 (rs6000_cpu_cpp_builtins): Adjust.
8739 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
8740 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
8741 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
8742 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
8743 -maltivec=be support.
8744 (rs6000_split_vec_extract_var): Adjust.
8745 (rs6000_split_v4si_init): Adjust.
8746 (swap_selector_for_mode): Delete.
8747 (altivec_expand_lvx_be, altivec_expand_stvx_be,
8748 altivec_expand_stvex_be): Delete.
8749 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
8750 -maltivec=be support.
8751 (rs6000_gimple_fold_builtin): Ditto.
8752 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
8754 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
8755 (TARGET_DIRECT_MOVE_64BIT): Adjust.
8756 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
8757 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
8758 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
8759 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
8760 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
8761 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
8762 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
8763 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
8764 anonymous split): Adjust.
8765 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
8766 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
8768 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
8770 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
8771 when --with-gxx-include-dir is also specified.
8772 * configure: Regenerate.
8774 2018-05-09 Jim Wilson <jimw@sifive.com>
8777 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
8778 * config/riscv/t-withmultilib: New.
8779 * config/riscv/withmultilib.h: New.
8780 * doc/install.texi: Document RISC-V --with-multilib-list support.
8782 2018-05-09 Richard Biener <rguenther@suse.de>
8784 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
8786 (vect_bb_vectorization_profitable_p): Adjust. Compute
8787 actual scalar cost using the cost vector and the add_stmt_cost
8790 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8792 PR rtl-optimization/85645
8793 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
8794 in the REG_CFA_REGISTER note for LR, don't leave it empty.
8796 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8798 PR rtl-optimization/85645
8799 * shrink-wrap.c (spread_components): Return a boolean saying if
8800 anything was changed.
8801 (try_shrink_wrapping_separate): Iterate spread_components until
8802 nothing changes anymore.
8804 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8806 PR rtl-optimization/85645
8807 * regrename.c (build_def_use): Also kill the chains that include the
8808 destination of a REG_CFA_REGISTER note.
8810 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8812 PR rtl-optimization/85645
8813 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
8814 insn that has a REG_CFA_REGISTER note.
8816 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
8818 * cfgexpand.c (expand_clobber): New function.
8819 (expand_gimple_stmt_1): Use it.
8820 * tree-vect-stmts.c (vect_clobber_variable): New function,
8822 (vectorizable_simd_clone_call): ...here.
8823 (vectorizable_store): Emit a clobber either side of an
8824 IFN_STORE_LANES sequence.
8825 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
8827 2018-05-09 Tom de Vries <tom@codesourcery.com>
8830 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
8831 (define_insn "trap_if_false"): Add exit after trap.
8833 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
8835 PR rtl-optimization/85638
8836 * bb-reorder.c: Include common/common-target.h.
8837 (create_forwarder_block): New function extracted from...
8838 (fix_up_crossing_landing_pad): ...here. Rename into...
8839 (dw2_fix_up_crossing_landing_pad): ...this.
8840 (sjlj_fix_up_crossing_landing_pad): New function.
8841 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
8842 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
8843 from both partitions and exit the loop after one iteration.
8845 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
8848 * doc/extend.texi (PowerPC Built-in Functions): Rename this
8850 (Basic PowerPC Built-in Functions): The new name of the
8851 subsection previously known as "PowerPC Built-in Functions".
8852 (Basic PowerPC Built-in Functions Available on all Configurations):
8854 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
8856 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
8858 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
8860 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
8863 2018-05-08 Jim Wilson <jimw@sifive.com>
8865 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
8866 (LD_EMUL_SUFFIX): New.
8867 (LINK_SPEC): Use it.
8869 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
8871 * doc/extend.texi (PowerPC Built-in Functions): Rename this
8873 (Basic PowerPC Built-in Functions): The new name of the
8874 subsection previously known as "PowerPC Built-in Functions".
8875 (Basic PowerPC Built-in Functions Available on all Configurations):
8877 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
8879 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
8881 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
8883 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
8886 2018-05-08 Jakub Jelinek <jakub@redhat.com>
8889 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
8890 after cmpelim optimization.
8892 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
8894 * config.gcc: Support "goldmont".
8895 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
8896 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8898 * config/i386/i386.c (m_GOLDMONT): Define.
8899 (processor_target_table): Add "goldmont".
8900 (PTA_GOLDMONT): Define.
8901 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
8902 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
8903 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
8904 (fold_builtin_cpu): Add "goldmont".
8905 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
8906 (ix86_option_override_internal): Add "goldmont".
8907 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
8908 (processor_type): Add PROCESSOR_GOLDMONT.
8909 * config/i386/i386.md: Add CPU "glm".
8910 * config/i386/glm.md: New file.
8911 * config/i386/x86-tune.def: Add m_GOLDMONT.
8912 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
8914 2018-05-08 Jakub Jelinek <jakub@redhat.com>
8917 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
8919 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
8920 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
8921 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
8924 * config/i386/i386.c (ix86_fold_builtin): Handle
8925 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
8928 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
8929 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
8931 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
8934 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
8935 (check_arch): Likewise.
8936 (check_fpu): Return the result rather than printing it.
8937 (end arch): Fix operator precedence.
8938 (end cpu): Likewise.
8939 (END): Print the result from check_fpu.
8941 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
8942 Alan Hayward <alan.hayward@arm.com>
8943 David Sherwood <david.sherwood@arm.com>
8945 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
8946 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
8947 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
8948 (*fcmuo<mode>_and): New patterns.
8950 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
8952 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
8953 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
8954 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
8955 (cmp_op, sve_imm_con): New code attributes.
8956 (SVE_COND_INT_CMP, imm_con): Delete.
8957 (cmp_op): Remove above unspecs from int attribute.
8958 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
8960 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
8961 comparison-specific unspecs.
8962 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
8963 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
8964 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
8965 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
8966 (*vec_fcm<cmp_op><mode>): Rename to...
8967 (*fcm<cmp_op><mode>): ...this and adjust likewise.
8968 (*vec_fcmuo<mode>): Rename to...
8969 (*fcmuo<mode>): ...this and adjust likewise.
8970 (*pred_fcm<cmp_op><mode>): New pattern.
8971 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
8972 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
8974 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
8976 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
8977 (aarch64_emit_sve_predicated_cond): New function.
8978 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
8979 (aarch64_emit_unspec_cond_or): Replace with...
8980 (aarch64_emit_sve_or_conds): ...this new function. Use
8981 aarch64_emit_sve_ptrue_op for the individual comparisons and
8982 aarch64_emit_binop to OR them together.
8983 (aarch64_emit_inverted_unspec_cond): Replace with...
8984 (aarch64_emit_sve_inverted_cond): ...this new function. Use
8985 aarch64_emit_sve_ptrue_op for the comparison and
8986 aarch64_emit_unop to invert the result.
8987 (aarch64_expand_sve_vec_cmp_float): Update after the above
8988 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
8990 2018-05-07 Nathan Sidwell <nathan@acm.org>
8992 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
8993 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
8994 (Backwards Compatibility): Likewise.
8996 2018-05-07 Luis Machado <luis.machado@linaro.org>
9000 2018-05-07 Luis Machado <luis.machado@linaro.org>
9002 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9003 <prefetch_dynamic_strides>: New const bool field.
9004 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9005 prefetch_dynamic_strides.
9006 (exynosm1_prefetch_tune): Likewise.
9007 (thunderxt88_prefetch_tune): Likewise.
9008 (thunderx_prefetch_tune): Likewise.
9009 (thunderx2t99_prefetch_tune): Likewise.
9010 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
9012 (aarch64_override_options_internal): Update to set
9013 PARAM_PREFETCH_DYNAMIC_STRIDES.
9014 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
9015 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
9016 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
9017 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
9018 prefetch-dynamic-strides setting.
9020 2018-05-07 Luis Machado <luis.machado@linaro.org>
9022 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9023 <minimum_stride>: New const int field.
9024 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9025 minimum_stride field.
9026 (exynosm1_prefetch_tune): Likewise.
9027 (thunderxt88_prefetch_tune): Likewise.
9028 (thunderx_prefetch_tune): Likewise.
9029 (thunderx2t99_prefetch_tune): Likewise.
9030 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
9031 (aarch64_override_options_internal): Update to set
9032 PARAM_PREFETCH_MINIMUM_STRIDE.
9033 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
9034 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
9035 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
9036 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
9037 stride is constant and is below the minimum stride threshold.
9039 2018-05-07 Luis Machado <luis.machado@linaro.org>
9041 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
9044 2018-05-07 Luis Machado <luis.machado@linaro.org>
9046 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9047 <prefetch_dynamic_strides>: New const bool field.
9048 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9049 prefetch_dynamic_strides.
9050 (exynosm1_prefetch_tune): Likewise.
9051 (thunderxt88_prefetch_tune): Likewise.
9052 (thunderx_prefetch_tune): Likewise.
9053 (thunderx2t99_prefetch_tune): Likewise.
9054 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
9056 (aarch64_override_options_internal): Update to set
9057 PARAM_PREFETCH_DYNAMIC_STRIDES.
9058 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
9059 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
9060 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
9061 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
9062 prefetch-dynamic-strides setting.
9064 2018-05-07 Luis Machado <luis.machado@linaro.org>
9066 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9067 <minimum_stride>: New const int field.
9068 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9069 minimum_stride field.
9070 (exynosm1_prefetch_tune): Likewise.
9071 (thunderxt88_prefetch_tune): Likewise.
9072 (thunderx_prefetch_tune): Likewise.
9073 (thunderx2t99_prefetch_tune): Likewise.
9074 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
9075 (aarch64_override_options_internal): Update to set
9076 PARAM_PREFETCH_MINIMUM_STRIDE.
9077 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
9078 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
9079 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
9080 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
9081 stride is constant and is below the minimum stride threshold.
9083 2018-05-06 Jakub Jelinek <jakub@redhat.com>
9086 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
9087 the type is addressable. Don't force op into register if it has
9090 2018-05-05 Roland McGrath <mcgrathr@google.com>
9093 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
9094 any section for which we don't know a specific type it should have,
9095 regardless of name. Previously this was done only for the exact
9096 names ".init_array", ".fini_array", and ".preinit_array".
9097 (default_elf_asm_named_section): Add comment about
9098 relationship with default_section_type_flags and SECTION_NOTYPE.
9099 (get_section): Don't consider it a type conflict if one side has
9100 SECTION_NOTYPE and the other doesn't, as long as neither has the
9101 SECTION_BSS et al used in the default_section_type_flags logic.
9103 2018-05-05 Tom de Vries <tom@codesourcery.com>
9106 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
9107 (workaround_barsyncs): New function.
9108 (nvptx_reorg): Use workaround_barsyncs.
9109 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
9110 (define_expand "nvptx_membar_cta"): New define_expand.
9111 (define_insn "*nvptx_membar_cta"): New insn.
9113 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
9115 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
9116 To improve optimization opportunities.
9117 * builtin-types.def: The new needed builtin types for the above.
9119 2018-05-04 Richard Biener <rguenther@suse.de>
9121 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
9122 * gimple-ssa-store-merging.c
9123 (imm_store_chain_info::output_merged_store): Remove redundant create,
9124 release split_store vector contents on failure.
9125 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
9126 scalar stmt vector on cache hit.
9128 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
9130 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
9132 * config.gcc (powerpc-xilinx-eabi*): Remove.
9133 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
9135 (fusion_addis_mem_combo_load): Ditto.
9136 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
9138 (rs6000_cpu_cpp_builtins): Ditto.
9139 * config/rs6000/rs6000-linux.c
9140 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
9141 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
9142 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
9144 (rs6000_setup_reg_addr_masks): Ditto.
9145 (rs6000_init_hard_regno_mode_ok): Ditto.
9146 (rs6000_option_override_internal): Ditto.
9147 (legitimate_lo_sum_address_p): Ditto.
9148 (rs6000_legitimize_address): Ditto.
9149 (rs6000_legitimize_reload_address): Ditto.
9150 (rs6000_legitimate_address_p): Ditto.
9151 (abi_v4_pass_in_fpr): Ditto.
9152 (setup_incoming_varargs): Ditto.
9153 (rs6000_gimplify_va_arg): Ditto.
9154 (rs6000_split_multireg_move): Ditto.
9155 (rs6000_savres_strategy): Ditto.
9156 (rs6000_emit_prologue_components): Ditto.
9157 (rs6000_emit_epilogue_components): Ditto.
9158 (rs6000_emit_prologue): Ditto.
9159 (rs6000_emit_epilogue): Ditto.
9160 (rs6000_elf_file_end): Ditto.
9161 (rs6000_function_value): Ditto.
9162 (rs6000_libcall_value): Ditto.
9163 * config/rs6000/rs6000.h: Ditto.
9164 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
9165 (TARGET_MINMAX): ... this. New.
9166 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
9167 * config/rs6000/rs6000.md: Remove Xilinx FP support.
9168 (*movsi_internal1_single): Delete.
9169 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
9170 mfpu=, mxilinx-fpu): Delete.
9171 * config/rs6000/singlefp.h: Delete.
9172 * config/rs6000/sysv4.h: Remove Xilinx FP support.
9173 * config/rs6000/t-rs6000: Ditto.
9174 * config/rs6000/t-xilinx: Delete.
9175 * config/rs6000/titan.md: Adjust for fp_type removal.
9176 * config/rs6000/vsx.md: Remove Xilinx FP support.
9177 (VStype_simple): Delete.
9178 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
9179 * config/rs6000/xfpu.h: Delete.
9180 * config/rs6000/xfpu.md: Delete.
9181 * config/rs6000/xilinx.h: Delete.
9182 * config/rs6000/xilinx.opt: Delete.
9183 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
9184 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
9186 2018-05-04 Tom de Vries <tom@codesourcery.com>
9189 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
9192 2018-05-04 Richard Biener <rguenther@suse.de>
9195 * tree-complex.c (update_complex_assignment): We are always in SSA form.
9196 (expand_complex_div_wide): Likewise.
9197 (expand_complex_operations_1): Likewise.
9198 (expand_complex_libcall): Preserve EH info of the original stmt.
9199 (tree_lower_complex): Handle removed blocks.
9200 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
9201 on complex multiplication and division libcall builtins.
9203 2018-05-04 Richard Biener <rguenther@suse.de>
9206 * fold-const.c (negate_expr_p): Restrict negation of operand
9207 zero of a division to when we know that can happen without
9209 (fold_negate_expr_1): Likewise.
9211 2018-05-04 Jakub Jelinek <jakub@redhat.com>
9214 * real.h (real_nextafter): Declare.
9215 * real.c (real_nextafter): New function.
9216 * fold-const-call.c (fold_const_nextafter): New function.
9217 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
9218 CASE_CFN_NEXTTOWARD.
9219 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
9220 even when arg1_mode is different from arg0_mode.
9222 2018-05-03 Nathan Sidwell <nathan@acm.org>
9224 * doc/extend.texi (Deprecated Features): Remove
9226 (Backwards Compatibility): Likewise.
9227 * doc/invoke.texi (C++ Language Options): Likewise.
9228 (C++ Dialect Options): Likewise.
9230 2018-05-03 Jakub Jelinek <jakub@redhat.com>
9233 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
9234 _mm512_mask_mullox_epi64): New intrinsics.
9236 2018-05-03 Tom de Vries <tom@codesourcery.com>
9239 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
9240 dump files): Add offload-tree.
9242 2018-05-03 Richard Biener <rguenther@suse.de>
9244 PR tree-optimization/85615
9245 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
9246 to loops not nested in BBs loop father to avoid creating multi-entry
9249 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9251 PR tree-optimization/70291
9252 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
9253 arguments. Change return type to tree. Emit libcall as a new
9254 statement rather than replacing existing one when inplace_p is true.
9255 (expand_complex_multiplication_components): New function.
9256 (expand_complex_multiplication): Expand floating-point complex
9257 multiplication using the above.
9258 (expand_complex_division): Rename inner_type parameter to type.
9259 Update expand_complex_libcall call-site.
9260 (expand_complex_operations_1): Update expand_complex_multiplication
9261 and expand_complex_division call-sites.
9263 2018-05-02 Jakub Jelinek <jakub@redhat.com>
9266 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
9267 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
9268 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
9269 the highest significant bit of the shift count mask is clear. In
9270 check whether and[sq]i3 is needed verify that all significant bits
9271 of the shift count other than the highest are set.
9273 2018-05-02 Tom de Vries <tom@codesourcery.com>
9276 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
9277 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
9278 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
9279 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
9280 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
9281 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
9282 __builtin_goacc_parlevel_size.
9284 2018-05-02 Richard Biener <rguenther@suse.de>
9286 PR tree-optimization/85597
9287 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
9288 do not use split vect_get_vec_defs call but call vect_get_slp_defs
9291 2018-05-02 Tom de Vries <tom@codesourcery.com>
9294 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
9295 dump files): Add ltrans-tree.
9297 2018-05-02 Tom de Vries <tom@codesourcery.com>
9300 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
9301 dump files): Add wpa-ipa.
9303 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
9305 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
9306 powerpc*-*-linux*paired* target.
9307 * config/rs6000/750cl.h: Delete.
9308 * config/rs6000/paired.h: Delete.
9309 * config/rs6000/paired.md: Delete.
9310 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
9312 * config/rs6000/rs6000-builtin.def: Remove paired float support.
9313 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
9314 comment. Remove paired float support.
9315 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
9316 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
9318 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
9319 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
9321 * config/rs6000/rs6000.c: Remove paired float support.
9322 (paired_expand_vector_init, paired_expand_vector_move,
9323 paired_emit_vector_compare, paired_emit_vector_cond_expr,
9324 (paired_expand_lv_builtin, paired_expand_stv_builtin,
9325 paired_expand_builtin, paired_expand_predicate_builtin,
9326 paired_init_builtins): Delete.
9327 * config/rs6000/rs6000.h: Remove paired float support.
9328 * config/rs6000/rs6000.md: Remove paired float support.
9329 (move_from_CR_ov_bit): Delete.
9330 * config/rs6000/rs6000.opt (mpaired): Delete.
9331 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
9332 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
9334 2018-05-02 Richard Biener <rguenther@suse.de>
9337 * gimplify.c (gimplify_save_expr): When in SSA form allow
9338 SAVE_EXPRs to compute to SSA vars.
9340 2018-05-02 Jakub Jelinek <jakub@redhat.com>
9343 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
9344 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
9345 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
9346 clobber operands[2], instead use a new pseudo. Formatting fixes.
9348 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
9350 PR tree-optimization/85586
9351 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
9352 exit early for statements in the same group if the accesses are
9355 2018-05-02 Tom de Vries <tom@codesourcery.com>
9358 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
9361 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
9363 PR tree-optimization/85143
9364 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
9366 2018-05-01 Tom de Vries <tom@codesourcery.com>
9369 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
9370 not found" error message.
9372 2018-05-01 Tom de Vries <tom@codesourcery.com>
9375 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
9376 * vec.c (test_ordered_remove_if): New function.
9377 (vec_c_tests): Call test_ordered_remove_if.
9378 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
9379 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
9380 * tree-vect-patterns.c (vect_pattern_recog_1): Use
9381 VEC_ORDERED_REMOVE_IF.
9383 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9385 PR tree-optimization/82665
9386 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
9387 pointer subtraction where arguments come from a memchr call.
9389 2018-05-01 Jakub Jelinek <jakub@redhat.com>
9391 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
9392 --push-state --as-needed and --pop-state instead of --as-needed and
9393 --no-as-needed if ld supports it.
9394 * configure: Regenerated.
9397 * doc/install.texi2html: Replace _002d with - and _002a with * in
9398 generated html files using sed.
9400 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9403 * gcc-rich-location.c (blank_line_before_p): New function.
9404 (use_new_line): New function.
9405 (gcc_rich_location::add_fixit_insert_formatted): New function.
9406 * gcc-rich-location.h
9407 (gcc_rich_location::add_fixit_insert_formatted): New function.
9409 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9411 * selftest.c (assert_streq): Rename "expected" and "actual" to
9412 "val1" and "val2". Extend NULL-handling to cover both inputs
9413 symmetrically, while still requiring both to be non-NULL for a pass.
9414 * selftest.h (assert_streq): Rename "expected" and "actual" to
9416 (ASSERT_EQ): Likewise.
9417 (ASSERT_EQ_AT): Likewise.
9418 (ASSERT_KNOWN_EQ): Likewise.
9419 (ASSERT_KNOWN_EQ_AT): Likewise.
9420 (ASSERT_NE): Likewise.
9421 (ASSERT_MAYBE_NE): Likewise.
9422 (ASSERT_MAYBE_NE_AT): Likewise.
9423 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
9424 the assertion to pass.
9425 (ASSERT_STREQ_AT): Likewise.
9427 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
9429 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
9430 interaction with -pie.
9432 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9434 * selftest.h: Fix alphabetization of per-source-file selftest
9437 2018-04-30 Jason Merrill <jason@redhat.com>
9439 PR c++/61982 - dead stores to destroyed objects.
9440 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
9443 2018-04-30 Jason Merrill <jason@redhat.com>
9445 * tree.c (build_clobber): New.
9446 * tree.h: Declare it.
9447 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
9449 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9451 * diagnostic-show-locus.c (layout::layout): Update for
9452 location_get_source_line returning a char_span.
9453 (struct char_span): Move to input.h.
9454 (struct correction): Update for fields in char_span becoming
9456 (struct source_line): Update for location_get_source_line
9457 returning a char_span.
9458 (layout::print_line): Likewise.
9459 * edit-context.c (edited_file::print_content): Likewise.
9460 (edited_file::print_diff_hunk): Likewise.
9461 (edited_file::print_run_of_changed_lines): Likewise.
9462 (edited_file::get_num_lines): Likewise.
9463 (edited_line::edited_line): Likewise.
9464 * final.c (asm_show_source): Likewise.
9465 * input.c (location_get_source_line): Convert return type
9466 from const char * to char_span, losing the final "line_len"
9468 (dump_location_info): Update for the above.
9469 (get_substring_ranges_for_loc): Likewise. Use a char_span
9470 when handling the literal within the line.
9471 (test_reading_source_line): Update for location_get_source_line
9472 returning a char_span.
9473 * input.h (class char_span): Move here from
9474 diagnostic-show-locus.c, converting from a struct to a class.
9475 Make data members private.
9476 (char_span::operator bool): New.
9477 (char_span::length): New.
9478 (char_span::get_buffer): New.
9479 (char_span::operator[]): New.
9480 (char_span::subspan): Make const.
9481 (char_span::xstrdup): New.
9482 (location_get_source_line): Convert return type from const char *
9483 to char_span, losing the final "line_size" param.
9485 2018-04-30 Jan Hubicka <jh@suse.cz>
9487 * lto-wrapper.c (ltrans_priorities): New static var.
9488 (cmp_priority): New.
9489 (run_gcc): Read priorities and if doing parallel build order
9490 the Makefile by them.
9492 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9494 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
9496 2018-04-30 Richard Biener <rguenther@suse.de>
9498 * tree-cfg.c (verify_address): Remove base argument, add
9499 flag whether to check TREE_ADDRESSABLE and do that.
9500 (verify_expr): Remove.
9501 (verify_types_in_gimple_reference): Add pieces from verify_expr.
9502 (verify_gimple_assign_single): Likewise.
9503 (verify_gimple_switch): Likewise.
9504 (verify_expr_location_1): Dereference tp once. Add (disabled)
9505 piece from verify_expr.
9506 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
9508 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
9510 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
9512 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
9514 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
9515 (small_data_pattern): Likewise.
9516 (arc_rewrite_small_data): Likewise.
9517 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
9518 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
9519 (get_symbol_alignment): New function.
9520 (legitimate_small_data_address_p): Likewise.
9521 (legitimate_scaled_address): Update, call
9522 legitimate_small_data_address_p.
9523 (output_sdata): New static variable.
9524 (arc_print_operand): Update how we handle small data operands.
9525 (arc_print_operand_address): Likewise.
9526 (arc_legitimate_address_p): Update, use
9527 legitimate_small_data_address_p.
9528 (arc_rewrite_small_data_p): Remove.
9529 (arc_rewrite_small_data_1): Likewise.
9530 (arc_rewrite_small_data): Likewise.
9531 (small_data_pattern): Likewise.
9532 (compact_sda_memory_operand): Update to use
9533 legitimate_small_data_address_p and get_symbol_alignment.
9534 (prepare_move_operands): Don't rewite sdata pattern.
9535 (prepare_extend_operands): Remove.
9536 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
9538 (zero_extendqisi2): Likewise.
9539 (zero_extendhisi2): Likewise.
9540 (extendqihi2): Likewise.
9541 (extendqisi2): Likewise.
9542 (extendhisi2): Likewise.
9546 * config/arc/constraints.md (Usd): Change it to memory constraint.
9548 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
9550 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
9551 as source of std instructions.
9552 * config/arc/arc.md (movsi_insn): Update pattern predicate to
9553 allow 6-bit constants as source for store instructions.
9554 (movdi_insn): Update instruction pattern to allow 6-bit constants
9555 as source for store instructions.
9557 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
9559 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
9561 2018-04-30 Nathan Sidwell <nathan@acm.org>
9562 Sandra Loosemore <sandra@codesourcery.com>
9564 * dumpfile.c (dump_open): Allow '-' for stdout.
9565 * doc/invoke.texi (Developer Options): Document dump filename
9566 determination early. Document stdin/stdout selection.
9568 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
9570 Microblaze Target: PIC data text relative
9572 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
9573 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
9575 * config/microblaze/microblaze.h (microblaze_constant_address_p):
9576 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
9577 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
9578 New addressing mode for data-text relative position indepenedent code.
9579 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
9580 'ADDRESS_SYMBOLIC_TXT_REL'.
9581 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
9582 (microblaze_legitimate_pic_operand): Exclude function calls from
9583 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
9584 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
9586 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
9587 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
9588 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
9589 for 'address + offset'.
9590 (microblaze_expand_prologue): Add new function prologue call for
9592 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
9593 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
9594 table in case of TARGET_PIC_DATA_TEXT_REL.
9595 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
9596 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
9597 Add new macros 'UNSPEC_TEXT',
9598 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
9599 + exclude function calls from 'UNSPEC_PLT' in case of data text
9601 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
9602 new target hook for generating address diff vector tables in case of
9604 * doc/tm.texi : Regenerate.
9605 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
9606 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
9607 of addr diff vector generation.
9608 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
9609 target hook definition.
9610 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
9611 Add default function for generate_pic_addr_diff_vec -> flag_pic.
9612 * doc/invoke.texi (Add new pic option): Add new microblaze pic
9613 option for data text relative.
9615 2018-04-30 Richard Biener <rguenther@suse.de>
9617 * tree-chrec.h (evolution_function_is_constant_p): Remove
9619 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
9621 2018-04-30 Richard Biener <rguenther@suse.de>
9624 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
9626 2018-04-30 Richard Biener <rguenther@suse.de>
9628 PR tree-optimization/28364
9629 PR tree-optimization/85275
9630 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
9631 copying first exit test.
9633 2018-04-28 Mark Wielaard <mark@klomp.org>
9635 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
9637 (dwarf_AT): Handle DW_AT_addr_base.
9638 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
9640 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
9643 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
9644 (*ashl<dwi>3_doubleword_mask_1): Ditto.
9645 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
9646 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
9648 2018-04-28 Richard Biener <rguenther@suse.de>
9650 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
9651 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
9652 to reflect use. Only add interesting stmts.
9654 2018-04-27 Martin Jambor <mjambor@suse.cz>
9657 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
9658 the jump function allows for passing through aggregate values.
9660 2018-04-27 David Malcolm <dmalcolm@redhat.com>
9662 * input.h (in_system_header_at): Convert from macro to inline
9664 (from_macro_expansion_at): Likewise.
9665 (from_macro_definition_at): Likewise.
9667 2018-04-27 Jeff Law <law@redhat.com>
9669 * config.gcc: Mark tile* targets as deprecated/obsolete.
9671 2018-04-27 Richard Biener <rguenther@suse.de>
9673 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
9676 2018-04-27 Richard Biener <rguenther@suse.de>
9678 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
9680 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
9682 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
9683 with Yd constraint. Set "preferred_for_speed" attribute from
9684 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
9686 (*movdi_internal): Ditto.
9687 (movti_interunit splitters): Remove
9688 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
9689 (movdi_interunit splitters): Ditto.
9690 * config/i386/constraints.md (Ye): Remove.
9691 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
9693 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9696 * config/aarch64/constraints.md (Usg): Limit to 31.
9699 2018-04-27 Jakub Jelinek <jakub@redhat.com>
9701 PR tree-optimization/85529
9702 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
9703 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
9704 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
9705 zero extension or masking of the MSB bit.
9706 (optimize_range_tests): Add FIRST_BB argument, pass it through
9707 to optimize_range_tests_var_bound.
9708 (maybe_optimize_range_tests, reassociate_bb): Adjust
9709 optimize_range_tests callers.
9711 2018-04-26 Richard Biener <rguenther@suse.de>
9712 Jakub Jelinek <jakub@redhat.com>
9714 * cgraph.h (symbol_table): Just declare debug method here.
9715 * symtab.c (symbol_table::debug): Define.
9717 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
9719 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
9721 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
9723 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
9724 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
9725 (*movdi_internal): Substitute Yi and Yj constraint with x
9726 and Ym and Yn constraint with y constraint. Update "isa"
9727 attribute and set "preferred_for_speed" attribute from
9728 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
9729 (*movsi_internal): Ditto.
9730 (*movdf_internal): Ditto.
9731 (*movsf_internal): Ditto.
9732 (*zero_extendsidi2): Ditto.
9733 * config/i386/sse.md (vec_set<mode>_0): Ditto.
9734 (sse2_loadld): Ditto.
9735 (*vec_extract<ssevecmodelower>_0): Ditto.
9736 (*vec_extractv4si_0_zext_sse4): Ditto.
9737 (vec_concatv2di): Ditto.
9738 (*vec_dup<mode>): Ditto.
9739 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
9740 * config/i386/constraints.md (Yi): Remove.
9745 2018-04-26 Nathan Sidwell <nathan@acm.org>
9747 * dumpfile.c (dump_open): New.
9748 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
9749 (dump_finish): Detect stdio/stderr by value not name.
9751 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
9753 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
9755 2018-04-26 Tom de Vries <tom@codesourcery.com>
9758 * config/nvptx/nvptx.c (verify_neutering_jumps)
9759 (verify_neutering_labels): New function
9760 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
9762 2018-04-26 Tom de Vries <tom@codesourcery.com>
9765 * config/nvptx/nvptx.c (needs_neutering_p): New function.
9766 (nvptx_single): Use needs_neutering_p to skip over insns that do not
9769 2018-04-26 Richard Biener <rguenther@suse.de>
9770 Tom de Vries <tom@codesourcery.com>
9773 * lto-streamer-out.c (output_function): Fixup loops if required to match
9774 discovery done in the reader.
9776 2018-04-26 Richard Biener <rguenther@suse.de>
9778 PR tree-optimization/85116
9779 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
9780 have a loop exit from the single latch predecessor. Remove
9781 case of header with just condition.
9782 (ch_base::copy_headers): Exclude infinite loops from any
9784 (pass_ch::execute): Record exits.
9786 2018-04-26 Richard Biener <rguenther@suse.de>
9788 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
9789 prologue cost vector and pass it to vect_get_load_cost.
9790 (vect_get_peeling_costs_all_drs): Likewise.
9791 (vect_peeling_hash_get_lowest_cost): Likewise.
9792 (vect_enhance_data_refs_alignment): Likewise.
9794 2018-04-26 Richard Biener <rguenther@suse.de>
9797 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
9798 checking of integer<->pointer conversions.
9799 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
9800 sign-/zero-extending pointer types.
9801 (expand_omp_for_static_chunk): Likewise.
9803 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
9804 Jean Lee <xiaoyur347@gmail.com>
9806 * config/mips/mips.c (mips_asan_shadow_offset): New function.
9807 (TARGET_ASAN_SHADOW_OFFSET): Define.
9808 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
9809 true for -fsanitize=address.
9811 2018-04-25 Mark Wielaard <mark@klomp.org>
9813 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
9816 2018-04-25 Jakub Jelinek <jakub@redhat.com>
9818 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
9819 than "alu", remove explicit "memory" and "imm_disp" attributes.
9820 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
9823 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
9824 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
9825 gen_lowpart_no_emit.
9827 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
9830 * config/i386/i386.c (ix86_expand_builtin): Change memory
9831 operand to XI, extend p0 to Pmode.
9832 * config/i386/i386.md: Change unspec volatile and operand
9833 1 mode to XI, change operand 0 mode to P.
9835 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9837 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
9838 GET_MODE_MASK before any checking.
9839 (nds32_can_use_bset_p): Likewise.
9840 (nds32_can_use_btgl_p): Likewise.
9842 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9844 * config/nds32/nds32-doubleword.md: New define_split pattern for
9845 illegal register number.
9847 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9849 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
9851 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9853 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
9855 2018-04-25 Richard Biener <rguenther@suse.de>
9857 * lto-streamer.h (LTO_major_version): Bump to 8.
9859 2018-04-25 Jakub Jelinek <jakub@redhat.com>
9861 * BASE-VER: Set to 9.0.0.
9863 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
9865 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
9866 in __abskf2 and __powikf2.
9868 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9871 * config/aarch64/constraints.md (Usg, Usj): New constraints.
9872 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
9873 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
9874 Use the above on operand 2. Reindent.
9875 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
9877 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
9880 * common/config/i386/i386-common.c (ix86_handle_option): Don't
9882 * config/i386/i386.opt (mcet): Removed.
9883 * doc/install.texi: Remove -mcet documentation.
9884 * doc/invoke.texi: Likewise.
9886 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
9889 * doc/install.texi: Remove -mcet from bootstrap-cet.
9891 2018-04-24 Jakub Jelinek <jakub@redhat.com>
9894 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
9895 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
9899 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
9900 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
9901 containing a CONST_VECTOR.
9903 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
9905 * doc/install.texi: Update newlib dependency for nvptx.
9907 2018-04-24 Jakub Jelinek <jakub@redhat.com>
9910 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
9911 instead of INTVAL when shifting x left.
9913 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
9915 PR tree-optimization/85478
9916 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
9917 vect_grouped_store_supported for single element vectors.
9919 2018-04-24 Richard Biener <rguenther@suse.de>
9922 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
9923 load cost increase to the case of non-constant step.
9925 2018-04-24 Jakub Jelinek <jakub@redhat.com>
9928 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
9929 destination if any_malformed_asm.
9931 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
9934 * expr.c (store_field): In the bitfield case, if the value comes from
9935 a function call and is returned in registers by means of a PARALLEL,
9936 do not change the mode of the temporary unless BLKmode and VOIDmode.
9938 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
9940 PR rtl-optimization/85423
9941 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
9942 dependencies to debug insns when the previous insn is non-debug.
9944 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
9946 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
9947 enums into a single definition.
9948 (fls): Fix predicates and printing.
9951 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
9953 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
9954 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
9955 and short u6 immediate.
9956 (check_if_valid_sleep_operand): Remove.
9957 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
9959 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9961 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
9962 flag_always_save_lp condition.
9963 * config/nds32/nds32.opt (malways-save-lp): New option.
9965 2018-04-22 Shiva Chen <shiva0217@gmail.com>
9967 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
9968 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
9969 * config/nds32/nds32.h
9970 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
9971 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
9973 2018-04-22 Shiva Chen <shiva0217@gmail.com>
9975 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
9976 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
9978 2018-04-22 Shiva Chen <shiva0217@gmail.com>
9979 Chung-Ju Wu <jasonwucj@gmail.com>
9981 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
9983 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
9984 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
9986 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9988 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
9990 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9992 * config/nds32/nds32-protos.h (nds32_data_alignment,
9993 nds32_local_alignment): Declare.
9994 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
9995 nds32_local_alignment): New functions.
9996 (TARGET_CONSTANT_ALIGNMENT): Define.
9997 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
9999 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10001 * config/nds32/nds32.c
10002 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
10003 (TARGET_MODES_TIEABLE_P): Likewise.
10005 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10007 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
10008 level Ofast and Og.
10010 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
10011 Chung-Ju Wu <jasonwucj@gmail.com>
10013 * config/nds32/constants.md (unspec_volatile_element): Add enum values
10014 for unaligned access.
10015 * config/nds32/nds32-intrinsic.c: Implementation of expanding
10017 * config/nds32/nds32-intrinsic.md: Likewise.
10018 * config/nds32/nds32_intrinsic.h: Likewise.
10019 * config/nds32/nds32.h (nds32_builtins): Likewise.
10020 * config/nds32/nds32.opt (munaligned-access): New option.
10021 * config/nds32/nds32.c (nds32_asm_file_start): Display
10022 flag_unaligned_access status.
10024 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
10026 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
10027 -mno-relax is present.
10028 * config/riscv/linux.h (LINK_SPEC): Ditto.
10030 2018-04-20 Martin Sebor <msebor@redhat.com>
10033 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
10035 (gimple_fold_builtin_stxcpy_chk): Same.
10036 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
10038 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
10041 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
10042 __powikf2 when long double is IEEE 128-bit.
10044 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
10046 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
10047 step to make sure stack always aligned.
10049 2018-04-20 Carl Love <cel@us.ibm.com>
10052 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
10053 size check for arg0.
10055 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
10056 Tom de Vries <tom@codesourcery.com>
10059 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
10060 Emit insns for calls too.
10061 (nvptx_find_par): Always look for worker-level predecessor insn.
10062 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
10064 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
10065 (nvptx_process_pars): Propagate frames for calls.
10067 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
10070 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
10072 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
10073 (ix86_handle_option): Don't handle OPT_mibt.
10074 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
10076 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
10078 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
10079 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
10080 (ix86_target_macros): Define __CET__ with flag_cf_protection
10081 for -fcf-protection.
10082 * config/i386/i386.c (isa2_opts): Remove -mibt.
10083 * config/i386/i386.h (TARGET_IBT): Removed.
10084 (TARGET_IBT_P): Likewise.
10085 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
10086 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
10087 * config/i386/i386.opt (mcet): Update help message.
10088 (mshstk): Likewise.
10090 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
10091 -mcet as an alias for -mshstk.
10093 2018-04-20 Richard Biener <rguenther@suse.de>
10095 PR middle-end/85475
10096 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
10097 complexity by forcing a single use of the multiply operand.
10099 2018-04-20 Martin Jambor <mjambor@suse.cz>
10102 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
10103 recursion dependency to only apply to non-clones.
10105 2018-04-20 Martin Jambor <mjambor@suse.cz>
10108 * ipa-cp.c (create_specialized_node): Check that clones of
10109 self-recursive edges exist during IPA-CP.
10111 2018-04-19 Toon Moene <toon@moene.org>
10113 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
10116 2018-04-19 Jakub Jelinek <jakub@redhat.com>
10118 PR tree-optimization/85467
10119 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
10120 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
10121 VECTOR_CST element to type.
10123 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
10126 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
10127 * config/i386/i386.md (builtin_setjmp_setup): Removed.
10128 (builtin_longjmp): Likewise.
10129 (save_stack_nonlocal): New pattern.
10130 (restore_stack_nonlocal): Likewise.
10132 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
10135 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
10136 Replace ASM_OUTPUT_LABEL with fprintf.
10138 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
10141 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
10142 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
10143 * config/i386/i386-c.c (ix86_target_macros_internal): Also
10144 define __IBT__ and __SHSTK__ for -fcf-protection.
10145 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
10147 (ix86_trampoline_init): Likewise.
10148 (x86_output_mi_thunk): Likewise.
10149 (ix86_notrack_prefixed_insn_p): Likewise.
10150 (ix86_option_override_internal): Don't disallow -fcf-protection.
10151 * config/i386/i386.md (rdssp<mode>): Also enable for
10153 (incssp<mode>): Likewise.
10154 (nop_endbr): Likewise.
10155 * config/i386/i386.opt (mcet): Change help message to built-in
10158 (mshstk): Likewise.
10159 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
10160 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
10161 enable CET built-in functions.
10163 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
10165 * common/config/i386/i386-common.c
10166 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
10167 OPTION_MASK_ISA_MOVDIRI_UNSET,
10168 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
10169 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
10170 * config.gcc (movdirintrin.h): New header.
10171 * config/i386/cpuid.h (bit_MOVDIRI,
10172 bit_MOVDIR64B): New bits.
10173 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
10175 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
10176 (VOID, PVOID, PCVOID)): New function types.
10177 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
10178 __builtin_ia32_directstoreu_u64,
10179 __builtin_ia32_movdir64b): New builtins.
10180 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
10181 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
10183 (ix86_valid_target_attribute_inner_p): Ditto.
10184 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
10185 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
10186 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
10187 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
10188 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
10189 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
10190 (movdiri<mode>, movdir64b_<mode>): New.
10191 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
10192 * config/i386/immintrin.h: Include movdirintrin.h.
10193 * config/i386/movdirintrin.h: New file.
10194 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
10196 2018-04-19 Richard Biener <rguenther@suse.de>
10198 PR middle-end/85455
10199 * cfg.c (clear_bb_flags): When loop state says we have
10200 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
10202 2018-04-19 Richard Biener <rguenther@suse.de>
10204 PR tree-optimization/84737
10205 * tree-vect-data-refs.c (vect_copy_ref_info): New function
10206 copying restrict info.
10207 (vect_setup_realignment): Use it.
10208 * tree-vectorizer.h (vect_copy_ref_info): Declare.
10209 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
10210 the first DR to all generated stores.
10211 (vectorizable_load): Likewise for loads.
10213 2018-04-19 Jakub Jelinek <jakub@redhat.com>
10215 PR tree-optimization/85446
10216 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
10217 the integral and pointer types to have the same precision.
10219 * doc/install.texi: Document --disable-cet being the default and
10222 2018-04-18 Martin Liska <mliska@suse.cz>
10224 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
10227 2018-04-18 Martin Liska <mliska@suse.cz>
10230 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
10233 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
10234 arguments if they are comparable.
10236 2018-04-18 Martin Liska <mliska@suse.cz>
10239 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
10242 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
10245 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
10248 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
10249 ENDBR after calling __morestack.
10251 2018-04-18 David Malcolm <dmalcolm@redhat.com>
10254 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
10255 by using gcc_base_ver to generate a gcc_driver_version, and use
10256 it when generating GCC_DRIVER_NAME.
10257 * configure: Regenerate.
10259 2018-04-18 Jakub Jelinek <jakub@redhat.com>
10262 * config.gcc: Obsolete powerpc*-*-*spe*.
10264 2018-04-17 Jakub Jelinek <jakub@redhat.com>
10267 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
10268 (stabstr_D): Change type of unum from unsigned int to
10269 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
10272 2018-04-17 Jim Wilson <jimw@sifive.com>
10275 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
10276 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
10277 Set arg_pointer_offset after using pretend_args_size.
10279 2018-04-17 Jakub Jelinek <jakub@redhat.com>
10281 PR rtl-optimization/85431
10282 * dse.c (record_store): Ignore zero width stores.
10285 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
10286 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
10287 __builtin_stack_restore rather than after it.
10288 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
10289 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
10290 argument instead of virtual_dynamic_stack_rtx.
10292 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
10294 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
10296 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
10297 Add note to error message to explain internal mapping of overloaded
10298 built-in function name to non-overloaded built-in function name.
10299 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
10302 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10305 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
10306 where the inputs overlap with the output.
10308 2018-04-17 Jakub Jelinek <jakub@redhat.com>
10310 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
10311 (=v, v) alternative and explicit "memory" attribute.
10312 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
10313 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
10315 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
10316 "sselog1" type instead of "sselog".
10317 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
10318 "sselog". Remove explicit "memory" attribute.
10319 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
10320 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
10322 (vec_extract_hi_v32hi): Merge all alternatives into one, use
10323 "sselog1" type instead of "sselog". Remove explicit "memory"
10325 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
10326 use "sselog1" type instead of "sselog". Remove explicit "memory"
10328 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
10329 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
10331 (vec_extract_hi_v64qi): Merge all alternatives into one, use
10332 "sselog1" type instead of "sselog". Remove explicit "memory"
10334 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
10335 use "sselog1" type instead of "sselog". Remove explicit "memory"
10339 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
10341 PR middle-end/85414
10342 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
10345 2018-04-17 Martin Jambor <mjambor@suse.cz>
10348 * ipa-cp.c (create_specialized_node): Call
10349 expand_all_artificial_thunks if necessary.
10351 2018-04-17 Martin Liska <mliska@suse.cz>
10354 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
10355 in message, remote space in between '_G' and '('.
10357 2018-04-17 Jakub Jelinek <jakub@redhat.com>
10360 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
10361 avx512f_vmcmp<mode>3<round_saeonly_name>,
10362 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
10363 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
10364 avx512f_rndscale<mode><round_saeonly_name>,
10365 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
10366 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
10367 Use %<iptr>2 instead of %2 for -masm=intel.
10368 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
10369 avx512f_vcvttss2usi<round_saeonly_name>,
10370 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
10372 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
10373 avx512f_vcvttsd2usi<round_saeonly_name>,
10374 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
10375 Use %q1 instead of %1 for -masm=intel.
10376 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
10377 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
10378 of %3 for -masm=intel.
10379 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
10381 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
10383 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
10385 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
10386 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
10388 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
10390 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
10391 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
10392 %g1 and one with %0 and %1.
10393 (avx512er_vmrcp28<mode><round_saeonly_name>,
10394 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
10395 %1 for -masm=intel.
10396 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
10397 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
10398 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
10399 of %0 and %{%4%} for -masm=intel.
10400 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
10401 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
10402 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
10403 order of %0 and %{%5%}%{z%} for -masm=intel.
10405 2018-04-17 Jan Hubicka <jh@suse.cz>
10408 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
10410 2018-04-17 Martin Liska <mliska@suse.cz>
10413 * multiple_target.c (create_dispatcher_calls): Set apostrophes
10414 for target_clone error message. Make default implementation
10415 clone to be a local declaration.
10416 (separate_attrs): Add new argument and check for an empty
10418 (expand_target_clones): Handle it.
10419 (ipa_target_clone): Make redirection just for target_clones
10422 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
10423 Tom de Vries <tom@codesourcery.com>
10425 PR middle-end/84955
10426 * omp-expand.c (expand_oacc_for): Add dummy false branch for
10427 tiled basic blocks without omp continue statements.
10429 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
10432 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
10433 vec_extract expression as having side effects to make sure it gets
10436 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
10439 * config/i386/i386.c (get_builtin_code_for_version): Check
10442 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
10445 * config.gcc: Support "skylake".
10446 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10448 * config/i386/i386.c (m_SKYLAKE): Define.
10449 (processor_target_table): Add "skylake".
10450 (ix86_option_override_internal): Add "skylake".
10451 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
10452 PROCESSOR_CANNONLAKE.
10453 (get_builtin_code_for_version): Fix priority for
10454 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
10455 PROCESSOR_SKYLAKE-AVX512.
10456 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
10457 (processor_type): Add PROCESSOR_SKYLAKE.
10459 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
10460 Jason Merrill <jason@redhat.com>
10463 * convert.c (convert_to_integer_1): Use direct recursion for
10464 enumeral types and types with a precision less than the number
10465 of bits in their mode.
10467 2018-04-16 Julia Koval <julia.koval@intel.com>
10470 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
10471 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
10473 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
10476 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
10477 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
10478 and -mno-direct-move.
10480 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
10483 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
10484 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
10485 Ensure negative shifts result in {0}.
10487 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
10489 PR rtl-optimization/79916
10490 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
10491 regs (if any) to define how to gnerate SD moves when LRA is in
10494 2018-04-13 Jakub Jelinek <jakub@redhat.com>
10496 PR rtl-optimization/85393
10497 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
10498 * except.c (expand_dw2_landing_pad_for_region): Make static.
10499 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
10500 a label and unconditional jump to old_bb, rather than
10501 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
10504 PR rtl-optimization/85376
10505 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
10506 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
10507 instead of a specific value.
10509 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
10510 Bin Cheng <bin.cheng@arm.com>
10512 PR tree-optimization/82965
10513 PR tree-optimization/83991
10514 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
10515 by_profile_only parameter.
10516 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
10517 information if the loop was predicted to iterate too many times.
10518 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
10520 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
10523 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
10526 2018-04-13 Martin Liska <mliska@suse.cz>
10527 Jakub Jelinek <jakub@redhat.com>
10529 PR middle-end/81657
10530 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
10531 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
10532 * builtins.c (expand_builtin_memory_copy_args): Use
10533 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
10534 handle dest_addr == pc_rtx.
10536 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
10539 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
10540 asked to not generate direct moves.
10541 (fix_trunc<mode>si2_stfiwx): Similar.
10542 (fix_trunc<mode>si2_internal): Similar.
10544 2018-04-12 Jakub Jelinek <jakub@redhat.com>
10547 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
10548 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
10549 lookup if dest in some wider mode is known to be const0_rtx and
10550 if so, record permanent equivalence for it to be ZERO_EXTEND of
10551 the narrower mode destination.
10553 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
10555 * lto-streamer-out.c (output_function): Revert 259346.
10556 * omp-expand.c (expand_oacc_for): Likewise.
10558 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
10560 PR rtl-optimization/85354
10561 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
10562 * sel-sched.c (sel_global_init): ... here.
10564 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
10567 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
10568 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
10569 mode for PE-COFF targets.
10570 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
10571 (i386_pe_asm_lto_end): Likewise.
10572 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
10573 (TARGET_ASM_LTO_END): Likewise.
10574 * config/i386/winnt.c (saved_debug_info_level): New static variable.
10575 (i386_pe_asm_lto_start): New function.
10576 (i386_pe_asm_lto_end): Likewise.
10578 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
10579 Richard Biener <rguenther@suse.de>
10581 PR middle-end/84955
10582 * lto-streamer-out.c (output_function): Fix CFG loop state before
10584 * omp-expand.c (expand_oacc_for): Handle calls to internal
10585 functions like regular functions.
10587 2018-04-12 Richard Biener <rguenther@suse.de>
10590 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
10591 for the early LTO debug to properly generate references to it
10592 during DIE emission. Do not re-use that for the skeleton for
10594 (dwarf2out_early_finish): Likewise.
10596 2018-04-12 Jakub Jelinek <jakub@redhat.com>
10599 * config/i386/sse.md
10600 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
10601 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
10602 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
10603 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
10604 and output is a reg, avoid creating invalid lowpart subreg, but
10605 instead split into a 512-bit move. Don't split if not AVX512VL,
10606 input is xmm16+ reg and output is a mem.
10607 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
10608 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
10609 xmm16+ reg and output is a mem.
10611 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10613 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
10614 also for flag_dwarf2_cfi_asm.
10616 2018-04-12 Jakub Jelinek <jakub@redhat.com>
10618 PR rtl-optimization/85342
10619 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
10620 a bool scalar var inside of the loop instead. Don't try to update
10621 recog_data.operand after failed apply_change_group.
10623 2018-04-12 Tom de Vries <tom@codesourcery.com>
10626 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
10627 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
10628 array with flexible array member as array without given dimension.
10629 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
10630 argument for undefined param to true.
10632 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
10635 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
10636 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
10637 from PowerPC section.
10638 * config/rs6000/sysv4.opt (mcall-): Improve help text.
10639 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
10640 help text that is too long.
10641 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
10642 help text that is too long.
10643 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
10644 help text that is too long.
10646 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
10648 * config/alpha/alpha.md (stack_probe_internal): Rename
10649 from "probe_stack". Update all callers.
10651 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
10653 PR rtl-optimization/84566
10654 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
10655 sched_macro_fuse_insns.
10657 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
10660 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
10661 (compute_block_dependences): ... from here.
10663 2018-04-11 Jakub Jelinek <jakub@redhat.com>
10665 PR tree-optimization/85331
10666 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
10667 from int to HOST_WIDE_INT.
10669 2018-04-11 Martin Jambor <mjambor@suse.cz>
10672 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
10673 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
10674 not the same as the source val.
10675 (cgraph_edge_brings_value_p): New parameter.
10676 (gather_edges_for_value): Pass destination value to
10677 cgraph_edge_brings_value_p.
10678 (perhaps_add_new_callers): Likewise.
10679 (get_info_about_necessary_edges): Likewise and exclude values brought
10680 only by self-recursive edges.
10681 (create_specialized_node): Redirect only clones of self-calling edges.
10682 (+self_recursive_pass_through_p): New function.
10683 (find_more_scalar_values_for_callers_subset): Use it.
10684 (find_aggregate_values_for_callers_subset): Likewise.
10685 (known_aggs_to_agg_replacement_list): Removed.
10686 (decide_whether_version_node): Re-calculate known constants for all
10687 remaining context clones.
10689 2018-04-11 Richard Biener <rguenther@suse.de>
10692 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
10693 from early DWARF output.
10694 (dwarf2out_early_finish): Output line info unconditionally into
10695 early DWARF and add reference to it.
10697 2018-04-11 Jakub Jelinek <jakub@redhat.com>
10700 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
10701 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
10702 other than V2DFmode using iptr mode attribute.
10703 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
10705 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
10707 PR rtl-optimization/84659
10708 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
10710 2018-04-11 Jakub Jelinek <jakub@redhat.com>
10713 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
10715 (output_loc_list): Pass address of a dummy size variable even in the
10716 locview handling loop.
10717 (index_location_lists): Add comment on why skip_loc_list_entry can't
10720 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
10723 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
10726 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
10729 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
10730 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
10731 and -mstring-compare-inline-limit.
10733 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
10736 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
10737 for stack clash protection in a register whenever we need it to be in
10740 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
10742 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
10743 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
10745 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
10748 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
10750 (mlong-double-): Ditto.
10751 * config/rs6000/sysv4.opt (msdata=): Ditto.
10752 (mtls-size=): Ditto.
10754 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10756 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
10757 erroneous entries for
10758 "vector int vec_ldl (int, long int *)", and
10759 "vector unsigned int vec_ldl (int, unsigned long int *)".
10760 Add comments and entries for
10761 "vector bool char vec_ldl (int, bool char *)",
10762 "vector bool short vec_ldl (int, bool short *)",
10763 "vector bool int vec_ldl (int, bool int *)",
10764 "vector bool long long vec_ldl (int, bool long long *)",
10765 "vector pixel vec_ldl (int, pixel *)",
10766 "vector long long vec_ldl (int, long long *)",
10767 "vector unsigned long long vec_ldl (int, unsigned long long *)".
10768 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
10769 type tree bool_long_long_type_node and correct definition of
10770 bool_V2DI_type_node to make reference to this new type tree.
10771 (rs6000_mangle_type): Replace erroneous reference to
10772 bool_long_type_node with bool_long_long_type_node.
10773 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
10774 comments to emphasize sign distinctions for char and int types and
10775 replace RS6000_BTI_bool_long constant with
10776 RS6000_BTI_bool_long_long constant. Also add comment to restrict
10777 use of RS6000_BTI_pixel.
10778 (bool_long_type_node): Remove this macro definition.
10779 (bool_long_long_type_node): New macro definition
10781 2018-04-10 Jakub Jelinek <jakub@redhat.com>
10783 PR rtl-optimization/85300
10784 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
10785 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
10786 simplify_unary_operation fails.
10788 2018-04-10 Martin Liska <mliska@suse.cz>
10790 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
10791 cgraph_edge and ipa_ref.
10793 2018-04-10 Jakub Jelinek <jakub@redhat.com>
10797 * config/i386/sse.md
10798 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
10799 computation of the VEC_MERGE selector from mask.
10800 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
10801 Fix decoding of the VEC_MERGE selector into mask.
10803 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
10805 PR tree-optimization/85286
10806 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
10808 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
10810 * final.c (final_1): Set insn_last_address as well as
10811 insn_current_address.
10813 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10816 * explow.c (emit_stack_probe): Call validize_mem on memory location
10817 before passing it to gen_probe_stack. Create address operand and
10818 legitimize it for the probe_stack_address case.
10820 2018-04-09 Jan Hubicka <jh@suse.cz>
10823 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
10824 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
10825 * tree.c (free_lang_data_in_type): Fix handling of binfos;
10827 (free_lang_data): Rebuild type inheritance graph.
10829 2018-04-09 Martin Sebor <msebor@redhat.com>
10831 * invoke.texi (-finline-small-functions): Mention other optimization
10833 (-findirect-inlining, -fpartial-inlining): Same.
10834 (-finline-functions-called-once): Same.
10835 (-freorder-blocks-and-partition): Same.
10837 2018-04-09 Jan Hubicka <jh@suse.cz>
10840 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
10841 jumps; choose last target that matches the criteria (i.e.
10842 no partition changes for non-crossing jumps).
10843 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
10844 support for redirecting crossing jumps to non-crossing.
10846 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
10848 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
10849 also for naked functions.
10851 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
10853 * config/arc/arc.md (add_shift): New pattern.
10854 (add_shift2): Likewise.
10855 (sub_shift): Likewise.
10856 (sub_shift_cmp0_noout): Likewise.
10857 (compare_si_ashiftsi): Likewise.
10858 (xbfu_cmp0_noout): New combine pattern.
10859 (xbfu_cmp0"): Likewise.
10860 (movsi_set_cc_insn): Place the predicable variant first.
10861 (commutative_binary_cmp0_noout): Remove clobber.
10862 (commutative_binary_cmp0): New pattern.
10863 (noncommutative_binary_cmp0): Likewise.
10864 (noncommutative_binary_cmp0_noout): Likewise.
10865 (noncommutative_binary_comparison_result_used): Removed.
10866 (rsub_cmp0): New pattern.
10867 (rsub_cmp0_noout): Likewise.
10868 (extzvsi): Changed, keep only meaningful variants.
10869 (SQH, SEZ): New iterators.
10870 (SQH_postfix): New mode attribute.
10871 (SEZ_prefix): New code attribute.
10872 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
10873 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
10874 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
10875 of numerical value.
10876 (noncommutative_operator): Check the availability of barrel
10879 2018-04-09 Richard Biener <rguenther@suse.de>
10881 PR tree-optimization/85284
10882 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
10883 Only use the niter constraining form of simple_iv when the exit
10884 is always executed.
10886 2018-04-09 Tom de Vries <tom@codesourcery.com>
10889 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
10890 (define_expand "*memory_barrier"): New define_expand.
10891 (define_insn "memory_barrier"): New insn.
10893 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10895 PR rtl-optimization/80463
10896 PR rtl-optimization/83972
10897 PR rtl-optimization/83480
10899 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
10900 correct producer for the insn.
10901 (tidy_control_flow): Fixup seqnos in case of debug insns.
10903 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10905 PR rtl-optimization/83913
10907 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
10908 different sched-times when merging exprs.
10910 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10912 PR rtl-optimization/83962
10914 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
10915 tidy_fallthru_edge and tidy_control_flow.
10917 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10919 PR rtl-optimization/83530
10921 * sel-sched.c (force_next_insn): New global variable.
10922 (remove_insn_for_debug): When force_next_insn is true, also leave only
10923 next insn in the ready list.
10924 (sel_sched_region): When the region wasn't scheduled, make another pass
10925 over it with force_next_insn set to 1.
10927 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
10929 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
10931 * config/nds32/constants.md (unspec_volatile_element): Add enum values
10932 for interrupt control.
10933 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
10934 functions for interrupt control.
10935 * config/nds32/nds32-intrinsic.md: Likewise.
10936 * config/nds32/nds32_intrinsic.h: Likewise.
10937 * config/nds32/nds32.h (nds32_builtins): Likewise.
10939 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
10941 * config/nds32/nds32.c (nds32_init_machine_status,
10942 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
10943 strict_aligned_p field.
10944 (nds32_expand_to_rtl_hook): New function.
10945 (TARGET_EXPAND_TO_RTL_HOOK): Define.
10946 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
10948 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10949 Chung-Ju Wu <jasonwucj@gmail.com>
10951 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
10952 * config/nds32/nds32-n7.md: New file.
10953 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
10954 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
10956 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
10957 * config/nds32/nds32.md (pipeline_model): Add n7.
10958 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
10959 * config/nds32/pipelines.md: Include n7 settings.
10961 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10962 Chung-Ju Wu <jasonwucj@gmail.com>
10964 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
10965 * config/nds32/nds32-e8.md: New file.
10966 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
10967 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
10969 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
10970 * config/nds32/nds32.md (pipeline_model): Add e8.
10971 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
10972 * config/nds32/pipelines.md: Include e8 settings.
10974 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10975 Chung-Ju Wu <jasonwucj@gmail.com>
10977 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
10978 * config/nds32/nds32-n8.md: New file.
10979 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
10980 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
10982 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
10983 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
10984 * config/nds32/nds32.md (pipeline_model): Add n8.
10985 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
10986 * config/nds32/pipelines.md: Include n8 settings.
10988 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10989 Chung-Ju Wu <jasonwucj@gmail.com>
10991 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
10992 * config/nds32/nds32-n9-2r1w.md: New file.
10993 * config/nds32/nds32-n9-3r2w.md: New file.
10994 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
10995 nds32_register_ports): New or modify for cpu n9.
10996 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
10998 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
10999 * config/nds32/nds32-utils.c: New file.
11000 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
11001 TARGET_MUL_SLOW): Define.
11002 * config/nds32/nds32.md (pipeline_model): New attribute.
11003 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
11004 New options that support cpu n9.
11005 * config/nds32/pipelines.md: Include n9 settings.
11006 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
11008 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
11010 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
11011 information if necessary.
11012 (output_cond_branch_compare_zero): Likewise.
11013 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
11014 (nds32_target_alignment): Refine for alignment.
11015 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
11016 (FUNCTION_BOUNDARY): Modify.
11017 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
11019 * config/nds32/nds32.opt (malways-align, malign-functions): New.
11021 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
11023 * config/nds32/constants.md (unspec_volatile_element): Add values for
11024 TLB operation and data prefetch.
11025 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
11026 functions for TLB operation and data prefetch.
11027 * config/nds32/nds32-intrinsic.md: Likewise.
11028 * config/nds32/nds32_intrinsic.h: Likewise.
11029 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
11030 (nds32_print_operand): Likewise.
11031 * config/nds32/nds32.h (nds32_builtins): Likewise.
11033 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
11034 Andrew Pinski <pinsika@gcc.gnu.org>
11036 PR middle-end/82976
11037 * match.pd: Use constant_boolean_node of correct type instead of
11038 boolean_true_node or boolean_false_node for simplifying
11039 pointer comparisons to zero.
11041 2018-04-07 Jakub Jelinek <jakub@redhat.com>
11043 PR tree-optimization/80021
11044 * tree.c (verify_type_variant): Make error call in verify_variant_match
11045 translatable and remove final full stop.
11047 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
11049 * config/nds32/constants.md (unspec_volatile_element): Add
11050 UNSPEC_VOLATILE_EH_RETURN.
11051 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
11052 nds32_output_stack_pop): Support dwarf exception handling process.
11053 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
11054 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
11055 exception handling process.
11056 (nds32_compute_stack_frame): Likewise.
11057 (nds32_return_addr_rtx): Likewise.
11058 (nds32_initial_elimination_offset): Likewise.
11059 (nds32_expand_prologue): Likewise.
11060 (nds32_expand_epilogue): Likewise.
11061 (nds32_dynamic_chain_address): New function.
11062 * config/nds32/nds32.h (machine_function): Add fields for dwarf
11063 exception handling.
11064 (DYNAMIC_CHAIN_ADDRESS): Define.
11065 (EH_RETURN_DATA_REGNO): Define.
11066 (EH_RETURN_STACKADJ_RTX): Define.
11067 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
11068 patterns for dwarf exception handling.
11070 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
11072 * config/nds32/nds32.h: Clean up obsolete macros.
11074 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11076 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
11077 Add enum values for particular instructions.
11078 * config/nds32/nds32-intrinsic.c: Implementation of expanding
11079 particular intrinsic functions.
11080 * config/nds32/nds32-intrinsic.md: Likewise.
11081 * config/nds32/nds32_intrinsic.h: Likewise.
11082 * config/nds32/nds32.h (nds32_builtins): Likewise.
11083 * config/nds32/nds32.md (type): Add pbsad and pbsada.
11084 (btst, ave): New patterns for particular instructions.
11086 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11088 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
11089 Add enum values for atomic load/store and memory sync.
11090 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
11092 * config/nds32/nds32-intrinsic.md: Likewise.
11093 * config/nds32/nds32_intrinsic.h: Likewise.
11094 * config/nds32/nds32.h (nds32_builtins): Likewise.
11096 2018-04-07 Jakub Jelinek <jakub@redhat.com>
11098 PR tree-optimization/85257
11099 * fold-const.c (native_encode_vector): If not all elts could fit
11100 and off is -1, return 0 rather than offset.
11101 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
11102 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
11103 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
11104 adjust buffer in native_interpret_expr call.
11106 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11108 * config/nds32/constants.md (unspec_volatile_element): Add cache
11109 control enum values.
11110 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
11111 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
11112 * config/nds32/nds32.c (nds32_cctl_names): New.
11113 (nds32_print_operand): Handle cache control register names.
11114 * config/nds32/nds32.h (nds32_builtins): New enum values.
11115 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
11117 * config/nds32/nds32.md (type): Add mmu.
11118 * config/nds32/pipelines.md (simple_insn): Add mmu.
11120 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
11122 * config/nds32/nds32.md (type): Remove call.
11123 * config/nds32/pipelines.md (simple_insn): Likewise.
11125 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11127 * config/nds32/constants.md (unspec_volatile_element): Add
11128 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
11129 UNSPEC_VOLATILE_FMFCFG.
11130 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
11131 description for fmfcfg and fmfcsr.
11132 (bdesc_1arg): Add fmtcsr.
11133 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
11134 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
11135 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
11136 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
11137 unspec_fmfcfg): New patterns.
11138 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
11139 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
11140 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
11141 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
11142 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
11143 __nds32__fmfcfg): Define.
11145 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11147 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
11148 intrinsic register names.
11149 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
11150 intrinsic register enum values and macros.
11152 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
11154 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
11155 for load/store addressing form.
11156 (nds32_print_operand_address): Likewise.
11158 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
11161 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
11162 based on LABEL_REF. Remove useless assertion.
11163 (pic_address_needs_scratch): Fix formatting.
11164 (sparc_legitimize_pic_address): Minor tweaks.
11165 (sparc_delegitimize_address): Adjust assertion accordingly.
11166 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
11167 into symbolic_operand.
11168 (movsi_high_pic_label_ref): Likewise.
11169 (movsi_lo_sum_pic_label_ref): Likewise.
11170 (movdi_pic_label_ref): Likewise.
11171 (movdi_high_pic_label_ref): Likewise.
11172 (movdi_lo_sum_pic_label_ref): Likewise.
11174 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
11176 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
11177 custom LIB_SPEC setup.
11179 2018-04-06 Ruslan Bukin <br@bsdpad.com>
11180 Kito Cheng <kito.cheng@gmail.com>
11182 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
11183 * config/riscv/freebsd.h: New.
11185 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
11187 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
11188 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
11191 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
11192 Kito Cheng <kito.cheng@gmail.com>
11194 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
11195 nds32_output_call, nds32_symbol_binds_local_p): New functions.
11196 * config/nds32/nds32-protos.h (nds32_output_call,
11197 nds32_output_return): Declare.
11198 * config/nds32/nds32.md: Refine all the call and return patterns.
11200 2018-04-06 Jakub Jelinek <jakub@redhat.com>
11203 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
11204 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
11206 PR rtl-optimization/84872
11207 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
11208 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
11209 EDGE_CROSSING edge.
11211 2018-04-06 Tamar Christina <tamar.christina@arm.com>
11213 * expr.c (copy_blkmode_to_reg): Revert 254862.
11214 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
11216 2018-04-06 Richard Biener <rguenther@suse.de>
11218 PR middle-end/85244
11219 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
11220 after seeing a component reference with an adjacent field. Treat
11221 refs to arrays at struct end of external decls similar to
11222 refs to unconstrained commons.
11224 2018-04-06 Jakub Jelinek <jakub@redhat.com>
11227 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
11228 look through SAVE_EXPRs with non-side-effects argument. Adjust
11230 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
11233 2018-04-06 Richard Biener <rguenther@suse.de>
11235 PR middle-end/85180
11236 * alias.c (find_base_term): New wrapper around find_base_term
11237 unwinding CSELIB_VAL_PTR changes.
11238 (find_base_term): Do not restore CSELIB_VAL_PTR during the
11241 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11243 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
11245 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
11246 constant definitions.
11247 ("nop"): lr 0,0 -> nopr r0
11248 ("nop_lr0", "nop_lr1"): New insn definitions.
11250 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
11252 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
11253 NDS32_V3PUSH_AVAILABLE_P macro.
11255 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
11256 Chung-Ju Wu <jasonwucj@gmail.com>
11258 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
11259 (nds32*-*-*): Add float and fpu_config into supported_defaults.
11260 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
11261 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
11262 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
11263 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
11264 * config/nds32/constraints.md: New constraints and checking for hard
11265 float configuration.
11266 * config/nds32/iterators.md: New mode iterator and attribute for hard
11267 float configuration.
11268 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
11270 * config/nds32/nds32-fpu.md: New file.
11271 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
11272 deal with hard float code generation.
11273 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
11275 (abi_type, float_reg_number): New enum type.
11276 * config/nds32/nds32-predicates.c: New predicates for hard float.
11277 * config/nds32/nds32-protos.h: Declare functions for hard float.
11278 * config/nds32/nds32.c: Implementation for hard float configuration.
11279 * config/nds32/nds32.h: Definitions for hard float configuration.
11280 * config/nds32/nds32.md: Include hard float machine description and
11281 modify patterns for hard float configuration.
11282 * config/nds32/nds32.opt: New options for hard float configuration.
11283 * config/nds32/predicates.md: New predicates for hard float
11286 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
11288 * common/config/nds32/nds32-common.c
11289 (nds32_option_optimization_table): Enable -mreleax-hint by default.
11291 2018-04-05 Jakub Jelinek <jakub@redhat.com>
11293 PR middle-end/85195
11294 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
11295 CONSTRUCTOR_ELT (ctor, ...)->value.
11297 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
11300 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
11302 2018-04-05 Tom de Vries <tom@codesourcery.com>
11305 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
11308 2018-04-05 Shiva Chen <shiva0217@gmail.com>
11309 Kito Cheng <kito.cheng@gmail.com>
11311 * config/nds32/constraints.md (U33): Fine-tune checking condition.
11312 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
11313 * config/nds32/nds32.h (nds32_16bit_address_type): Add
11314 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
11316 2018-04-05 Shiva Chen <shiva0217@gmail.com>
11317 Kito Cheng <kito.cheng@gmail.com>
11319 * config/nds32/constraints.md (Ufe): New memory constraint.
11320 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
11321 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
11322 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
11324 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
11325 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
11327 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11329 * config/nds32/nds32.md: Use optimize_size in the condition for
11330 alu-shift instructions.
11332 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11334 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
11336 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11338 * config/nds32/nds32.md (negsi2): Refine pattern.
11340 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
11341 Chung-Ju Wu <jasonwucj@gmail.com>
11343 * config/nds32/iterators.md (shift_rotate): New code iterator.
11344 (shift): New code attribute.
11345 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
11346 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
11347 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
11348 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
11349 bit-wise operations.
11350 (andsi3, *andsi3): Ditto.
11351 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
11352 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
11353 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
11354 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
11355 nds32_ior_operand, nds32_xor_operand): New predicates.
11357 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11359 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
11360 (addsi3, subsi3): ... this.
11362 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11364 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
11366 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11368 * config/nds32/nds32.md: Adjust indention.
11370 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
11372 * config/nds32/nds32.md (feature): New attribute.
11374 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11376 * config/nds32/nds32.md (subtype): New attribute.
11378 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
11381 * config/arm/arm-builtins.c (arm_expand_builtin): Change
11382 expansion to perform a bitwise AND of the argument followed by a
11383 boolean negation of the result.
11385 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
11387 PR rtl-optimization/84878
11388 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
11389 the basic block. Assert the use reference is not artificial and that
11390 it has an associated insn.
11392 2018-04-04 Michael Matz <matz@suse.de>
11394 * builtins.c (compute_objsize): Pass correct operand
11395 to array_at_struct_end_p.
11397 2018-04-04 Richard Biener <rguenther@suse.de>
11400 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
11401 from contexts for DINFO_LEVEL_TERSE and below.
11403 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
11405 * config/nds32/nds32-doubleword.md (move_<mode>): Require
11406 resiter_operand condition.
11407 * config/nds32/nds32.md (*move<mode>): Ditto.
11409 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
11410 Monk Chiang <sh.chiang04@gmail.com>
11412 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
11414 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11416 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
11418 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11419 Kito Cheng <kito.cheng@gmail.com>
11421 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
11422 nds32_cond_code_str, output_cond_branch,
11423 output_cond_branch_compare_zero, nds32_expand_cbranch,
11424 nds32_expand_cstore, nds32_expand_movcc,
11425 nds32_output_cbranchsi4_equality_zero,
11426 nds32_output_cbranchsi4_equality_reg,
11427 nds32_output_cbranchsi4_equality_reg_or_const_int,
11428 nds32_output_cbranchsi4_greater_less_zero: New functions.
11429 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
11430 nds32_expand_cstore, nds32_expand_movcc,
11431 nds32_output_cbranchsi4_equality_zero,
11432 nds32_output_cbranchsi4_equality_reg,
11433 nds32_output_cbranchsi4_equality_reg_or_const_int,
11434 nds32_output_cbranchsi4_greater_less_zero): Declare.
11435 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
11436 nds32_rimm11s_operand): New predicates.
11437 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
11438 * config/nds32/nds32.md: Rewrite all the branch and conditional move
11441 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
11443 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
11444 * config/nds32/nds32.md: Ditto.
11445 * config/nds32/pipelines.md: Ditto.
11447 2018-04-04 Richard Biener <rguenther@suse.de>
11449 PR tree-optimization/85168
11450 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
11451 propagating abnormals.
11453 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11455 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
11457 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11458 Kito Cheng <kito.cheng@gmail.com>
11460 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
11461 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
11462 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
11463 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
11464 * config/nds32/nds32.md (sibcall_internal): New.
11465 (sibcall_register): Remove.
11466 (sibcall_immediate): Remove.
11467 (sibcall_value_internal): New.
11468 (sibcall_value_register): Remove.
11469 (sibcall_value_immediate): Remove.
11470 * config/nds32/predicates.md (nds32_general_register_operand): New.
11471 (nds32_call_address_operand): New.
11473 2018-04-03 Jakub Jelinek <jakub@redhat.com>
11475 PR rtl-optimization/85167
11476 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
11477 bb_defs if *split_p, instead preinitialize it to NULL.
11479 PR tree-optimization/85156
11480 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
11481 evaluating the argument multiple times.
11483 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
11485 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
11487 (_mm_cvtpd_ps): Likewise.
11488 (_mm_cvttpd_epi32): Likewise.
11489 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
11490 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
11491 vector, pixel, and bool following altivec.h include.
11493 2018-04-03 Martin Sebor <msebor@redhat.com>
11495 * doc/extend.texi (Common Function Attributes): Clarify.
11496 (const attribute): Likewise.
11497 (pure attribute): Likewise.
11499 2018-04-03 Jakub Jelinek <jakub@redhat.com>
11502 * config/i386/i386.c (ix86_expand_vector_set): Use
11503 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
11505 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
11507 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
11508 instructions when changing rounding bits to preserve precision bits
11509 in the x87 control word.
11511 2018-04-03 Martin Liska <mliska@suse.cz>
11513 PR tree-optimization/82491
11514 * rtl.h (strip_offset_and_add): Replace += suboffset with
11515 poly_uint64 () + suboffset.
11517 2018-03-29 Martin Liska <mliska@suse.cz>
11518 Martin Jambor <mjambor@suse.cz>
11521 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
11522 param_type is not an integral or pointer type.
11524 2018-04-03 Richard Biener <rguenther@suse.de>
11526 * sese.h (recompute_all_dominators): Remove.
11528 2018-04-02 Martin Sebor <msebor@redhat.com>
11530 * doc/invoke.texi (-Wrestrict): Fix typos.
11532 2018-04-02 Jim Wilson <jimw@sifive.com>
11534 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
11535 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
11536 (<optab>di3, <optab>si3_extend): Likewise.
11537 (<optab>si3_mask, <optab>si3_mask_1): New.
11538 (<optab>di3_mask, <optab>di3_mask_1): New.
11539 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
11540 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
11541 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
11543 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
11545 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
11548 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
11550 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
11551 (nds32_canonicalize_comparison): New function.
11553 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
11554 Kito Cheng <kito.cheng@gmail.com>
11555 Kuan-Lin Chen <kuanlinchentw@gmail.com>
11557 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
11558 * config/nds32/constants.md (unspec_volatile_element): Add
11559 UNSPEC_VOLATILE_RELAX_GROUP.
11560 * config/nds32/nds32-relax-opt.c: New file.
11561 * config/nds32/nds32-predicates.c
11562 (nds32_symbol_load_store_p): New function.
11563 * config/nds32/nds32-protos.h
11564 (nds32_symbol_load_store_p): Declare function.
11565 (make_pass_nds32_relax_opt): Declare new rtl pass function.
11566 * config/nds32/nds32.c
11567 (nds32_register_pass): New function to register pass.
11568 (nds32_register_passes): New function to register passes.
11569 * config/nds32/nds32.md (relax_group): New pattern.
11570 * config/nds32/nds32.opt (mrelax-hint): New option.
11571 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
11573 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
11575 * config/nds32/t-nds32: Modify files dependency.
11577 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
11579 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
11580 (PROFILE_HOOK): Define its implementation.
11582 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
11584 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
11585 type and 32-bit size.
11587 2018-04-01 Jakub Jelinek <jakub@redhat.com>
11589 PR middle-end/85090
11590 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
11591 (V_128_256): New mode iterator.
11592 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
11593 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
11594 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
11596 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
11597 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
11599 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
11602 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
11603 NaN inputs correctly.
11605 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
11608 * config/rs6000/vsx.md (??r): New mode attribute.
11609 (*vsx_mov<mode>_64bit): Use it.
11610 (*vsx_mov<mode>_32bit): Likewise.
11612 2018-03-30 Martin Sebor <msebor@redhat.com>
11614 PR tree-optimization/84818
11615 * builtins.c (check_access): Use warning_n.
11617 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11620 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
11622 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
11625 2018-03-30 Julia Koval <julia.koval@intel.com>
11628 * x86-tune.def (movx, partial_reg_dependency): Enable for
11631 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
11633 PR inline-asm/84985
11634 * lra-constraints.c (process_alt_operands): Move setting
11635 this_alternative_matches below.
11637 2018-03-29 Martin Liska <mliska@suse.cz>
11640 * doc/invoke.texi: Document how LTO works with debug info.
11641 Describe auto-load support of binutils. Mention 'x86-64'
11642 as valid option value of -march option.
11644 2018-03-29 Jakub Jelinek <jakub@redhat.com>
11646 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
11649 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
11650 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
11651 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
11654 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
11657 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
11658 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
11659 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
11660 for RS6000_BTM_POWERPC64.
11661 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
11662 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
11663 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
11666 (DIVDEU): Likewise.
11668 2018-03-28 Carl Love <cel@us.ibm.com>
11671 2017-09-27 Carl Love <cel@us.ibm.com>
11673 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
11674 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
11675 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
11678 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11680 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
11681 instead of __vector bool.
11682 (_mm_max_pu8): Likewise.
11683 (_mm_min_pi16): Likewise.
11685 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
11688 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
11689 (DIVWEUO): Likewise.
11690 (DIVDEO): Likewise.
11691 (DIVDEUO): Likewise.
11692 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
11693 DIVWEUO and DIVDEUO.
11694 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
11695 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
11696 (div_extend): Likewise.
11697 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
11699 (__builtin_divweuo): Likewise.
11700 (__builtin_divdeo): Likewise.
11701 (__builtin_divdeuo): Likewise.
11703 2018-03-28 Jakub Jelinek <jakub@redhat.com>
11706 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
11707 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
11709 PR tree-optimization/82004
11710 * gimple-match-head.c (optimize_pow_to_exp): New function.
11711 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
11712 Don't fold to exp if optimize_pow_to_exp is false.
11714 2018-03-28 Martin Liska <mliska@suse.cz>
11717 * calls.c (initialize_argument_information): Fix trailing space.
11718 * common.opt: Fix typo and provide better explanation for
11719 -fsanitize-coverage option.
11720 * config/i386/i386.opt: Fix typo.
11722 2018-03-28 Jakub Jelinek <jakub@redhat.com>
11723 Martin Liska <mliska@suse.cz>
11726 * gimplify.c (asan_poison_variable): Don't do the check for
11727 gimplify_omp_ctxp here.
11728 (gimplify_decl_expr): Do it here.
11729 (gimplify_target_expr): Likewise.
11731 2018-03-28 Martin Liska <mliska@suse.cz>
11734 * config/i386/i386.c (ix86_function_arg_advance): Do not call
11735 chkp_type_bounds_count if MPX is not enabled.
11737 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
11739 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
11741 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
11744 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
11745 function to create the function decl for complex long double
11746 multiply and divide for -mabi=ieeelongdouble.
11747 (init_float128_ieee): Call it.
11749 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
11752 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
11753 -fcf-protection=branch -mibt.
11754 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
11756 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11759 * config/arm/arm.c (arm_valid_symbolic_address): Handle
11760 arm_word_relocations.
11762 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
11765 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
11766 extern array declarations.
11768 2018-03-27 Richard Biener <rguenther@suse.de>
11770 PR middle-end/84067
11771 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
11772 explicit single_use checks.
11774 2018-03-27 Richard Biener <rguenther@suse.de>
11776 PR tree-optimization/85082
11777 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
11780 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11782 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
11783 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
11784 Turn on fasynchronous-unwind-tables and funwind-tables.
11786 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
11789 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
11790 (*bmi_blsr_<mode>_ccz): Ditto.
11792 2018-03-26 Tom de Vries <tom@codesourcery.com>
11794 PR tree-optimization/85063
11795 * omp-general.c (offloading_function_p): New function. Factor out
11797 * omp-offload.c (pass_omp_target_link::gate): ... here.
11798 * omp-general.h (offloading_function_p): Declare.
11799 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
11800 with attribute omp declare target for offloading functions.
11802 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
11804 PR tree-optimization/84005
11805 * tree-data-ref.h (get_base_for_alignment): Declare.
11806 * tree-data-ref.c (get_base_for_alignment_1): New function.
11807 (get_base_for_alignment): Likewise.
11808 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
11809 get_base_for_alignment to find a suitable base object, instead
11810 of always using drb->base_address.
11812 2018-03-23 Jakub Jelinek <jakub@redhat.com>
11814 PR inline-asm/85022
11815 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
11816 known size by default.
11818 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
11820 PR inline-asm/85030
11821 * lra-constraints.c (process_alt_operands): Don't match BLKmode
11822 and non BLKmode operands.
11824 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11827 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
11828 Clean up attributes.
11830 2018-03-23 Richard Biener <rguenther@suse.de>
11833 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
11834 we are going to emit early debug for LTO.
11836 2018-03-23 Jakub Jelinek <jakub@redhat.com>
11838 PR inline-asm/85034
11839 * function.c (match_asm_constraints_1): Don't optimize if input
11840 doesn't satisfy general_operand predicate for output's mode.
11842 PR inline-asm/85022
11843 * alias.c (write_dependence_p): Don't require for x_canonicalized
11844 non-VOIDmode if x has VOIDmode.
11847 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
11848 just don't try to optimize it rather than assert it never happens.
11850 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
11852 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
11853 macro expansions for definition of ST_INTERNAL_<mode> and
11854 LD_INTERNAL_<mode> builtins.
11855 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
11857 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
11859 (altivec_expand_st_builtin): Likewise.
11860 (altivec_expand_builtin): Remove calls to deleted functions.
11861 (rs6000_address_for_altivec): Delete this function.
11862 * config/rs6000/vector.md: Remove expands for
11863 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
11865 2018-03-22 Sudakshina Das <sudi.das@arm.com>
11868 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
11869 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
11870 re-computing once computed.
11871 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
11872 (arm_init_machine_status): Initialize
11873 machine->static_chain_stack_bytes.
11875 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
11878 * doc/extend.texi: Add four new prototypes for vec_ld.
11879 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
11880 definitions for more logical presentation.
11881 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
11882 entries for V1TI variants of __builtin_altivec_ld builtin.
11883 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
11884 handling of V1TI variant of LVX icode pattern.
11885 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
11886 (rs6000_gimple_fold_builtin): Likewise.
11887 (altivec_init_builtins): Add code to define
11888 __builtin_altivec_lvx_v1ti function.
11890 2018-03-22 Jakub Jelinek <jakub@redhat.com>
11892 PR inline-asm/84941
11893 * function.c (match_asm_constraints_1): Don't do the optimization
11894 if input isn't a REG, SUBREG, MEM or constant.
11896 2018-03-22 Tom de Vries <tom@codesourcery.com>
11898 PR tree-optimization/84956
11899 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
11900 bb_has_abnormal_pred.
11902 2018-03-22 Jakub Jelinek <jakub@redhat.com>
11905 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
11906 DECL_INITIAL (decl) to decl at the end.
11907 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
11908 adjust the comment.
11910 2018-03-21 Joseph Myers <joseph@codesourcery.com>
11912 * doc/extend.texi (__builtin_tgmath): Document when complex
11913 integer types are treated as _Complex _Float64.
11915 2018-03-21 Tom de Vries <tom@codesourcery.com>
11917 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
11919 2018-03-21 Jakub Jelinek <jakub@redhat.com>
11921 PR tree-optimization/84960
11922 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
11923 if it is ENTRY block, move them into single succ of ENTRY in that case.
11925 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
11927 PR tree-optimization/84811
11928 * poly-int.h (poly_span_traits): Remove the T3 parameter and
11929 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
11930 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
11931 (known_subrange_p): Update accordingly. Cast each value involved
11932 in the size comparison, rather than casting the result of the
11935 2018-03-21 Jakub Jelinek <jakub@redhat.com>
11937 PR tree-optimization/84982
11938 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
11939 by flipping the least significant bit rather than all bits from
11940 bitpos to bitpos + bitsize - 1.
11942 2018-03-21 Nathan Sidwell <nathan@acm.org>
11944 * doc/extend.texi (Deprecated Features): Remove mention of
11945 long-deleted deprecations.
11947 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11950 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
11951 * configure: Regenerate.
11953 2018-03-21 Tom de Vries <tom@codesourcery.com>
11955 PR tree-optimization/83126
11956 * tree-parloops.c (num_phis): New function.
11957 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
11959 2018-03-21 Nathan Sidwell <nathan@acm.org>
11961 * doc/extend.texi (Deprecated Features): Update deprecated flags,
11962 mention anon-struct/union members and trailing attributes.
11964 2018-03-21 Bin Cheng <bin.cheng@arm.com>
11966 PR tree-optimization/84969
11967 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
11968 builtin memset partitions if they set different rhs values.
11970 2018-03-21 Jakub Jelinek <jakub@redhat.com>
11972 PR rtl-optimization/84989
11973 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
11974 VEC_DUPLICATE with scalar result mode.
11976 2018-03-21 Martin Liska <mliska@suse.cz>
11979 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
11980 not intended return statement.
11982 2018-03-21 Martin Liska <mliska@suse.cz>
11985 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
11986 (chkp_find_bound_slots_1): Limit number of iterations.
11988 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
11991 * Minor grammar fixes for x86 options.
11993 2018-03-20 Jakub Jelinek <jakub@redhat.com>
11996 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
11997 holding REG_CFA_RESTORE notes, instead turn them into a USE.
11999 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
12002 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
12003 (altivec_lvx_<mode>_1op): Likewise.
12004 (altivec_stvx_<mode>_2op): Likewise.
12005 (altivec_stvx_<mode>_1op): Likewise.
12006 (altivec_lvx_<VM2:mode>): New define_expand.
12007 (altivec_stvx_<VM2:mode>): Likewise.
12008 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
12009 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
12010 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
12011 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
12012 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
12013 (rs6000_gen_lvx): Likewise.
12014 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
12015 (altivec_expand_stv_builtin): Likewise.
12016 (altivec_expand_builtin): Likewise.
12017 * config/rs6000/vector.md: Likewise.
12019 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12022 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
12025 2018-03-20 Richard Biener <rguenther@suse.de>
12028 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
12029 sign-conversions as zero, fall back to standard scalar_stmt
12032 2018-03-20 Martin Liska <mliska@suse.cz>
12035 * predict.c (rebuild_frequencies): Handle case when we have
12036 PROFILE_ABSENT, but flag_guess_branch_prob is false.
12038 2018-03-20 Jakub Jelinek <jakub@redhat.com>
12041 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
12042 flag_section_anchors.
12043 * varasm.c (use_blocks_for_decl_p): Remove hack for
12044 dw2_force_const_mem.
12047 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
12049 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
12050 be created, use lowpart_subreg of operands[0] rather than operands[0]
12052 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
12053 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
12054 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
12055 and n constraint instead of aarch64_shift_imm_di and Usd.
12056 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
12057 (*aarch64_<optab>_reg_minus<mode>3): ... this.
12059 2018-03-20 Sudakshina Das <sudi.das@arm.com>
12062 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
12063 to favor GPR over NEON registers.
12064 (<shift>di3_neon): Likewise.
12066 2018-03-20 Tom de Vries <tom@codesourcery.com>
12069 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
12070 (nvptx_process_pars): Emit bar.sync asap and alap.
12072 2018-03-20 Tom de Vries <tom@codesourcery.com>
12075 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
12076 seen_label if seen_label is already set.
12078 2018-03-20 Jakub Jelinek <jakub@redhat.com>
12081 * config/i386/i386.c (fold_builtin_cpu): For features above 31
12082 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
12083 Use 1U instead of 1. Formatting fixes.
12086 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
12087 instead of TREE_TYPE (s1) for the return value.
12089 2018-03-19 Jakub Jelinek <jakub@redhat.com>
12091 PR tree-optimization/84946
12092 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
12093 bitsize + bitsize in poly_uint64 rather than poly_int64.
12096 * dwarf2asm.c: Include fold-const.c.
12097 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
12098 of decl rather than decl itself.
12100 PR rtl-optimization/84643
12101 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
12103 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
12106 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
12107 calling assemble_variable.
12109 2018-03-19 Sudakshina Das <sudi.das@arm.com>
12112 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
12113 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
12115 2018-03-19 Jim Wilson <jimw@sifive.com>
12118 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
12119 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
12120 (riscv_first_stack_step): Likewise.
12121 (riscv_option_override): Use STACK_BOUNDARY instead of
12122 MIN_STACK_BOUNDARY.
12123 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
12124 MIN_STACK_BOUNDARY.
12125 (BIGGEST_ALIGNMENT): Set to 128.
12126 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
12127 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
12130 2018-03-19 Richard Biener <rguenther@suse.de>
12132 PR tree-optimization/84933
12133 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
12134 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
12136 2018-03-19 Richard Biener <rguenther@suse.de>
12138 PR tree-optimization/84859
12139 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
12140 (cond_if_else_store_replacement): Perform sinking operation on
12141 single-store BBs regardless of MAX_STORES_TO_SINK setting.
12142 Generalize what a BB with a single eligible store is.
12144 2018-03-19 Richard Biener <rguenther@suse.de>
12146 PR tree-optimization/84929
12147 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
12148 chrec_is_positive against non-chrec arg.
12150 2018-03-19 Tamar Christina <tamar.christina@arm.com>
12153 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
12155 2018-03-18 Martin Liska <mliska@suse.cz>
12157 PR rtl-optimization/84635
12158 * regrename.c (build_def_use): Use matches_mode only when
12161 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
12163 PR tree-optimization/84913
12164 * tree-vect-loop.c (vectorizable_reduction): Don't try to
12165 vectorize chains of COND_EXPRs.
12167 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
12169 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
12171 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
12173 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
12175 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
12177 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
12179 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
12180 Kito Cheng <kito.cheng@gmail.com>
12182 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
12183 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
12184 (nds32_adjust_reg_alloc_order): New function.
12185 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
12187 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
12189 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
12190 nds32_print_operand, nds32_print_operand_address): Use
12191 HOST_WIDE_INT_PRINT_DEC instead.
12193 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
12195 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
12197 2018-03-17 Jakub Jelinek <jakub@redhat.com>
12200 * config/i386/i386.c (initial_ix86_tune_features,
12201 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
12202 unsigned long long.
12203 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
12204 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
12205 rather than 1u << ix86_tune. Formatting fix.
12206 (ix86_option_override_internal): Change ix86_arch_mask from
12207 unsigned int to unsigned HOST_WIDE_INT, initialize to
12208 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
12209 (ix86_function_specific_restore): Likewise.
12211 2018-03-16 Jakub Jelinek <jakub@redhat.com>
12214 * postreload.c (reload_combine_recognize_pattern): Perform
12215 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
12216 truncate_int_for_mode the result for the destination's mode.
12219 * hsa-gen.c (mem_type_for_type): Fix comment typo.
12220 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
12222 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
12225 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
12228 * lra-assigns.c (lra_split_hard_reg_for): Don't use
12229 regno_allocno_class_array and sorted_pseudos.
12230 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
12231 insns where regno is used.
12233 2018-03-16 Martin Liska <mliska@suse.cz>
12236 * multiple_target.c (create_dispatcher_calls): Redirect
12237 reference in the symbol table.
12239 2018-03-16 Martin Liska <mliska@suse.cz>
12242 * multiple_target.c (create_dispatcher_calls): Redirect also
12245 2018-03-16 Jakub Jelinek <jakub@redhat.com>
12249 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
12250 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
12251 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
12253 2018-03-16 Julia Koval <julia.koval@intel.com>
12255 * doc/invoke.texi (Skylake Server): Add CLWB.
12256 Cannonlake): Remove CLWB.
12258 2018-03-16 Jakub Jelinek <jakub@redhat.com>
12260 PR tree-optimization/84841
12261 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
12263 (FLOAT_ONE_CONST_TYPE): Define.
12264 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
12265 (sort_by_operand_rank): Put entries with higher constant_type last
12266 rather than first to match comments.
12268 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
12270 * config/nios2/nios2.md (movsi_internal): Fix thinko in
12273 2018-03-15 Jakub Jelinek <jakub@redhat.com>
12276 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
12277 check and use address of target always.
12279 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
12282 * config/i386/i386.c (indirect_thunk_needed): Update comments.
12283 (indirect_thunk_bnd_needed): Likewise.
12284 (indirect_thunks_used): Likewise.
12285 (indirect_thunks_bnd_used): Likewise.
12286 (indirect_return_needed): New.
12287 (indirect_return_bnd_needed): Likewise.
12288 (output_indirect_thunk_function): Add a bool argument for
12290 (output_indirect_thunk_function): Don't generate alias for
12291 function return thunk.
12292 (ix86_code_end): Call output_indirect_thunk_function to generate
12293 function return thunks.
12294 (ix86_output_function_return): Set indirect_return_bnd_needed
12295 and indirect_return_needed instead of indirect_thunk_bnd_needed
12296 and indirect_thunk_needed.
12298 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
12300 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
12301 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
12302 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
12304 2018-03-15 David Malcolm <dmalcolm@redhat.com>
12305 Paul Hua <paul.hua.gm@gmail.com>
12308 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
12310 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
12312 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
12313 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
12314 resp. SFmode cases.
12316 2018-03-15 Tamar Christina <tamar.christina@arm.com>
12319 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
12320 instead of GET_MODE_SIZE when comparing Units.
12322 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
12325 * varasm.c (hash_section): Return an unchangeble hash value
12326 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
12327 Return !aarch64_can_use_per_function_literal_pools_p ().
12329 2018-03-15 Jakub Jelinek <jakub@redhat.com>
12332 * optabs.c (emit_conditional_move): Pass address of cmode's copy
12333 rather than address of cmode as last argument to prepare_cmp_insn.
12335 2018-03-15 Julia Koval <julia.koval@intel.com>
12337 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
12338 F_AVX512VNNI, F_AVX512BITALG): New.
12340 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
12343 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
12344 insn for floating-point loads and stores.
12346 2018-03-14 Carl Love <cel@us.ibm.com>
12348 * config/rs6000/rs6000-c.c: Add macro definitions for
12349 ALTIVEC_BUILTIN_VEC_PERMXOR.
12350 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
12351 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
12352 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
12353 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
12355 * config/doc/extend.texi: Add prototypes for vec_permxor.
12357 2018-03-14 David Malcolm <dmalcolm@redhat.com>
12360 * diagnostic-show-locus.c (class layout_point): Convert m_line
12361 from int to linenum_type.
12362 (line_span::comparator): Use linenum "compare" function when
12363 comparing line numbers.
12364 (test_line_span): New function.
12365 (layout_range::contains_point): Convert param "row" from int to
12367 (layout_range::intersects_line_p): Likewise.
12368 (layout::will_show_line_p): Likewise.
12369 (layout::print_source_line): Likewise.
12370 (layout::should_print_annotation_line_p): Likewise.
12371 (layout::print_annotation_line): Likewise.
12372 (layout::print_leading_fixits): Likewise.
12373 (layout::annotation_line_showed_range_p): Likewise.
12374 (struct line_corrections): Likewise for field m_row.
12375 (line_corrections::line_corrections): Likewise for param "row".
12376 (layout::print_trailing_fixits): Likewise.
12377 (layout::get_state_at_point): Likewise.
12378 (layout::get_x_bound_for_row): Likewise.
12379 (layout::print_line): Likewise.
12380 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
12381 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
12382 * input.c (selftest::test_linenum_comparisons): New function.
12383 (selftest::input_c_tests): Call it.
12384 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
12385 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
12386 * selftest.h (ASSERT_GT): New macro.
12387 (ASSERT_GT_AT): New macro.
12388 (ASSERT_LT): New macro.
12389 (ASSERT_LT_AT): New macro.
12391 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
12393 PR rtl-optimization/84780
12394 * combine.c (distribute_links): Don't make a link based on pc_rtx.
12396 2018-03-14 Martin Liska <mliska@suse.cz>
12398 * tree.c (record_node_allocation_statistics): Use
12399 get_stats_node_kind.
12400 (get_stats_node_kind): New function extracted from
12401 record_node_allocation_statistics.
12402 (free_node): Use get_stats_node_kind.
12404 2018-03-14 Richard Biener <rguenther@suse.de>
12406 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
12407 that the value-set of ANTIC_IN doesn't grow.
12410 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
12412 (BB_VISITED_WITH_VISITED_SUCCS): New define.
12413 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
12415 2018-03-14 Julia Koval <julia.koval@intel.com>
12417 * config.gcc (icelake-client, icelake-server): New.
12419 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
12420 (initial_ix86_arch_features): Ditto.
12421 (PTA_SKYLAKE): Add SGX.
12422 (PTA_ICELAKE): Remove.
12423 (PTA_ICELAKE_CLIENT): New.
12424 (PTA_ICELAKE_SERVER): New.
12425 (ix86_option_override_internal): Split up icelake on icelake client and
12427 (get_builtin_code_for_version): Ditto.
12428 (fold_builtin_cpu): Ditto.
12429 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
12430 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
12431 * config/i386/i386.h (processor_type): Ditto.
12432 * doc/invoke.texi: Ditto.
12434 2018-03-14 Jakub Jelinek <jakub@redhat.com>
12437 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
12438 INTEGER_CST offset, add it together with bitpos / 8 and
12439 sign extend based on POINTER_SIZE.
12443 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
12446 * config/i386/constraints.md (Yc): New register constraint.
12447 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
12448 Use Yc constraint for alternative 2 of operand 0. Remove
12449 preferred_for_speed attribute.
12451 2018-03-14 Richard Biener <rguenther@suse.de>
12453 PR tree-optimization/84830
12454 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
12455 with the old one to avoid oscillations.
12457 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
12460 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
12462 (assign_by_spills): Return a flag of reload assignment failure.
12463 Do not process the reload assignment failures. Do not spill other
12464 reload pseudos if they has the same reg class. Update n if
12466 (lra_assign): Add a return arg. Set up from the result of
12467 assign_by_spills call.
12468 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
12469 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
12470 usage_insns if it is not NULL.
12471 (spill_hard_reg_in_range): New function.
12472 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
12473 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
12474 function prototypes.
12475 (lra_assign): Change prototype.
12476 * lra.c (lra): Add code to deal with fails by splitting hard reg
12479 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
12481 * config/riscv/riscv.opt (mrelax): New option.
12482 * config/riscv/riscv.c (riscv_file_start): Emit ".option
12483 "norelax" when riscv_mrelax is disabled.
12484 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
12486 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12489 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
12490 reassociation for int modes.
12492 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
12494 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
12495 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
12497 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
12498 * config/aarch64/aarch64-sve.md
12499 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
12500 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
12501 (*extend<mode><Vwide>2): Rename to...
12502 (aarch64_sve_extend<mode><Vwide>2): ...this.
12503 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
12504 renaming the old pattern to...
12505 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
12507 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
12508 define_expand, renaming the old pattern to...
12509 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
12510 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
12511 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
12512 account when deciding which SVE instruction the optab should use.
12513 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
12515 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
12517 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
12518 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
12519 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
12520 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
12521 (tlsdesc_small_<mode>): Turn a define_expand and use
12522 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
12523 (tlsdesc_small_advsimd_<mode>): ...this.
12524 (tlsdesc_small_sve_<mode>): New pattern.
12526 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
12528 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
12529 (UNSPEC_UMUL_HIGHPART): New constants.
12530 (MUL_HIGHPART): New int iteraor.
12531 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
12532 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
12534 (*<su>mul<mode>3_highpart): New define_insn.
12536 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
12539 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
12542 2018-03-13 Martin Liska <mliska@suse.cz>
12545 * (sem_item_optimizer::sem_item_optimizer): Initialize new
12547 (sem_item_optimizer::~sem_item_optimizer): Release it.
12548 (sem_item_optimizer::merge_classes): Register variable aliases.
12549 (sem_item_optimizer::fixup_pt_set): New function.
12550 (sem_item_optimizer::fixup_points_to_sets): Likewise.
12551 * ipa-icf.h: Declare new variables and functions.
12553 2018-03-13 Jakub Jelinek <jakub@redhat.com>
12555 PR middle-end/84834
12556 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
12557 integer_pow2p@2 and test integer_pow2p in condition.
12558 (A < 0 ? C : 0): Similarly for @1.
12560 PR middle-end/84831
12561 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
12562 characters starting at p contain '\0' character, don't look beyond
12566 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
12567 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
12570 * reg-stack.c (change_stack): Change update_end var from int to
12571 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
12572 also call set_block_for_insn on the newly added insns and rescan.
12575 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
12576 on the last operand.
12579 * tree.c (stabilize_reference_1): Return save_expr (e) for
12580 STATEMENT_LIST even if it doesn't have side-effects.
12582 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
12584 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
12586 2018-03-12 Renlin Li <renlin.li@arm.com>
12588 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
12589 aarch64_output_scalar_simd_mov_immediate.
12591 2018-03-12 Martin Sebor <msebor@redhat.com>
12593 PR tree-optimization/83456
12594 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
12595 for perfectly overlapping calls to memcpy.
12596 (gimple_fold_builtin_memory_chk): Same.
12597 (gimple_fold_builtin_strcpy): Handle no-warning.
12598 (gimple_fold_builtin_stxcpy_chk): Same.
12599 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
12601 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
12603 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
12604 parameter. Use it for SFmode.
12605 (rs6000_function_arg_advance_1): Adjust.
12606 (rs6000_function_arg): Adjust.
12607 (rs6000_gimplify_va_arg): Pass false for that new parameter.
12609 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
12611 PR rtl-optimization/84169
12612 PR rtl-optimization/84780
12613 * combine.c (can_combine_p): Check for a 2-insn combination whether
12614 the destination register is used between the two insns, too.
12616 2018-03-12 Richard Biener <rguenther@suse.de>
12618 PR tree-optimization/84803
12619 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
12620 for refs DR analysis didn't process.
12622 2018-03-12 Richard Biener <rguenther@suse.de>
12624 PR tree-optimization/84777
12625 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
12626 force-vectorize loops ignore whether we are optimizing for size.
12628 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
12630 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
12631 (TARGET_MD_ASM_ADJUST): Define.
12633 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
12634 Kito Cheng <kito.cheng@gmail.com>
12635 Chung-Ju Wu <jasonwucj@gmail.com>
12637 * config/nds32/nds32.c (nds32_compute_stack_frame,
12638 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
12639 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
12640 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
12641 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
12642 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
12643 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
12644 * config/nds32/nds32.md (prologue, epilogue): Use macro
12645 NDS32_V3PUSH_AVAILABLE_P to do checking.
12647 2018-03-11 Jakub Jelinek <jakub@redhat.com>
12650 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
12651 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
12652 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
12653 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
12654 addition of most attributes on !orig_type_die or the attribute not
12655 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
12657 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
12658 Chung-Ju Wu <jasonwucj@gmail.com>
12660 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
12661 __NDS32_VH__ macro.
12662 * config/nds32/nds32.opt (mvh): New option.
12664 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
12665 Chung-Ju Wu <jasonwucj@gmail.com>
12667 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
12669 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
12670 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
12673 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
12674 Chung-Ju Wu <jasonwucj@gmail.com>
12676 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
12678 * config/nds32/nds32-multiple.md (strlensi): New pattern.
12679 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
12681 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
12682 Kito Cheng <kito.cheng@gmail.com>
12683 Chung-Ju Wu <jasonwucj@gmail.com>
12685 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
12686 UNSPEC_FFMISM and UNSPEC_FLMISM.
12687 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
12688 for ffb, ffmism and flmism.
12689 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
12690 (unspec_ffmism): Ditto.
12691 (unspec_flmism): Ditto.
12692 (nds32_expand_builtin_impl): Check if string extension is available.
12693 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
12694 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
12696 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
12699 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
12702 * lra-assigns.c (assign_by_spills): Return a flag of reload
12703 assignment failure. Do not process the reload assignment
12704 failures. Do not spill other reload pseudos if they has the same
12706 (lra_assign): Add a return arg. Set up from the result of
12707 assign_by_spills call.
12708 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
12709 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
12710 usage_insns if it is not NULL.
12711 (spill_hard_reg_in_range): New function.
12712 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
12713 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
12714 function prototypes.
12715 (lra_assign): Change prototype.
12716 * lra.c (lra): Add code to deal with fails by splitting hard reg
12719 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
12722 * config/i386/i386.opt: Replace Enforcment with Enforcement.
12724 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
12727 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
12728 (dw_val_node): Add val_symbolic_view.
12729 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
12730 (symview_upper_bound): New.
12731 (new_line_info_table): Initialize symviews_since_reset.
12732 (dwarf2out_source_line): Count symviews_since_reset and set
12733 symview_upper_bound.
12734 (dw_val_equal_p): Handle symview.
12735 (add_AT_symview): New.
12736 (print_dw_val): Handle symview.
12737 (attr_checksum, attr_checksum_ordered): Likewise.
12738 (same_dw_val_p, size_of_die): Likewise.
12739 (value_format, output_die): Likewise.
12740 (add_high_low_attributes): Use add_AT_symview for entry_view.
12741 (dwarf2out_finish): Reset symview_upper_bound, clear
12744 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
12747 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
12748 Add strict argument and use it.
12749 (rs6000_split_multireg_move): Update for new strict argument.
12750 (mem_operand_gpr): Disallow all non-offsettable addresses.
12751 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
12753 2018-03-09 Jakub Jelinek <jakub@redhat.com>
12756 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
12757 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
12758 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
12761 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
12762 decl, use remap_type if we want to use the type.
12764 2018-03-09 Martin Sebor <msebor@redhat.com>
12766 PR tree-optimization/84526
12767 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
12769 (builtin_access::generic_overlap): Be prepared to handle non-array
12772 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
12774 PR rtl-optimization/84682
12775 * lra-constraints.c (process_address_1): Check is_address flag
12776 for address constraints.
12777 (process_alt_operands): Likewise.
12778 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
12779 preprocess_constraints.
12780 * recog.h (preprocess_constraints): Add oploc parameter.
12782 * recog.c (preprocess_constraints): Test address_operand for
12783 CT_ADDRESS constraints.
12785 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
12788 * lra-assigns.c (assign_by_spills): Return a flag of reload
12789 assignment failure. Do not process the reload assignment
12790 failures. Do not spill other reload pseudos if they has the same
12792 (lra_assign): Add a return arg. Set up from the result of
12793 assign_by_spills call.
12794 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
12795 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
12796 usage_insns if it is not NULL.
12797 (spill_hard_reg_in_range): New function.
12798 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
12799 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
12800 function prototypes.
12801 (lra_assign): Change prototype.
12802 * lra.c (lra): Add code to deal with fails by splitting hard reg
12805 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12808 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
12809 Accept complain bool parameter. Only emit errors if it is true.
12810 (arm_parse_cpu_option_name): Likewise.
12811 (arm_target_thumb_only): Adjust callers of the above.
12812 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
12813 prototype to take a default true bool parameter.
12814 (arm_parse_arch_option_name): Likewise.
12816 2018-03-09 David Malcolm <dmalcolm@redhat.com>
12817 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
12821 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
12822 * configure: Regenerate.
12823 * configure.ac ("linker --version-script option"): New.
12824 ("linker soname option"): New.
12826 2018-03-09 Richard Biener <rguenther@suse.de>
12828 PR tree-optimization/84775
12829 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
12830 immediate uses of predicate stmts and mark them modified.
12833 PR tree-optimization/84178
12834 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
12836 (version_loop_for_if_conversion): Delay update_ssa call.
12837 (tree_if_conversion): Delay update_ssa until after predicate
12840 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
12843 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
12844 when the function accesses prior frames.
12846 2018-03-08 Jakub Jelinek <jakub@redhat.com>
12849 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
12850 gen_llsym, otherwise call maybe_gen_llsym.
12852 PR inline-asm/84742
12853 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
12854 has ',' character inside of it.
12856 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12859 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
12860 as clobbering CC_REGNUM.
12862 2018-03-08 Richard Biener <rguenther@suse.de>
12864 PR middle-end/84552
12865 * tree-scalar-evolution.c: Include tree-into-ssa.h.
12866 (follow_copies_to_constant): Do not follow SSA names registered
12869 2018-03-08 Richard Biener <rguenther@suse.de>
12871 PR tree-optimization/84178
12872 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
12874 (version_loop_for_if_conversion): Delay update_ssa call.
12875 (tree_if_conversion): Delay update_ssa until after predicate
12878 2018-03-08 David Malcolm <dmalcolm@redhat.com>
12880 PR tree-optimization/84178
12881 * tree-if-conv.c (release_bb_predicate): Remove the
12882 the assertion that the stmts have NULL use_ops.
12883 Discard the statements, asserting that they haven't
12884 yet been added to a BB.
12886 2018-03-08 Richard Biener <rguenther@suse.de>
12888 PR tree-optimization/84746
12889 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
12890 (phi_translate): Pass in destination ANTIC_OUT set.
12891 (phi_translate_1): Likewise. For a simplified result lookup
12892 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
12893 (phi_translate_set): Adjust.
12894 (do_pre_regular_insertion): Likewise.
12895 (do_pre_partial_partial_insertion): Likewise.
12897 2018-03-08 Martin Liska <mliska@suse.cz>
12899 PR gcov-profile/84735
12900 * doc/gcov.texi: Document usage of profile files.
12901 * gcov-io.h: Document changes in the format.
12903 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
12907 * dwarf2out.c (struct dw_line_info_table): Update comments for
12909 (FORCE_RESET_NEXT_VIEW): New.
12910 (FORCE_RESETTING_VIEW_P): New.
12911 (RESETTING_VIEW_P): Check for -1 too.
12912 (ZERO_VIEW_P): Likewise.
12913 (new_line_info_table): Force-reset next view.
12914 (dwarf2out_begin_function): Likewise.
12915 (dwarf2out_source_line): Simplify zero_view_p initialization.
12916 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
12917 view directly. Omit view when omitting .loc at line 0.
12919 2018-03-08 Jakub Jelinek <jakub@redhat.com>
12921 PR tree-optimization/84740
12922 * tree-switch-conversion.c (process_switch): Call build_constructors
12923 only if info.phi_count is non-zero.
12925 PR tree-optimization/84739
12926 * tree-tailcall.c (find_tail_calls): Check call arguments against
12927 DECL_ARGUMENTS (current_function_decl) rather than
12928 DECL_ARGUMENTS (func) when checking for tail recursion.
12930 2018-03-07 Jakub Jelinek <jakub@redhat.com>
12932 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
12933 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
12934 Volker Reichelt's entry and add entries for people that perform
12935 GCC fuzzy testing and report numerous bugs.
12937 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
12940 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
12941 readonly data in sdata, if that is disabled.
12942 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
12943 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
12944 -mreadonly-in-sdata option.
12946 2018-03-07 Martin Sebor <msebor@redhat.com>
12948 PR tree-optimization/84468
12949 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
12950 basic block when looking for nul assignment.
12952 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
12955 * except.h (output_function_exception_table): Adjust prototype.
12956 * except.c (output_function_exception_table): Remove FNNAME parameter
12957 and add SECTION parameter. Ouput one part of the table at a time.
12958 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
12959 the first part of the exception table and emit unwind directives.
12960 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
12961 (i386_pe_seh_cold_init): Likewise.
12962 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
12963 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
12964 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
12965 (ix86_output_call_insn): Emit a nop in one more case for SEH.
12966 * config/i386/winnt.c: Include except.h.
12967 (struct seh_frame_state): Add reg_offset, after_prologue and
12968 in_cold_section fields.
12969 (i386_pe_seh_end_prologue): Set seh->after_prologue.
12970 (i386_pe_seh_cold_init): New function.
12971 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
12972 to seh->in_cold_section.
12973 (seh_emit_push): Record the offset of the push.
12974 (seh_emit_save): Record the offet of the save.
12975 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
12976 Test seh->after_prologue to disregard the epilogue.
12977 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
12978 (i386_pe_end_cold_function): New function.
12980 2018-03-07 Jakub Jelinek <jakub@redhat.com>
12983 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
12984 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
12987 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
12989 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
12990 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
12992 PR middle-end/84723
12993 * multiple_target.c: Include tree-inline.h and intl.h.
12994 (expand_target_clones): Diagnose and fail if node->definition and
12995 !tree_versionable_function_p (node->decl).
12997 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
12999 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
13001 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
13002 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
13003 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
13005 2018-03-06 Jakub Jelinek <jakub@redhat.com>
13008 * combine.c (try_combine): Use reg_or_subregno instead of handling
13009 just paradoxical SUBREGs and REGs.
13011 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
13013 * config/arc/arc.c (arc_finalize_pic): Remove function.
13014 (arc_must_save_register): We use single base PIC register, remove
13015 checks to save/restore the PIC register.
13016 (arc_expand_prologue): Likewise.
13017 * config/arc/arc-protos.h (arc_set_default_type_attributes):
13019 (arc_verify_short): Likewise.
13020 (arc_attr_type): Likewise.
13021 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
13022 (walk_stores): Likewise.
13023 (arc_address_cost): Make it static.
13024 (arc_verify_short): Likewise.
13025 (branch_dest): Likewise.
13026 (arc_attr_type): Likewise.
13027 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
13028 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
13029 (arc_final_prescan_insn): Remove inserting the nops due to
13030 hardware hazards. It is done in reorg step.
13031 (insn_length_variant_t): Remove.
13032 (insn_length_parameters_t): Likewise.
13033 (arc_insn_length_parameters): Likewise.
13034 (arc_get_insn_variants): Likewise.
13035 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
13037 2018-03-06 Jakub Jelinek <jakub@redhat.com>
13039 PR inline-asm/84683
13040 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
13043 PR tree-optimization/84687
13044 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
13046 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
13048 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13050 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
13051 Rename to ppc_speculation_barrier.
13052 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
13053 __builtin_ppc_speculation_barrier.
13055 2018-03-05 Jakub Jelinek <jakub@redhat.com>
13058 * combine.c (combine_simplify_rtx): Don't try to simplify if
13059 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
13062 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
13064 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
13065 to 32 bytes when compiling for POWER9.
13067 2018-03-05 Jakub Jelinek <jakub@redhat.com>
13070 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
13071 regparm >= 3 with no arg reg available also for calls with
13072 flag_force_indirect_call. Pass decl to ix86_function_regparm.
13075 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
13077 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
13079 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
13082 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
13084 2018-03-05 Richard Biener <rguenther@suse.de>
13086 PR tree-optimization/84486
13087 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
13088 When inserting a __builtin_assume_aligned call set the LHS
13089 SSA name alignment info accordingly.
13091 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
13093 PR tree-optimization/84114
13094 * config/aarch64/aarch64.c (aarch64_reassociation_width)
13095 Avoid reassociation of FLOAT_MODE addition.
13097 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
13099 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
13100 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
13101 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
13102 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
13103 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
13104 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
13105 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
13107 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
13108 __builtin_ia32_wbinvd): New builtins.
13109 (SPECIAL_ARGS2): New.
13110 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
13111 (SPECIAL_ARGS2): New.
13112 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
13113 (ix86_valid_target_attribute_inner_p): Ditto.
13114 (ix86_init_mmx_sse_builtins): Add special_args2.
13115 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
13116 TARGET_WBNOINVD_P): New.
13117 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
13118 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
13119 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
13120 * config/i386/immintrin.h (_wbinvd): New intrinsic.
13121 * config/i386/pconfigintrin.h: New file.
13122 * config/i386/wbnoinvdintrin.h: Ditto.
13123 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
13125 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
13127 2018-03-05 Richard Biener <rguenther@suse.de>
13129 PR tree-optimization/84670
13130 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
13132 (BB_VISITED_WITH_VISITED_SUCCS): New define.
13133 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
13134 (compute_antic_aux): Only assert the number of values in ANTIC_IN
13135 doesn't grow if all successors (recursively) were visited at least
13138 2018-03-05 Richard Biener <rguenther@suse.de>
13140 PR tree-optimization/84650
13141 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
13142 if executed in the loop pipeline.
13144 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
13146 * doc/configfiles.texi (Configuration Files): Move info about
13147 conditionalizing $target-protos.h to...
13148 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
13149 differs from $target-protos.h.
13151 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
13152 Chung-Ju Wu <jasonwucj@gmail.com>
13154 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
13155 * config/nds32/nds32-multiple.md (setmemsi): Define.
13156 * config/nds32/nds32-memory-manipulation.c
13157 (nds32_gen_dup_4_byte_to_word_value): New.
13158 (emit_setmem_word_loop): New.
13159 (emit_setmem_byte_loop): New.
13160 (nds32_expand_setmem_loop): New.
13161 (nds32_expand_setmem_loop_v3m): New.
13162 (nds32_expand_setmem_unroll): New.
13163 (nds32_expand_setmem): New.
13165 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
13166 Chung-Ju Wu <jasonwucj@gmail.com>
13168 * config/nds32/nds32-memory-manipulation.c
13169 (nds32_emit_load_store): New.
13170 (nds32_emit_post_inc_load_store): New.
13171 (nds32_emit_mem_move): New.
13172 (nds32_emit_mem_move_block): New.
13173 (nds32_expand_movmemsi_loop_unknown_size): New.
13174 (nds32_expand_movmemsi_loop_known_size): New.
13175 (nds32_expand_movmemsi_loop): New.
13176 (nds32_expand_movmemsi_unroll): New.
13177 (nds32_expand_movmemqi): Rename ...
13178 (nds32_expand_movmemsi): ... to this.
13179 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
13180 (movmemsi): ... to this.
13181 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
13182 (nds32_expand_movmemsi): ... to this.
13184 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
13185 Monk Chiang <sh.chiang04@gmail.com>
13186 Chung-Ju Wu <jasonwucj@gmail.com>
13188 * config/nds32/nds32-protos.h
13189 (nds32_expand_load_multiple): New arguments.
13190 (nds32_expand_store_multiple): Ditto.
13191 (nds32_valid_multiple_load_store): Rename ...
13192 (nds32_valid_multiple_load_store_p): ... to this.
13193 * config/nds32/nds32-memory-manipulation.c
13194 (nds32_expand_load_multiple): Refine implementation.
13195 (nds32_expand_store_multiple): Ditto.
13196 * config/nds32/nds32-multiple.md
13197 (load_multiple): Update nds32_expand_load_multiple interface.
13198 (store_multiple): Update nds32_expand_store_multiple interface.
13199 * config/nds32/nds32-predicates.c
13200 (nds32_valid_multiple_load_store): Rename ...
13201 (nds32_valid_multiple_load_store_p): ... to this and refine
13203 * config/nds32/predicates.md
13204 (nds32_load_multiple_and_update_address_operation): New predicate.
13205 (nds32_store_multiple_and_update_address_operation): New predicate.
13207 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
13208 Chung-Ju Wu <jasonwucj@gmail.com>
13210 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
13211 (combo): New attribute.
13212 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
13214 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
13216 * config/nds32/nds32.opt: Change -mcmodel= default value.
13218 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
13219 Monk Chiang <sh.chiang04@gmail.com>
13220 Chung-Ju Wu <jasonwucj@gmail.com>
13222 * config/nds32/constants.md (unspec_element): New enum.
13223 * config/nds32/constraints.md (Umw): New constraint.
13224 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
13225 * config/nds32/nds32-intrinsic.md: Likewise.
13226 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
13227 (nds32_valid_smw_lwm_base_p): New.
13228 (nds32_output_smw_single_word): New.
13229 (nds32_output_lmw_single_word): New.
13230 (nds32_expand_unaligned_load): New.
13231 (nds32_expand_unaligned_store): New.
13232 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
13233 (nds32_output_smw_single_word): Declare.
13234 (nds32_output_lmw_single_word): Declare.
13235 (nds32_expand_unaligned_load): Declare.
13236 (nds32_expand_unaligned_store): Declare.
13237 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
13238 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
13239 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
13240 NDS32_BUILTIN_UASTORE_DW.
13241 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
13244 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
13245 Kito Cheng <kito.cheng@gmail.com>
13246 Chung-Ju Wu <jasonwucj@gmail.com>
13248 * config/nds32/nds32-intrinsic.c
13249 (nds32_expand_builtin_null_ftype_reg): Delete.
13250 (nds32_expand_builtin_reg_ftype_imm): Ditto.
13251 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
13252 (nds32_read_argument): New.
13253 (nds32_legitimize_target): Ditto.
13254 (nds32_legitimize_argument): Ditto.
13255 (nds32_check_constant_argument): Ditto.
13256 (nds32_expand_unop_builtin): Ditto.
13257 (nds32_expand_unopimm_builtin): Ditto.
13258 (nds32_expand_binop_builtin): Ditto.
13259 (nds32_builtin_decl_impl): Ditto.
13260 (builtin_description): Ditto.
13261 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
13262 (nds32_init_builtins_impl): Ditto.
13263 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
13264 (nds32_builtin_decl): New.
13265 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
13266 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
13268 2018-03-02 Jeff Law <law@redhat.com>
13270 * reorg.c (stop_search_p): Handle DEBUG_INSN.
13271 (redundant_insn, fill_simple_delay_slots): Likewise.
13272 (fill_slots_from_thread): Likewise.
13273 * resource.c (mark_referenced_resources): Likewise.
13274 (mark_set_resources, find_dead_or_set_registers): Likewise.
13276 2018-03-02 Jakub Jelinek <jakub@redhat.com>
13278 * substring-locations.h (format_warning_va): Formatting fix for
13279 ATTRIBUTE_GCC_DIAG.
13280 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
13282 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
13283 * substring-locations.c: Include intl.h.
13284 (format_warning_va): Turned into small wrapper around
13285 format_warning_n_va, renamed to ...
13286 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
13287 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
13289 (format_warning_at_substring_n): New function.
13290 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
13291 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
13292 format_warning_at_substring with just a shorter name instead of
13293 const function pointer.
13294 (fmtwarn_n): New function.
13295 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
13296 appropriate, get rid of all the fmtstr temporaries, move conditionals
13297 with G_() wrapped string literals directly into fmtwarn arguments,
13298 cast dir.len to (int), formatting fixes.
13300 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
13302 * doc/invoke.texi: Remove "Cilk Plus" references.
13304 2018-03-02 Jakub Jelinek <jakub@redhat.com>
13305 Richard Biener <rguenther@suse.de>
13308 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
13309 for error or warning attributes if CALL_FROM_THUNK_P is set.
13312 2018-03-02 Jakub Jelinek <jakub@redhat.com>
13315 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
13316 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
13319 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
13320 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
13322 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
13323 instead of -1U in last predictors element's probability member.
13325 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
13328 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
13329 arguments if they are comparable.
13331 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
13333 PR tree-optimization/84634
13334 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
13335 masks and masked_loop_p with a single loop_masks, making sure it's
13336 null for bb vectorization.
13338 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
13340 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
13341 (vect_analyze_data_ref_access): Use loop->safe_len rather than
13342 loop->force_vectorize to check whether there is no alias.
13344 2018-03-02 Jakub Jelinek <jakub@redhat.com>
13347 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
13349 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
13351 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
13352 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
13353 instead of a loop around prev_real_insn.
13354 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
13357 PR inline-asm/84625
13358 * config/i386/i386.c (ix86_print_operand): Use conditional
13359 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
13362 2018-03-02 Richard Biener <rguenther@suse.de>
13364 PR tree-optimization/84427
13365 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
13366 (bitmap_set_subtract_values): Rewrite to handle multiple
13369 (prune_clobbered_mems): Likewise.
13370 (phi_translate): Take edge instead of pred/phiblock.
13371 (phi_translate_1): Likewise.
13372 (phi_translate_set): Likewise. Insert all translated
13373 exprs for a value into the set, keeping possibly multiple
13374 expressions per value.
13375 (compute_antic_aux): Adjust for phi_translate changes.
13376 When intersecting union the expressions and prune those
13377 not in the final value set, keeping possibly multiple
13378 expressions per value. Do not use value-insertion
13379 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
13380 all expressions. Add verification that the value-sets
13381 only shrink during iteration.
13382 (compute_partial_antic_aux): Adjust for the phi_translate changes.
13383 (do_pre_regular_insertion): Likewise.
13384 (do_pre_partial_partial_insertion): Likewise.
13386 2018-03-02 Richard Biener <rguenther@suse.de>
13389 * config/darwin.c (saved_debug_info_level): New static global.
13390 (darwin_asm_lto_start): Disable debug info generation for LTO out.
13391 (darwin_asm_lto_end): Restore debug info generation settings.
13393 2018-03-01 Martin Liska <mliska@suse.cz>
13396 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
13397 volatile arguments.
13399 2018-03-01 Richard Biener <rguenther@suse.de>
13402 * dwarf2out.c (gen_variable_die): Properly handle late VLA
13403 type annotation with LTO when debug was disabled at compile-time.
13405 2018-03-01 Matthew Fortune <mfortune@gmail.com>
13407 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
13409 (mips_final_postscan_insn): Likewise.
13411 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
13413 PR rtl-optimization/84528
13414 * alias.c (init_alias_target): Add commentary.
13415 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
13416 a unique base value if the frame pointer is not eliminated
13417 to the stack pointer.
13419 2018-03-01 Tom de Vries <tom@codesourcery.com>
13421 PR rtl-optimization/83327
13422 * lra-int.h (hard_regs_spilled_into): Declare.
13423 * lra.c (hard_regs_spilled_into): Define.
13424 (init_reg_info): Init hard_regs_spilled_into.
13425 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
13426 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
13427 (process_bb_lives): Handle hard_regs_spilled_into.
13428 (lra_create_live_ranges_1): Before doing liveness propagation, clear
13429 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
13431 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
13433 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
13434 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
13435 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
13436 * config/rs6000/aix72.h: New file.
13438 2018-02-28 Jakub Jelinek <jakub@redhat.com>
13440 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
13441 instead of warning_at with conditional singular and plural messages
13445 * stor-layout.c (update_alignment_for_field): For
13446 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
13447 && !DECL_PACKED (field), do the alignment update, just use
13448 only desired_align instead of MAX (type_align, desired_align)
13450 (place_field): Don't do known_align < desired_align handling
13451 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
13452 is non-NULL, instead do it after rli->prev_field handling and
13453 only if not within a bitfield word. For DECL_PACKED (field)
13454 use type_align of BITS_PER_UNIT.
13456 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
13458 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
13459 superfluous parentheses and trailing spaces.
13461 2018-02-28 Richard Biener <rguenther@suse.de>
13463 PR tree-optimization/84584
13464 * graphite-scop-detection.c (scop_detection::add_scop): Discard
13465 SCoPs with fake exit edge.
13467 2018-02-28 Martin Liska <mliska@suse.cz>
13470 * timevar.c (timer::print): Fix format to properly print 100%
13473 2018-02-28 Richard Biener <rguenther@suse.de>
13475 PR middle-end/84607
13476 * genmatch.c (capture_info::walk_match): Do not mark
13477 captured expressions without operands as expr_p given
13478 they act more like predicates and should be subject to
13479 "lost tail" side-effect preserving.
13481 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
13483 PR rtl-optimization/81611
13484 * auto-inc-dec.c (attempt_change): Move dead note from
13485 mem_insn if it's the next use of regno
13486 (find_address): Take address use of reg holding
13487 non-incremented value. Add parm to limit search to the named
13489 (merge_in_block): Attempt to use a mem insn that is the next
13490 use of the original regno.
13492 2018-02-27 Martin Sebor <msebor@redhat.com>
13495 * doc/invoke.texi (-Wmissing-attributes): New option.
13496 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
13498 2018-02-27 Martin Sebor <msebor@redhat.com>
13500 PR translation/84207
13501 * diagnostic-core.h (warning_n, error_n, inform_n): Change
13502 n argument to unsigned HOST_WIDE_INT.
13503 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
13504 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
13505 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
13506 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
13508 2018-02-27 Richard Biener <rguenther@suse.de>
13510 PR tree-optimization/84512
13511 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
13512 Do not use the estimate returned from record_stmt_cost for
13513 the scalar iteration cost but sum properly using add_stmt_cost.
13515 2018-02-27 Richard Biener <rguenther@suse.de>
13517 PR tree-optimization/84466
13518 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
13519 Adjust last change to less strictly validate use operands.
13521 2018-02-27 Martin Liska <mliska@suse.cz>
13523 PR gcov-profile/84548
13524 * gcov.c (process_file): Allow partial overlap and consider it
13525 also as group functions.
13526 (output_lines): Properly calculate range of lines for a group.
13528 2018-02-27 Martin Liska <mliska@suse.cz>
13530 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
13531 'ggc' suffixes. Change first column width.
13532 (timer::print): Fix formatting of the column.
13534 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
13536 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
13537 preserve inline entry blocks for the sake of debug inline
13538 entry point markers alone.
13539 (remove_unused_locals): Suggest in comments a better place to
13540 force the preservation of inline entry blocks that are
13541 otherwise unused, but do not preserve them.
13543 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
13545 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
13547 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
13550 * config/i386/constraints.md (Bs): Replace
13551 ix86_indirect_branch_register with
13552 TARGET_INDIRECT_BRANCH_REGISTER.
13554 * config/i386/i386.md (indirect_jump): Likewise.
13555 (tablejump): Likewise.
13556 (*sibcall_memory): Likewise.
13557 (*sibcall_value_memory): Likewise.
13558 Peepholes of indirect call and jump via memory: Likewise.
13559 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
13560 (*sibcall_value_GOT_32): Likewise.
13561 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13562 (GOT_memory_operand): Likewise.
13563 (call_insn_operand): Likewise.
13564 (sibcall_insn_operand): Likewise.
13565 (GOT32_symbol_operand): Likewise.
13566 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
13568 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
13570 PR rtl-optimization/83496
13571 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
13572 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
13573 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
13574 redundant insn, if any.
13575 (relax_delay_slots): Likewise.
13576 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
13578 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
13580 PR tree-optimization/83965
13581 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
13582 that grouped statements are part of a reduction chain. Return
13583 true if the statement is not marked as a reduction itself but
13584 is part of a group.
13585 (vect_recog_dot_prod_pattern): Don't check whether the statement
13586 is part of a group here.
13587 (vect_recog_sad_pattern): Likewise.
13588 (vect_recog_widen_sum_pattern): Likewise.
13590 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
13593 * final.c (rest_of_clean_state): Also look for calls inside sequences.
13595 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
13598 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
13600 (ix86_output_indirect_function_return): New prototype.
13601 (ix86_split_simple_return_pop_internal): Likewise.
13602 * config/i386/i386.c (indirect_return_via_cx): New.
13603 (indirect_return_via_cx_bnd): Likewise.
13604 (indirect_thunk_name): Handle return va CX_REG.
13605 (output_indirect_thunk_function): Create alias for
13606 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
13607 (ix86_output_indirect_jmp): Remove the bool argument.
13608 (ix86_output_indirect_function_return): New function.
13609 (ix86_split_simple_return_pop_internal): Likewise.
13610 * config/i386/i386.md (*indirect_jump): Don't pass false
13611 to ix86_output_indirect_jmp.
13612 (*tablejump_1): Likewise.
13613 (simple_return_pop_internal): Change it to define_insn_and_split.
13614 Call ix86_split_simple_return_pop_internal to split it for
13615 -mfunction-return=.
13616 (simple_return_indirect_internal): Call
13617 ix86_output_indirect_function_return instead of
13618 ix86_output_indirect_jmp.
13620 2018-02-26 Jakub Jelinek <jakub@redhat.com>
13623 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
13624 memset and value initialization afterwards.
13626 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
13628 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
13630 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
13633 * common/config/aarch64/aarch64-common.c
13634 (aarch_option_optimization_table[]): Switch
13635 off fomit-frame-pointer
13637 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
13638 Chung-Ju Wu <jasonwucj@gmail.com>
13640 * config/nds32/nds32-multiple.md (load_multiple): Disallow
13642 (store_multiple): Ditto.
13644 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
13646 * config.gcc: Add --with-cpu support for nds32 target.
13647 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
13648 * config/nds32/nds32.opt: Add -mcpu= option.
13650 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
13652 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
13653 isel=yes): Warn for these deprecated options.
13655 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
13657 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
13658 ISA_2_5_MASKS_EMBEDDED.
13660 2018-02-23 Jakub Jelinek <jakub@redhat.com>
13662 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
13663 p->max as pointers rather than using iterative_hash_expr.
13665 2018-02-23 Carl Love <cel@us.ibm.com>
13667 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
13668 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
13670 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
13671 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
13672 P8V_BUILTIN_VEC_VUNSIGNED2.
13674 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
13677 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
13678 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
13680 * lra-constraints.c (curr_insn_transform): Set up
13681 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
13683 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
13684 * lra-eliminations.c (spill_pseudos): Ditto.
13685 (process_insn_for_elimination): Ditto.
13686 * lra-lives.c (reg_early_clobber_p): Use the new macros.
13687 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
13688 LRA_NON_CLOBBERED_ALT.
13690 2018-02-22 Martin Sebor <msebor@redhat.com>
13692 PR tree-optimization/84480
13693 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
13694 to maybe_diag_stxncpy_trunc. Call it.
13695 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
13696 from gimple_fold_builtin_strcpy. Print inlining stack.
13697 (handle_builtin_stxncpy): Print inlining stack.
13698 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
13700 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
13703 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
13704 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
13705 and -fcheck-pointer-bounds are used together.
13706 (indirect_thunk_prefix): New enum.
13707 (indirect_thunk_need_prefix): New function.
13708 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
13709 "_nt" instead of "_bnd" for NOTRACK prefix.
13710 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
13711 (output_indirect_thunk_function): Likewise.
13713 (ix86_code_end): Update output_indirect_thunk_function calls.
13714 (ix86_output_indirect_branch_via_reg): Replace
13715 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
13716 (ix86_output_indirect_branch_via_push): Likewise.
13717 (ix86_output_function_return): Likewise.
13718 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
13719 incompatible with -fcf-protection=branch and
13720 -fcheck-pointer-bounds.
13722 2018-02-22 Steve Ellcey <sellcey@cavium.com>
13725 * config/aarch64/aarch64.c (aarch64_print_address_internal):
13726 Change gcc_assert call to output_operand_lossage.
13728 2018-02-22 Steve Ellcey <sellcey@cavium.com>
13730 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
13732 2018-02-22 DJ Delorie <dj@redhat.com>
13733 Sebastian Perta <sebastian.perta@renesas.com>
13734 Oleg Endo <olegendo@gcc.gnu.org>
13736 * config/rx/rx.c (rx_rtx_costs): New function.
13737 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
13739 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
13741 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
13743 2018-02-22 Martin Liska <mliska@suse.cz>
13746 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
13747 Add "native" as a possible value.
13749 2018-02-22 Martin Liska <mliska@suse.cz>
13752 * config/i386/i386.c (ix86_option_override_internal):
13753 Add "native" as a possible value for -march and -mtune.
13755 2018-02-22 Jakub Jelinek <jakub@redhat.com>
13758 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
13759 to all type variants.
13761 PR tree-optimization/84503
13762 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
13763 width as info->bitpos + info->bitsize - start.
13764 (merged_store_group::merge_overlapping): Simplify width computation.
13765 (check_no_overlap): New function.
13766 (imm_store_chain_info::try_coalesce_bswap): Compute expected
13767 start + width and last_order of the group, fail if check_no_overlap
13769 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
13770 to group if check_no_overlap fails.
13772 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
13774 * config/rs6000/altivec.md: Delete contraint arguments to
13775 define_expand, define_split, and define_peephole2, and in
13776 define_insn_and_split if always unused.
13777 * config/rs6000/darwin.md: Ditto.
13778 * config/rs6000/dfp.md: Ditto.
13779 * config/rs6000/rs6000.md: Ditto.
13780 * config/rs6000/sync.md: Ditto.
13781 * config/rs6000/vector.md: Ditto.
13782 * config/rs6000/vsx.md: Ditto.
13784 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
13786 * config/rs6000/altivec.md: Write output control strings as braced
13787 blocks instead of double-quoted strings.
13788 * config/rs6000/darwin.md: Ditto.
13789 * config/rs6000/rs6000.md: Ditto.
13790 * config/rs6000/vector.md: Ditto.
13791 * config/rs6000/vsx.md: Ditto.
13793 2018-02-21 Jason Merrill <jason@redhat.com>
13795 PR c++/84314 - ICE with templates and fastcall attribute.
13796 * attribs.c (build_type_attribute_qual_variant): Remove assert.
13798 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
13800 * ipa-cp.c (determine_versionability): Fix comment typos.
13802 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
13805 * ipa-cp.c (determine_versionability): Do not version functions caling
13808 2018-02-21 Martin Liska <mliska@suse.cz>
13811 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
13812 Add "native" as a possible value.
13813 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
13814 the macro when native cpu detection is available.
13816 2018-02-21 Martin Liska <mliska@suse.cz>
13819 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
13820 Add "native" as a possible value.
13821 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
13822 when native cpu detection is available.
13824 2018-02-21 Jakub Jelinek <jakub@redhat.com>
13825 Martin Sebor <msebor@redhat.com>
13827 PR tree-optimization/84478
13828 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
13830 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
13831 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
13832 support which is conservatively correct, for 2 only stay conservative
13833 for maxlen. Formatting and comment capitalization fixes. Add STRICT
13834 argument to the 2 argument get_range_strlen, adjust 6 arg
13835 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
13837 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
13838 (gimple_fold_builtin_strlen): Pass true as last argument to
13841 2018-02-20 Martin Sebor <msebor@redhat.com>
13843 PR middle-end/84095
13844 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
13845 (builtin_memref::set_base_and_offset): Same. Handle inner references.
13846 (builtin_memref::builtin_memref): Factor out parts into
13847 set_base_and_offset and call it.
13849 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
13851 PR middle-end/84406
13852 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
13853 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
13854 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
13855 search at the associated MODE_INT.
13857 2018-02-20 Jeff Law <law@redhat.com>
13859 PR middle-end/82123
13860 PR tree-optimization/81592
13861 PR middle-end/79257
13862 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
13863 for range data rather than using global data.
13864 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
13865 range data rather than using global data.
13866 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
13867 pass it to children as needed.
13868 (struct directive::fmtresult): Similarly.
13869 (struct directive::set_width): Similarly.
13870 (struct directive::set_precision): Similarly.
13871 (format_integer, format_directive, parse_directive): Similarly.
13872 (format_none): Accept unnamed vr_values parameter.
13873 (format_percent, format_floating, format_character): Similarly.
13874 (format_string, format_plain): Similarly.
13875 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
13876 the EVRP range analyzer for range data rather than using global data.
13877 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
13878 gimple-ssa-evrp-analyze.h
13879 (class sprintf_dom_walker): Add after_dom_children member function.
13880 Add evrp_range_analyzer member.
13881 (sprintf_dom_walker::before_dom_children): Call into the EVRP
13882 range analyzer as needed.
13883 (sprintf_dom_walker::after_dom_children): New member function.
13884 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
13886 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
13887 (evrp_range_analyzer::pop_to_marker): Likewise.
13889 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
13891 PR tree-optimization/84419
13892 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
13893 with the required type if its current type is compatible but
13896 2018-02-20 Jakub Jelinek <jakub@redhat.com>
13898 PR middle-end/82004
13899 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
13900 after vectorization.
13902 2018-02-20 Martin Liska <mliska@suse.cz>
13905 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
13906 possible values if we don't have a hint.
13908 2018-02-20 Martin Liska <mliska@suse.cz>
13912 * final.c (shorten_branches): Build align_tab array with one
13914 * opts.c (finish_options): Add alignment option limit check.
13915 (MAX_CODE_ALIGN): Likewise.
13916 (MAX_CODE_ALIGN_VALUE): Likewise.
13917 * doc/invoke.texi: Document maximum allowed option value for
13918 all -falign-* options.
13920 2018-02-19 Jakub Jelinek <jakub@redhat.com>
13923 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
13924 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
13925 * var-tracking.c (emit_note_insn_var_location): Remove all references
13926 to NOTE_INSN_CALL_ARG_LOCATION.
13927 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
13928 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
13929 Use copy_rtx_if_shared.
13930 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
13931 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
13932 (dwarf2out_var_location): Remove handling of
13933 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
13935 * final.c (final_scan_insn): Remove all references to
13936 NOTE_INSN_CALL_ARG_LOCATION.
13937 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
13938 before dumping final insns.
13939 * except.c (emit_note_eh_region_end): Remove all references to
13940 NOTE_INSN_CALL_ARG_LOCATION.
13941 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
13942 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
13943 * config/arc/arc.c (hwloop_optimize): Likewise.
13944 * config/arm/arm.c (create_fix_barrier): Likewise.
13945 * config/s390/s390.c (s390_chunkify_start): Likewise.
13946 * config/sh/sh.c (find_barrier): Likewise.
13947 * config/i386/i386.c (rest_of_insert_endbranch,
13948 ix86_seh_fixup_eh_fallthru): Likewise.
13949 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
13950 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
13951 * config/frv/frv.c (frv_function_prologue): Likewise.
13952 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
13954 (note_outside_basic_block_p): Remove all references to
13955 NOTE_INSN_CALL_ARG_LOCATION.
13956 * gengtype.c (adjust_field_rtx_def): Likewise.
13957 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
13959 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
13960 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
13963 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
13966 PR tree-optimization/84452
13967 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
13968 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
13971 2018-02-19 Martin Liska <mliska@suse.cz>
13974 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
13976 2018-02-19 Martin Liska <mliska@suse.cz>
13977 Richard Sandiford <richard.sandiford@linaro.org>
13979 PR tree-optimization/82491
13980 * gimple-fold.c (get_base_constructor): Make earlier bail out
13983 2018-02-19 Carl Love <cel@us.ibm.com>
13985 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
13986 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
13988 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
13989 P8V_BUILTIN_VEC_NEG.
13991 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
13993 * config/rl78/rl78.md (movdf): New define expand.
13995 2018-02-19 Martin Liska <mliska@suse.cz>
13998 * doc/invoke.texi: Fix typo.
13999 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
14001 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
14003 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
14004 handle rs6000_single_float and rs6000_double_float specially for
14007 2018-02-16 Jeff Law <law@redhat.com>
14009 * config/rx/rx.c (add_pop_cfi_notes): New function.;
14010 (pop_regs): Use it.
14012 2018-02-16 Jakub Jelinek <jakub@redhat.com>
14015 * ipa-inline.c (inline_small_functions): Fix a typo.
14017 2018-02-16 Nathan Sidwell <nathan@acm.org>
14019 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
14021 2018-02-16 Carl Love <cel@us.ibm.com>
14023 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
14024 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
14025 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
14026 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
14027 expansion to P8V_BUILTIN_VEC_FLOAT2.
14029 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
14031 PR rtl-optimization/70023
14032 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
14033 src_regno into account.
14035 2018-02-16 Carl Love <cel@us.ibm.com>
14037 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
14038 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
14039 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
14040 * config/rs6000/rs6000.c: Remove case statements for
14041 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
14042 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
14043 and P9V_BUILTIN_VEC_VINSERT4B.
14044 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
14045 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
14046 * config/rs6000/vsx.md:
14047 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
14050 2018-02-16 Carl Love <cel@us.ibm.com>
14052 * config/rs6000/altivec.h: Add builtin names vec_extract4b
14054 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
14056 * config/rs6000/rs6000-c.c: Add the definitions for
14057 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
14058 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
14059 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
14060 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
14061 definition for insert4b and define insn *insert3b_internal.
14062 * doc/extend.texi: Add documentation for vec_extract4b.
14064 2018-02-16 Nathan Sidwell <nathan@acm.org>
14066 * doc/extend.texi (Backwards Compatibility): Mention friend
14067 injection. Note for-scope is deprecated.
14068 * doc/invoke.texi (-ffriend-injection): Deprecate.
14070 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
14072 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
14073 that moved to I2, also allow destinations that are a paradoxical
14074 subreg (instead of a normal reg).
14076 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
14079 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
14082 2018-02-16 Richard Biener <rguenther@suse.de>
14084 PR tree-optimization/84037
14085 PR tree-optimization/84016
14087 * config/i386/i386.c (ix86_builtin_vectorization_cost):
14088 Adjust vec_construct for the fact we need additional higher latency
14089 128bit inserts for AVX256 and AVX512 vector builds.
14090 (ix86_add_stmt_cost): Scale vector construction cost for
14093 2018-02-16 Richard Biener <rguenther@suse.de>
14095 PR tree-optimization/84417
14096 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
14097 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
14098 (non_rewritable_lvalue_p): Likewise, use poly-ints.
14100 2018-02-16 Martin Liska <mliska@suse.cz>
14103 * internal-fn.def (ASAN_CHECK): Set proper flags.
14104 (ASAN_MARK): Likewise.
14106 2018-02-16 Julia Koval <julia.koval@intel.com>
14108 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
14109 from PTA_CANNONLAKE.
14111 2018-02-16 Jakub Jelinek <jakub@redhat.com>
14114 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
14115 Use ++iter rather than iter++ for std::list iterators.
14116 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
14117 defer deleting them until all nodes in the forest are processed. Do
14118 free even leaf nodes. Change to_process into auto_vec.
14121 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
14122 * vec.h (vec_default_construct): Use memset instead of placement new
14123 if BROKEN_VALUE_INITIALIZATION is defined.
14124 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
14125 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
14128 PR rtl-optimization/83723
14129 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
14130 * lra.c (lra_substitute_pseudo): Likewise. If true, use
14131 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
14133 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
14135 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
14137 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
14139 PR rtl-optimization/81443
14140 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
14141 from inner REGs to paradoxical SUBREGs.
14143 2018-02-16 Richard Biener <rguenther@suse.de>
14145 PR tree-optimization/84399
14146 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
14147 For operands we can analyze at their definition make sure we can
14148 analyze them at each use as well.
14150 2018-02-16 Richard Biener <rguenther@suse.de>
14152 PR tree-optimization/84190
14153 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
14154 volatile accesses if the decl isn't volatile.
14156 2018-02-15 Jason Merrill <jason@redhat.com>
14158 PR c++/84314 - ICE with templates and fastcall attribute.
14159 * attribs.c (build_type_attribute_qual_variant): Don't clobber
14160 TYPE_CANONICAL on an existing type.
14162 2018-02-15 Jakub Jelinek <jakub@redhat.com>
14164 PR tree-optimization/84383
14165 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
14166 dstoff nor call operand_equal_p if dstbase is NULL.
14168 PR tree-optimization/84334
14169 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
14170 also a CONSTANT_CLASS_P, punt.
14172 2018-02-14 Jim Wilson <jimw@sifive.com>
14174 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
14175 first SMALL_OPERAND check. New local min_second_step. Move assert
14176 to where locals are set. Add TARGET_RVC support.
14177 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
14179 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
14181 * doc/invoke.texi: Correct -Wformat-overflow code sample.
14183 2018-02-14 Martin Sebor <msebor@redhat.com>
14185 PR tree-optimization/83698
14186 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
14187 arrays constrain the offset range to their bounds.
14188 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
14189 (builtin_access::overlap): Avoid setting the size of overlap if it's
14191 (maybe_diag_overlap): Also consider arrays when deciding what values
14192 of offsets to include in diagnostics.
14194 2018-02-14 Martin Sebor <msebor@redhat.com>
14197 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
14198 that correspond to the kind of a declaration.
14200 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
14203 * config/pa/pa.md: Load address of PIC label using the linkage table
14204 if the label is nonlocal.
14206 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
14208 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
14209 warning message if user requests -maltivec=be.
14210 * doc/invoke.texi: Document deprecation of -maltivec=be.
14212 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
14215 * config/rs6000/rs6000-c.c: Update definitions for
14216 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
14217 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
14219 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
14222 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
14223 add _get_ssp intrinsics. Remove argument from
14224 __builtin_ia32_rdssp[d|q].
14225 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
14226 * config/i386/i386-builtin.def: Remove argument from
14227 __builtin_ia32_rdssp[d|q].
14228 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
14229 ix86_expand_special_args_builtin for _rdssp[d|q].
14230 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
14231 Clear register before usage.
14232 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
14233 Add documentation for new _get_ssp and _inc_ssp intrinsics.
14235 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
14237 PR tree-optimization/84357
14238 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
14239 operand 1 of an ARRAY_REF too.
14241 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
14244 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
14245 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
14247 (set_of_reg): New struct.
14248 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
14249 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
14250 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
14252 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
14253 Split into bitclr, bitset, bitinvert patterns if appropriate.
14254 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
14255 use rx_fuse_in_memory_bitop.
14256 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
14257 to named insn, correct maximum insn length.
14259 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
14262 * machmode.def: Define a complex mode for PARTIAL_INT.
14263 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
14265 * doc/rtl.texi: Document CSPImode.
14266 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
14268 (msp430_hard_regno_nregs_with_padding): Likewise.
14270 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
14273 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
14275 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
14277 PR rtl-optimization/84169
14278 * combine.c (try_combine): New variable split_i2i3. Set it to true if
14279 we generated a parallel as new i3 and we split that to new i2 and i3
14280 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
14281 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
14282 those to i2, not i1. Partially rewrite this scan code.
14284 2018-02-13 Jakub Jelinek <jakub@redhat.com>
14287 * stor-layout.c (place_field): For variable length fields, adjust
14288 offset_align afterwards not just based on the field's alignment,
14289 but also on the size.
14291 PR middle-end/84309
14292 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
14293 of exps and logs in the use_exp2 case.
14295 2018-02-13 Jeff Law <law@redhat.com>
14297 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
14298 entry for "vector".
14300 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
14303 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
14307 * common.opt (gas-loc-support, gas-locview-support): New.
14308 (ginline-points, ginternal-reset-location-views): New.
14309 * doc/invoke.texi: Document them. Use @itemx where intended.
14310 (gvariable-location-views): Adjust.
14311 * target.def (reset_location_view): New.
14312 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
14313 (TARGET_RESET_LOCATION_VIEW): New.
14314 * doc/tm.texi: Rebuilt.
14315 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
14316 (dwarf2out_default_as_locview_support): New.
14317 (output_asm_line_debug_info): Use option variables.
14318 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
14319 (output_loc_list): Likewise.
14320 (add_high_low_attributes): Check option variables.
14321 Don't output entry view attribute in strict mode.
14322 (gen_inlined_subroutine_die): Check option variables.
14323 (dwarf2out_inline_entry): Likewise.
14324 (init_sections_and_labels): Likewise.
14325 (dwarf2out_early_finish): Likewise.
14326 (maybe_reset_location_view): New, from...
14327 (dwarf2out_var_location): ... here. Call it.
14328 * debug.h (dwarf2out_default_as_loc_support): Declare.
14329 (dwarf2out_default_as_locview_support): Declare.
14330 * hooks.c (hook_int_rtx_insn_0): New.
14331 * hooks.h (hook_int_rtx_insn_0): Declare.
14332 * toplev.c (process_options): Take -gas-loc-support and
14333 -gas-locview-support from dwarf2out. Enable
14334 -gvariable-location-views by default only with locview
14335 assembler support. Enable -ginternal-reset-location-views by
14336 default only if the target defines the corresponding hook.
14337 Enable -ginline-points by default if location views are
14338 enabled; force it disabled if statement frontiers are
14340 * tree-inline.c (expand_call_inline): Check option variables.
14341 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
14343 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
14345 PR tree-optimization/84321
14346 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
14347 handling. Also check whether the anti-range contains any values
14348 that satisfy the mask; switch to a VR_RANGE if not.
14350 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
14353 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
14355 2018-02-13 Martin Jambor <mjambor@suse.cz>
14358 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
14359 of call statements, also set location of a load to a temporary.
14361 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
14363 * config/rl78/rl78.c (add_vector_labels): New function.
14364 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
14365 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
14366 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
14367 which checks that no arguments are passed.
14368 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
14369 * doc/extend.texi: Documentation for the new attribute.
14371 2018-02-13 Andreas Schwab <schwab@suse.de>
14373 * config/riscv/linux.h (CPP_SPEC): Define.
14375 2018-02-13 Jakub Jelinek <jakub@redhat.com>
14378 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
14379 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
14380 OPTION_MASK_ISA_AES as first argument to def_builtin_const
14381 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
14382 instead of OPTION_MASK_ISA_PCLMUL as first argument to
14383 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
14384 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
14385 temporarily for AES and PCLMUL builtins.
14387 PR tree-optimization/84339
14388 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
14389 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
14392 PR middle-end/84309
14393 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
14394 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
14395 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
14397 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
14399 * omp-simd-clone.h: New file.
14400 * omp-simd-clone.c: Include omp-simd-clone.h.
14401 (expand_simd_clones): No longer static.
14402 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
14403 cgraph.h and omp-simd-clone.h.
14404 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
14405 (vect_recog_widen_shift_pattern): Formatting fix.
14406 (vect_pattern_recog_1): Don't check optab for calls.
14409 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
14410 operands[2] into a REG before using gen_lowpart on it.
14412 2018-02-12 Jeff Law <law@redhat.com>
14415 * config/sh/sh.c (find_barrier): Consider a sibling call
14418 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
14419 successfully back substituting a reg.
14421 2018-02-12 Richard Biener <rguenther@suse.de>
14423 PR tree-optimization/84037
14424 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
14425 parameter, move visited init to caller.
14426 (vect_slp_analyze_operations): Separate cost from validity
14427 check, initialize visited once for all instances.
14428 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
14430 * tree-vect-stmts.c (vect_model_simple_cost): Make early
14432 (vect_model_promotion_demotion_cost): Likewise.
14433 (vectorizable_bswap): Guard cost modeling with !slp_node
14434 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
14436 (vectorizable_call): Likewise.
14437 (vectorizable_conversion): Likewise.
14438 (vectorizable_assignment): Likewise.
14439 (vectorizable_shift): Likewise.
14440 (vectorizable_operation): Likewise.
14441 (vectorizable_store): Likewise.
14442 (vectorizable_load): Likewise.
14443 (vectorizable_condition): Likewise.
14444 (vectorizable_comparison): Likewise.
14446 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
14449 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
14450 (ASAN_MARK): Fix fnspec to account for return value, change pointer
14451 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
14453 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
14455 PR middle-end/83665
14456 * params.def (inline-min-speedup): Increase from 8 to 15.
14457 (max-inline-insns-auto): Decrease from 40 to 30.
14458 * ipa-split.c (consider_split): Add some buffer for function to
14459 be considered inlining candidate.
14460 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
14463 2018-02-12 Richard Biener <rguenther@suse.de>
14465 PR tree-optimization/84037
14466 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
14467 matched stmts if we cannot swap the non-matched ones.
14469 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
14471 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
14472 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
14473 _mm_maskz_scalef_round_ss): New intrinsics.
14474 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
14475 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
14476 __builtin_ia32_scalefss_round): Remove.
14477 (__builtin_ia32_scalefsd_mask_round,
14478 __builtin_ia32_scalefss_mask_round): New intrinsics.
14479 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
14480 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
14481 ((match_operand:VF_128 2 "<round_nimm_predicate>"
14482 "<round_constraint>")): Changed to ...
14483 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
14484 "<round_scalar_constraint>")): ... this.
14485 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
14486 %0, %1, %2<round_op3>}"): Changed to ...
14487 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
14488 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
14489 %2<round_scalar_mask_op3>}"): ... this.
14490 * config/i386/subst.md (round_scalar_nimm_predicate): New.
14492 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
14494 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
14495 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
14496 (_mm_maskz_sqrt_round_ss): New intrinsics.
14497 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
14498 (__builtin_ia32_sqrtsd_mask_round)
14499 (__builtin_ia32_sqrtss_mask_round): New builtins.
14500 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
14501 (__builtin_ia32_sqrtss_round): Remove.
14502 (__builtin_ia32_sqrtsd_mask_round)
14503 (__builtin_ia32_sqrtss_mask_round): New builtins.
14504 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
14505 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
14506 ((match_operand:VF_128 1 "vector_operand"
14507 "xBm,<round_constraint>")): Changed to ...
14508 ((match_operand:VF_128 1 "vector_operand"
14509 "xBm,<round_scalar_constraint>")): ... this.
14510 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
14511 %0, %2, %<iptr>1<round_op3>}): Changed to ...
14512 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
14513 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
14514 %<iptr>1<round_scalar_mask_op3>}): ... this.
14515 ((set_attr "prefix" "<round_prefix>")): Changed to ...
14516 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
14518 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
14521 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
14522 Cast vec_cmpeq result to correct type.
14523 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
14524 Cast vec_cmpgt result to correct type.
14526 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
14528 * final.c (final_scan_insn_1): Renamed from...
14529 (final_scan_insn): ... this. New wrapper, to recover
14530 seen from the outermost call in recursive ones.
14531 * config/sparc/sparc.c (output_return): Drop seen from call.
14532 (output_sibcall): Likewise.
14533 * config/visium/visium.c (output_branch): Likewise.
14535 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
14537 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
14540 2018-02-10 Alan Modra <amodra@gmail.com>
14543 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
14544 Specify LR as an input.
14546 2018-02-10 Jakub Jelinek <jakub@redhat.com>
14549 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
14550 remove_member_access_dummy_vars): New functions.
14551 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
14552 lower_omp_1, execute_lower_omp): Use them.
14554 PR rtl-optimization/84308
14555 * shrink-wrap.c (spread_components): Release todo vector.
14557 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
14559 PR rtl-optimization/57193
14560 * ira-color.c (struct allocno_color_data): Add member
14561 conflict_allocno_hard_prefs.
14562 (update_conflict_allocno_hard_prefs): New.
14563 (bucket_allocno_compare_func): Add a preference based on
14564 conflict_allocno_hard_prefs.
14565 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
14566 (color_allocnos): Remove a dead code. Initiate
14567 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
14569 2018-02-09 Jakub Jelinek <jakub@redhat.com>
14572 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
14573 constraint from =wa to wa. Avoid a subreg on the output operand,
14574 instead use a pseudo and subreg it in a move.
14575 (p9_xxbrd_<mode>): Changed to ...
14576 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
14577 (p9_xxbrd_v2df): New expander.
14578 (p9_xxbrw_<mode>): Changed to ...
14579 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
14580 (p9_xxbrw_v4sf): New expander.
14582 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
14584 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
14586 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
14589 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
14590 multiply in 32-bit mode.
14591 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
14592 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
14595 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
14597 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
14598 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
14599 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
14600 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
14602 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
14605 * dwarf2out.c (is_trivial_indirect_ref): New function.
14606 (dwarf2out_late_global_decl): Do not generate a location
14607 attribute for variables that have a non-trivial DECL_VALUE_EXPR
14608 and that are not defined in the current unit.
14610 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
14612 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
14613 instead of a libcall for UNORDERED.
14615 2018-02-09 Tamar Christina <tamar.christina@arm.com>
14618 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
14619 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
14621 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14624 * config/s390/s390.c (s390_set_current_function): Invoke
14625 s390_indirect_branch_settings also if fndecl didn't change.
14627 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
14629 * config/rs6000/rs6000.md (blockage): Set length to zero.
14631 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
14633 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
14635 2018-02-09 Jakub Jelinek <jakub@redhat.com>
14638 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
14639 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
14643 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
14644 PARALLEL incoming that failed vt_get_decl_and_offset check.
14646 PR middle-end/84237
14647 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
14648 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
14650 (get_variable_section): For decls in named .bss* sections pass true as
14651 second argument to bss_initializer_p.
14653 2018-02-09 Marek Polacek <polacek@redhat.com>
14654 Jakub Jelinek <jakub@redhat.com>
14657 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
14658 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
14659 Sync some changes from cxx_fold_indirect_ref.
14661 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
14663 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
14665 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
14666 (BLOCK_INLINE_ENTRY_LABEL): New.
14667 (dwarf2out_var_location): Disregard inline entry markers.
14668 (inline_entry_data): New struct.
14669 (inline_entry_data_hasher): New hashtable type.
14670 (inline_entry_data_hasher::hash): New.
14671 (inline_entry_data_hasher::equal): New.
14672 (inline_entry_data_table): New variable.
14673 (add_high_low_attributes): Add DW_AT_entry_pc and
14674 DW_AT_GNU_entry_view attributes if a pending entry is found
14675 in inline_entry_data_table. Add old entry_pc attribute only
14676 if debug nonbinding markers are disabled.
14677 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
14678 markers are enabled.
14679 (block_within_block_p, dwarf2out_inline_entry): New.
14680 (dwarf2out_finish): Check that no entries remained in
14681 inline_entry_data_table.
14682 * final.c (reemit_insn_block_notes): Handle inline entry notes.
14683 (final_scan_insn, notice_source_line): Likewise.
14684 (rest_of_clean_state): Skip inline entry markers.
14685 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
14687 * gimple.c (gimple_build_debug_inline_entry): New.
14688 * gimple.h (enum gimple_debug_subcode): Add
14689 GIMPLE_DEBUG_INLINE_ENTRY.
14690 (gimple_build_debug_inline_entry): Declare.
14691 (gimple_debug_inline_entry_p): New.
14692 (gimple_debug_nonbind_marker_p): Adjust.
14693 * insn-notes.def (INLINE_ENTRY): New.
14694 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
14695 inline entry marker notes.
14696 (print_insn): Likewise.
14697 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
14698 (INSN_DEBUG_MARKER_KIND): Likewise.
14699 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
14700 * tree-inline.c (expand_call_inline): Build and insert
14701 debug_inline_entry stmt.
14702 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
14703 inline entry blocks early, if nonbind markers are enabled.
14704 (dump_scope_block): Dump fragment info.
14705 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
14706 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
14707 (gimple_build_debug_inline_entry): New.
14708 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
14709 Enable/disable inline entry points too.
14710 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
14711 (DEBUG_INSN): Describe inline entry markers.
14713 * common.opt (gvariable-location-views): New.
14714 (gvariable-location-views=incompat5): New.
14715 * config.in: Rebuilt.
14716 * configure: Rebuilt.
14717 * configure.ac: Test assembler for view support.
14718 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
14719 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
14720 * dwarf2out.c (var_loc_view): New typedef.
14721 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
14722 (dwarf2out_locviews_in_attribute): New.
14723 (dwarf2out_locviews_in_loclist): New.
14724 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
14725 (enum dw_line_info_opcode): Add LI_adv_address.
14726 (struct dw_line_info_table): Add view.
14727 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
14728 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
14729 (zero_view_p): New variable.
14730 (ZERO_VIEW_P): New macro.
14731 (output_asm_line_debug_info): New.
14732 (struct var_loc_node): Add view.
14733 (add_AT_view_list, AT_loc_list): New.
14734 (add_var_loc_to_decl): Add view param. Test it against last.
14735 (new_loc_list): Add view params. Record them.
14736 (AT_loc_list_ptr): Handle loc and view lists.
14737 (view_list_to_loc_list_val_node): New.
14738 (print_dw_val): Handle dw_val_class_view_list.
14739 (size_of_die): Likewise.
14740 (value_format): Likewise.
14741 (loc_list_has_views): New.
14742 (gen_llsym): Set vl_symbol too.
14743 (maybe_gen_llsym, skip_loc_list_entry): New.
14744 (dwarf2out_maybe_output_loclist_view_pair): New.
14745 (output_loc_list): Output view list or entries too.
14746 (output_view_list_offset): New.
14747 (output_die): Handle dw_val_class_view_list.
14748 (output_dwarf_version): New.
14749 (output_compilation_unit_header): Use it.
14750 (output_skeleton_debug_sections): Likewise.
14751 (output_rnglists, output_line_info): Likewise.
14752 (output_pubnames, output_aranges): Update version comments.
14753 (output_one_line_info_table): Output view numbers in asm comments.
14754 (dw_loc_list): Determine current endview, pass it to new_loc_list.
14755 Call maybe_gen_llsym.
14756 (loc_list_from_tree_1): Adjust.
14757 (add_AT_location_description): Create view list attribute if
14758 needed, check it's absent otherwise.
14759 (convert_cfa_to_fb_loc_list): Adjust.
14760 (maybe_emit_file): Call output_asm_line_debug_info for test.
14761 (dwarf2out_var_location): Reset views as needed. Precompute
14762 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
14763 attribute. Set view.
14764 (new_line_info_table): Reset next view.
14765 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
14766 (dwarf2out_source_line): Likewise. Output view resets and labels to
14767 the assembler, or select appropriate line info opcodes.
14768 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
14769 (optimize_string_length): Catch it. Adjust.
14770 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
14771 dw_val_class_view_list, and remove it if no longer needed.
14772 (hash_loc_list): Hash view numbers.
14773 (loc_list_hasher::equal): Compare them.
14774 (optimize_location_lists): Check whether a view list symbol is
14775 needed, and whether the locview attribute is present, and
14776 whether they match. Remove the locview attribute if no longer
14778 (index_location_lists): Call skip_loc_list_entry for test.
14779 (dwarf2out_finish): Call output_asm_line_debug_info for test.
14780 Use output_dwarf_version.
14781 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
14782 (struct dw_val_node): Add val_view_list.
14783 * final.c (SEEN_NEXT_VIEW): New.
14784 (set_next_view_needed): New.
14785 (clear_next_view_needed): New.
14786 (maybe_output_next_view): New.
14787 (final_start_function): Rename to...
14788 (final_start_function_1): ... this. Take pointer to FIRST,
14789 add SEEN parameter. Emit param bindings in the initial view.
14790 (final_start_function): Reintroduce SEEN-less interface.
14791 (final): Rename to...
14792 (final_1): ... this. Take SEEN parameter. Output final pending
14793 next view at the end.
14794 (final): Reintroduce seen-less interface.
14795 (final_scan_insn): Output pending next view before switching
14796 sections or ending a block. Mark the next view as needed when
14797 outputting variable locations. Notify debug backend of section
14798 changes, and of location view changes.
14799 (rest_of_handle_final): Adjust.
14800 * toplev.c (process_options): Autodetect value for debug variable
14801 location views option. Warn on incompat5 without -gdwarf-5.
14802 * doc/invoke.texi (gvariable-location-views): New.
14803 (gvariable-location-views=incompat5): New.
14804 (gno-variable-location-views): New.
14806 2018-02-08 David Malcolm <dmalcolm@redhat.com>
14808 PR tree-optimization/84136
14809 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
14810 that the result of find_edge is non-NULL.
14812 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
14815 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
14816 storing integer register in SImode. Fix cost of 256 and 512
14817 byte aligned SSE register store.
14819 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
14821 * config/i386/i386.c (ix86_multiplication_cost): Fix
14822 multiplication cost for TARGET_AVX512DQ.
14824 2018-02-08 Marek Polacek <polacek@redhat.com>
14826 PR tree-optimization/84238
14827 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
14830 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
14832 PR tree-optimization/84265
14833 * tree-vect-stmts.c (vectorizable_store): Don't treat
14834 VMAT_CONTIGUOUS accesses as grouped.
14835 (vectorizable_load): Likewise.
14837 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
14839 PR tree-optimization/81635
14840 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
14841 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
14842 (test_round_for_mask): New functions.
14843 (wide_int_cc_tests): Call test_round_for_mask.
14844 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
14845 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
14846 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
14847 range returned by get_range_info.
14849 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
14852 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
14853 * symtab.c: Include builtins.h
14854 (symtab_node::output_to_lto_symbol_table_p): Move here
14855 from lto-streamer-out.c:output_symbol_p.
14856 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
14857 (output_symbol_p): Move all logic to symtab.c
14858 (produce_symtab): Update.
14860 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14862 * config/s390/s390-opts.h (enum indirect_branch): Define.
14863 * config/s390/s390-protos.h (s390_return_addr_from_memory)
14864 (s390_indirect_branch_via_thunk)
14865 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
14866 (enum s390_indirect_branch_type): Define.
14867 * config/s390/s390.c (struct s390_frame_layout, struct
14868 machine_function): Remove.
14869 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
14870 (indirect_branch_table_label_no, indirect_branch_table_name):
14872 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
14873 (enum s390_indirect_branch_option): Define.
14874 (s390_return_addr_from_memory): New function.
14875 (s390_handle_string_attribute): New function.
14876 (s390_attribute_table): Add new attribute handler.
14877 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
14878 (s390_indirect_branch_via_thunk): New function.
14879 (s390_indirect_branch_via_inline_thunk): New function.
14880 (s390_function_ok_for_sibcall): When jumping via thunk disallow
14881 sibling call optimization for non z10 compiles.
14882 (s390_emit_call): Force indirect branch target to be a single
14883 register. Add r1 clobber for non-z10 compiles.
14884 (s390_emit_epilogue): Emit return jump via return_use expander.
14885 (s390_reorg): Handle JUMP_INSNs as execute targets.
14886 (s390_option_override_internal): Perform validity checks for the
14887 new command line options.
14888 (s390_indirect_branch_attrvalue): New function.
14889 (s390_indirect_branch_settings): New function.
14890 (s390_set_current_function): Invoke s390_indirect_branch_settings.
14891 (s390_output_indirect_thunk_function): New function.
14892 (s390_code_end): Implement target hook.
14893 (s390_case_values_threshold): Implement target hook.
14894 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
14896 * config/s390/s390.h (struct s390_frame_layout)
14897 (struct machine_function): Move here from s390.c.
14898 (TARGET_INDIRECT_BRANCH_NOBP_RET)
14899 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
14900 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
14901 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
14902 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
14903 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
14904 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
14905 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
14906 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
14907 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
14908 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
14909 (mnemonic attribute): Add values which aren't recognized
14911 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
14912 pattern for branch conversion. Fix mnemonic attribute.
14913 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
14914 indirect branch via thunk if requested.
14915 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
14916 ("*indirect_jump"): Disable for branch conversion using out of
14918 ("indirect_jump_via_thunk<mode>_z10")
14919 ("indirect_jump_via_thunk<mode>")
14920 ("indirect_jump_via_inlinethunk<mode>_z10")
14921 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
14922 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
14923 ("casesi_jump_via_inlinethunk<mode>_z10")
14924 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
14925 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
14926 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
14927 ("*indirect2_jump"): Disable for branch conversion.
14928 ("casesi_jump"): Turn into expander and expand patterns for branch
14930 ("return_use"): New expander.
14931 ("*return"): Emit return via thunk and rename it to ...
14932 ("*return<mode>"): ... this one.
14933 * config/s390/s390.opt: Add new options and and enum for the
14936 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
14938 * lra-constraints.c (match_reload): Unconditionally use
14939 gen_lowpart_SUBREG, rather than selecting between that
14940 and equivalent gen_rtx_SUBREG code.
14942 2018-02-08 Richard Biener <rguenther@suse.de>
14944 PR tree-optimization/84233
14945 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
14946 changed flag instead of boguously re-using phi_inserted.
14948 2018-02-08 Martin Jambor <mjambor@suse.cz>
14950 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
14951 static local variables.
14953 2018-02-08 Richard Biener <rguenther@suse.de>
14955 PR tree-optimization/84278
14956 * tree-vect-stmts.c (vectorizable_store): When looking for
14957 smaller vector types to perform grouped strided loads/stores
14958 make sure the mode is supported by the target.
14959 (vectorizable_load): Likewise.
14961 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
14963 * config/aarch64/aarch64.c (aarch64_components_for_bb):
14964 Increase LDP/STP opportunities by adding adjacent callee-saves.
14966 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
14968 PR rtl-optimization/84068
14969 PR rtl-optimization/83459
14970 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
14972 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
14974 PR tree-optimization/84224
14975 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
14976 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
14977 non-zero arguments.
14979 2018-02-07 Iain Sandoe <iain@codesourcery.com>
14982 * config/rs6000/altivec.md (*restore_world): Remove LR use.
14983 * config/rs6000/predicates.md (restore_world_operation): Adjust op
14984 count, remove one USE.
14986 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
14988 * doc/install.texi (Configuration): Document the
14989 --with-long-double-format={ibm,ieee} PowerPC configuration
14993 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
14994 Convert from define_expand to be define_insn_and_split. Rework
14995 float/double/_Float128 conversions to QI/HI/SImode to work with
14996 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
14997 conversions to QI/HImode types did a store and then a load to
14998 truncate the value. For conversions to VSX registers, don't split
14999 the insn, instead emit the code directly. Use the code iterator
15000 any_fix to combine signed and unsigned conversions.
15001 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
15002 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
15003 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
15004 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
15005 (fix_<mode>di2_hw): Likewise.
15006 (fixuns_<mode>di2_hw): Likewise.
15007 (fix_<mode>si2_hw): Likewise.
15008 (fixuns_<mode>si2_hw): Likewise.
15009 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
15010 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
15011 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
15012 fix<uns>_trunc<SFDF:mode>si2_p8.
15013 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
15014 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
15015 (fix<uns>_<mode>_mem): Likewise.
15016 (fctiw<u>z_<mode>_mem): Likewise.
15017 (fix<uns>_<mode>_mem): Likewise.
15018 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
15019 the register allocator from doing a direct move to the GPRs to do
15020 a store, and instead use the ISA 3.0 store byte/half-word from
15021 vector register instruction. For IEEE 128-bit floating point,
15022 also optimize stores of 32-bit ints.
15023 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
15025 2018-02-07 Alan Hayward <alan.hayward@arm.com>
15027 * genextract.c (push_pathstr_operand): New function to support
15029 (walk_rtx): Call push_pathstr_operand.
15030 (print_path): Support [a-zA-Z].
15032 2018-02-07 Richard Biener <rguenther@suse.de>
15034 PR tree-optimization/84037
15035 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
15036 (cse_and_gimplify_to_preheader): Declare.
15037 (vect_get_place_in_interleaving_chain): Likewise.
15038 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15040 (_loop_vec_info::~_loop_vec_info): Delete it.
15041 (cse_and_gimplify_to_preheader): New function.
15042 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
15043 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
15044 (vectorizable_load): Likewise. For grouped stores always base
15045 the IV on the first element.
15046 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
15047 condition before gimplifying.
15049 2018-02-07 Jakub Jelinek <jakub@redhat.com>
15051 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
15052 *DIV_EXPR and *MOD_EXPR.
15054 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
15057 * config/i386/i386.c (ix86_option_override_internal): Mask out
15058 the CF_SET bit when checking -fcf-protection.
15060 2018-02-07 Tom de Vries <tom@codesourcery.com>
15063 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
15066 2018-02-07 Richard Biener <rguenther@suse.de>
15068 PR tree-optimization/84204
15069 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
15072 PR tree-optimization/84205
15073 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
15074 special-case isl_ast_op_zdiv_r.
15076 PR tree-optimization/84223
15077 * graphite-scop-detection.c (gather_bbs::before_dom_children):
15078 Only add conditions from within the region.
15079 (gather_bbs::after_dom_children): Adjust.
15081 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
15084 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
15085 * config/avr/avr.md: Only post-reload split REG-REG moves if
15086 either register is GENERAL_REG_P.
15088 2018-02-07 Jakub Jelinek <jakub@redhat.com>
15090 PR tree-optimization/84235
15091 * tree-ssa-scopedtables.c
15092 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
15093 if the subtraction is performed in floating point type where NaNs are
15094 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
15095 build 1. Formatting fix.
15097 2018-02-06 Jakub Jelinek <jakub@redhat.com>
15100 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
15101 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
15102 and skip it regardless of bb boundaries. Use CALL_P macro,
15103 don't test INSN_P (insn) together with CALL_P or JUMP_P check
15104 unnecessarily, formatting fix.
15106 2018-02-06 Michael Collison <michael.collison@arm.com>
15108 * config/arm/thumb2.md:
15109 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
15110 (*thumb_mov_notscc): Ditto.
15112 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
15115 * config/rs6000/rs6000.md (su code attribute): Use "u" for
15116 unsigned_fix, not "s".
15118 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15120 * configure.ac (gcc_fn_eh_frame_ro): New function.
15121 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
15122 correct .eh_frame permissions.
15123 * configure: Regenerate.
15125 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
15127 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
15128 irrelevant options.
15130 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15132 * config/rs6000/rs6000.c (rs6000_option_override_internal):
15133 Display warning message for -mno-speculate-indirect-jumps.
15135 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
15137 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
15139 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
15141 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
15143 PR tree-optimization/84225
15144 * tree-eh.c (find_trapping_overflow): Only call
15145 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
15147 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
15150 * config/i386/i386.c: Reimplement the check of possible options
15151 -mibt/-mshstk conbination. Change error messages.
15152 * doc/invoke.texi: Fix a typo: remove extra '='.
15154 2018-02-06 Marek Polacek <polacek@redhat.com>
15156 PR tree-optimization/84228
15157 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
15159 2018-02-06 Tamar Christina <tamar.christina@arm.com>
15162 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
15163 emitted arch directives.
15164 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
15165 __ARM_FEATURE_COPROC before changing architectures.
15167 2018-02-06 Richard Biener <rguenther@suse.de>
15169 * config/i386/i386.c (print_reg): Fix typo.
15170 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
15172 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
15174 * configure: Regenerate.
15176 2018-02-05 Martin Sebor <msebor@redhat.com>
15178 PR tree-optimization/83369
15179 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
15182 2018-02-05 Martin Liska <mliska@suse.cz>
15184 * doc/invoke.texi: Cherry-pick upstream r323995.
15186 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
15188 * ira.c (ira_init_register_move_cost): Adjust comment.
15190 2018-02-05 Martin Liska <mliska@suse.cz>
15192 PR gcov-profile/84137
15193 * doc/gcov.texi: Fix typo in documentation.
15195 2018-02-05 Martin Liska <mliska@suse.cz>
15197 PR gcov-profile/83879
15198 * doc/gcov.texi: Document necessity of --dynamic-list-data when
15199 using dlopen functionality.
15201 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
15203 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
15204 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
15205 _mm_maskz_range_ss, _mm_mask_range_round_ss,
15206 _mm_maskz_range_round_ss): New intrinsics.
15207 (__builtin_ia32_rangesd128_round)
15208 (__builtin_ia32_rangess128_round): Remove.
15209 (__builtin_ia32_rangesd128_mask_round,
15210 __builtin_ia32_rangess128_mask_round): New builtins.
15211 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
15212 __builtin_ia32_rangess128_round): Remove.
15213 (__builtin_ia32_rangesd128_mask_round,
15214 __builtin_ia32_rangess128_mask_round): New builtins.
15215 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
15216 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
15217 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
15218 "<round_saeonly_constraint>")): Changed to ...
15219 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
15220 "<round_saeonly_scalar_constraint>")): ... this.
15221 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
15222 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
15223 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
15224 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
15225 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
15227 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
15229 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
15231 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
15232 Remove all values except native, 8540 and 8548.
15234 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
15236 * config/i386/i386.c (ix86_output_function_return): Pass
15237 INVALID_REGNUM, instead of -1, as invalid register number to
15238 indirect_thunk_name and output_indirect_thunk.
15240 2018-02-02 Julia Koval <julia.koval@intel.com>
15242 * config.gcc: Add -march=icelake.
15243 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
15244 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
15245 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
15246 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
15247 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
15248 (processor_target_table): Add icelake.
15249 (ix86_option_override_internal): Handle new PTAs.
15250 (get_builtin_code_for_version): Handle icelake.
15251 (M_INTEL_COREI7_ICELAKE): New.
15252 (fold_builtin_cpu): Handle icelake.
15253 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
15254 * doc/invoke.texi: Add -march=icelake.
15256 2018-02-02 Julia Koval <julia.koval@intel.com>
15258 * config/i386/i386.c (ix86_option_override_internal): Change flags type
15259 to wide_int_bitmask.
15260 * wide-int-bitmask.h: New.
15262 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
15265 * config/i386/i386.md: Replace Pmode with word_mode in
15266 builtin_setjmp_setup and builtin_longjmp to support x32.
15268 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
15272 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
15274 (rs6000_supported_cpu_names): New static variable.
15275 (linux_cpu_translation_table): Likewise.
15276 (elf_platform) <cpu>: Define new static variable and use it.
15277 Translate kernel AT_PLATFORM name to canonical name if needed.
15278 Error if platform name is unknown.
15280 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
15283 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
15285 2018-02-01 Jeff Law <law@redhat.com>
15288 * config/i386/i386.c (release_scratch_register_on_entry): Add new
15289 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
15290 the scratch if RELEASE_VIA_POP is false.
15291 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
15292 If we have to save a temporary register, decrement SIZE appropriately.
15293 Pass new arguments to release_scratch_register_on_entry.
15294 (ix86_adjust_stack_and_probe): Likewise.
15295 (ix86_emit_probe_stack_range): Pass new arguments to
15296 release_scratch_register_on_entry.
15298 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
15300 PR rtl-optimization/84157
15301 * combine.c (change_zero_ext): Use REG_P predicate in
15302 front of HARD_REGISTER_P predicate.
15304 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
15306 * config/avr/avr.c (avr_option_override): Move disabling of
15307 -fdelete-null-pointer-checks to...
15308 * common/config/avr/avr-common.c (avr_option_optimization_table):
15311 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15313 PR tree-optimization/81635
15314 * tree-data-ref.c (split_constant_offset_1): For types that
15315 wrap on overflow, try to use range info to prove that wrapping
15318 2018-02-01 Renlin Li <renlin.li@arm.com>
15321 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
15322 TAILCALL_ADDR_REGS.
15323 (aarch64_register_move_cost): Likewise.
15324 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
15325 TAILCALL_ADDR_REGS.
15326 (REG_CLASS_NAMES): Likewise.
15327 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
15328 TAILCALL_ADDR_REGS. Remove IP registers.
15329 * config/aarch64/aarch64.md (Ucs): Update register constraint.
15331 2018-02-01 Richard Biener <rguenther@suse.de>
15333 * domwalk.h (dom_walker::dom_walker): Add additional constructor
15334 for specifying RPO order and allow NULL for that.
15335 * domwalk.c (dom_walker::dom_walker): Likewise.
15336 (dom_walker::walk): Handle NULL RPO order.
15337 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
15339 (rewrite_update_dom_walker): Likewise.
15340 (mark_def_dom_walker): Likewise.
15342 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15344 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
15345 (aarch64_maybe_expand_sve_subreg_move): Declare.
15346 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
15347 * config/aarch64/predicates.md (aarch64_any_register_operand): New
15349 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
15350 that are semantically a reverse operation.
15351 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
15352 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
15353 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
15355 (aarch64_can_change_mode_class): For big-endian, forbid changes
15356 between two SVE modes if they have different element sizes.
15358 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15360 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
15361 the TImode handling for big-endian targets.
15363 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15365 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
15366 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
15368 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
15369 Remove BSWAP handing for big-endian targets and use the form of
15370 LD1RQ appropariate for the mode.
15372 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15374 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
15375 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
15376 duplicated element.
15378 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15381 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
15382 check for operands that need to go through aarch64_sve_reload_be.
15384 2018-02-01 Jakub Jelinek <jakub@redhat.com>
15386 PR tree-optimization/81661
15387 PR tree-optimization/84117
15388 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
15389 * tree-eh.c: Include gimplify.h.
15390 (find_trapping_overflow, replace_trapping_overflow,
15391 rewrite_to_non_trapping_overflow): New functions.
15392 * tree-vect-loop.c: Include tree-eh.h.
15393 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
15394 * tree-data-ref.c: Include tree-eh.h.
15395 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
15397 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
15399 PR rtl-optimization/84123
15400 * combine.c (change_zero_ext): Check if hard register satisfies
15401 can_change_dest_mode before calling gen_lowpart_SUBREG.
15403 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
15406 * ira.c (ira_init_register_move_cost): Remove assert.
15408 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
15410 PR rtl-optimization/84071
15411 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
15412 * doc/tm.texi: Regenerate.
15414 2018-01-31 Richard Biener <rguenther@suse.de>
15416 PR tree-optimization/84132
15417 * tree-data-ref.c (analyze_miv_subscript): Properly
15418 check whether evolution_function_is_affine_multivariate_p
15419 before calling gcd_of_steps_may_divide_p.
15421 2018-01-31 Julia Koval <julia.koval@intel.com>
15424 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
15425 * config/i386/i386.md (rdpid_rex64) New.
15426 (rdpid): Make 32bit only.
15428 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
15431 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
15432 an IDENTIFIER_NODE for FUNCTION_TYPE's.
15434 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
15437 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15439 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
15441 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
15443 PR rtl-optimization/84071
15444 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
15445 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
15447 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
15449 * config/arc/arc.c (arc_handle_aux_attribute): New function.
15450 (arc_attribute_table): Add 'aux' attribute.
15451 (arc_in_small_data_p): Consider aux like variables.
15452 (arc_is_aux_reg_p): New function.
15453 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
15454 (arc_get_aux_arg): New function.
15455 (prepare_move_operands): Handle aux-register access.
15456 (arc_handle_aux_attribute): New function.
15457 * doc/extend.texi (ARC Variable attributes): Add subsection.
15459 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
15461 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
15462 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
15463 (arc_attribute_table): Add 'uncached' attribute.
15464 (arc_print_operand): Print '.di' flag for uncached memory
15466 (arc_in_small_data_p): Do not consider for small data the uncached
15468 (arc_is_uncached_mem_p): New function.
15469 * config/arc/predicates.md (compact_store_memory_operand): Check
15470 for uncached memory accesses.
15471 (nonvol_nonimm_operand): Likewise.
15472 * doc/extend.texi (ARC Type Attribute): New subsection.
15474 2018-01-31 Jakub Jelinek <jakub@redhat.com>
15477 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
15478 falign-loops=): Add Optimization flag.
15480 2018-01-30 Jeff Law <law@redhat.com>
15483 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
15484 INT_REGISTERS_SAVED. Check it prior to calling
15485 get_scratch_register_on_entry.
15486 (ix86_adjust_stack_and_probe): Similarly.
15487 (ix86_emit_probe_stack_range): Similarly.
15488 (ix86_expand_prologue): Corresponding changes.
15490 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15493 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
15494 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
15496 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
15499 * lra-constraints.c (curr_insn_transform): Process AND in the
15502 2018-01-30 Jakub Jelinek <jakub@redhat.com>
15504 PR rtl-optimization/83986
15505 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
15506 dependence against last_pending_memory_flush in addition to
15507 pending_jump_insns.
15509 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
15511 PR tree-optimization/81611
15512 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
15515 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15518 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
15521 2018-01-30 Richard Biener <rguenther@suse.de>
15522 Jakub Jelinek <jakub@redhat.com>
15524 PR tree-optimization/84111
15525 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
15526 inner loops added during recursion, as they don't have up-to-date
15529 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
15532 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
15533 (can_inline_edge_by_limits_p): ... here.
15534 (can_early_inline_edge_p, check_callers,
15535 update_caller_keys, update_callee_keys, recursive_inlining,
15536 add_new_edges_to_heap, speculation_useful_p,
15537 inline_small_functions,
15538 inline_small_functions, flatten_function,
15539 inline_to_all_callers_1): Update.
15541 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
15543 * profile-count.c (profile_count::combine_with_ipa_count): Handle
15546 2018-01-30 Richard Biener <rguenther@suse.de>
15548 PR tree-optimization/83008
15549 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
15550 invariant and constant vector uses in stmts when they need
15551 more than one stmt.
15553 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15556 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
15557 * configure: Regenerate.
15559 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
15561 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
15563 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
15564 Use gen_rtx_REG rather than gen_lowpart.
15566 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
15568 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
15569 rather than 0 when creating partial subregs.
15571 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
15573 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
15576 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
15579 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
15580 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
15581 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
15582 flags. This restores the settings used before the 2017-07-24.
15583 Turning off pre increment/decrement/modify allows IVOPTS to
15584 optimize DF/SF loops where the index is an int.
15586 2018-01-29 Richard Biener <rguenther@suse.de>
15587 Kelvin Nilsen <kelvin@gcc.gnu.org>
15590 * tree-vect-stmts.c (vectorizable_call): Don't call
15591 targetm.vectorize_builtin_md_vectorized_function if callee is
15594 2018-01-22 Carl Love <cel@us.ibm.com>
15596 * doc/extend.tex: Fix typo in second arg in
15597 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
15599 2018-01-29 Richard Biener <rguenther@suse.de>
15601 PR tree-optimization/84086
15602 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
15603 (flush_ssaname_freelist): When SSA names were released reset
15604 the SCEV hash table.
15606 2018-01-29 Richard Biener <rguenther@suse.de>
15608 PR tree-optimization/84057
15609 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
15610 removed paths when removing edges.
15612 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
15614 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
15615 -mfunction-return=@var{choice}.
15617 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
15619 PR diagnostic/84034
15620 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
15621 Handle CR like TAB.
15622 (layout::print_source_line): Likewise.
15623 (test_get_line_width_without_trailing_whitespace): Add test cases.
15625 2018-01-27 Jakub Jelinek <jakub@redhat.com>
15627 PR middle-end/84040
15628 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
15631 2018-01-26 Jim Wilson <jimw@sifive.com>
15633 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
15635 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
15638 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15640 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
15641 and CMP + SUB-immediate -> SUBS.
15643 2018-01-26 Martin Sebor <msebor@redhat.com>
15645 PR tree-optimization/83896
15646 * tree-ssa-strlen.c (get_string_len): Rename...
15647 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
15648 Avoid assuming length is constant.
15649 (handle_char_store): Use HOST_WIDE_INT for string length.
15651 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
15654 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
15655 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
15657 2018-01-26 Richard Biener <rguenther@suse.de>
15659 PR rtl-optimization/84003
15660 * dse.c (record_store): Only record redundant stores when
15661 the earlier store aliases at least all accesses the later one does.
15663 2018-01-26 Jakub Jelinek <jakub@redhat.com>
15665 PR rtl-optimization/83985
15666 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
15667 REG_CFA_RESTORE insns.
15668 (delete_unmarked_insns): Don't ignore separate shrink wrapping
15669 REG_CFA_RESTORE insns here.
15672 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
15673 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
15675 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15677 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
15678 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
15679 (arc_init): Likewise.
15680 (arc_override_options): Likewise.
15681 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
15683 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
15685 * config/arc/arc.h (TARGET_DBNZ): Define.
15686 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
15687 properly set the tune attribute.
15688 (dbnz): Use TARGET_DBNZ guard.
15689 * config/arc/arc.opt (mtune): Add core3 option.
15691 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15693 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
15694 recognize new pic like addresses.
15695 (arc_delegitimize_address): Clean up.
15697 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15699 * config/arc/arc-arches.def: Option mrf16 valid for all
15701 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
15702 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
15703 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
15704 * config/arc/arc-tables.opt: Regenerate.
15705 * config/arc/arc.c (arc_conditional_register_usage): Handle
15706 reduced register file case.
15707 (arc_file_start): Set must have build attributes.
15708 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
15709 mrf16 option value.
15710 * config/arc/arc.opt (mrf16): Add new option.
15711 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
15712 * config/arc/genmultilib.awk: Handle new mrf16 option.
15713 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
15714 * config/arc/t-multilib: Regenerate.
15715 * doc/invoke.texi (ARC Options): Document mrf16 option.
15717 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15719 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
15720 * config/arc/arc.c (arc_handle_secure_attribute): New function.
15721 (arc_attribute_table): Add 'secure_call' attribute.
15722 (arc_print_operand): Print secure call operand.
15723 (arc_function_ok_for_sibcall): Don't optimize tail calls when
15725 (arc_is_secure_call_p): New function. * config/arc/arc.md
15726 (call_i): Add support for sjli instruction.
15727 (call_value_i): Likewise.
15728 * config/arc/constraints.md (Csc): New constraint.
15730 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15731 John Eric Martin <John.Martin@emmicro-us.com>
15733 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
15734 * config/arc/arc.c (_arc_jli_section): New struct.
15735 (arc_jli_section): New type.
15736 (rc_jli_sections): New static variable.
15737 (arc_handle_jli_attribute): New function.
15738 (arc_attribute_table): Add jli_always and jli_fixed attribute.
15739 (arc_file_end): New function.
15740 (TARGET_ASM_FILE_END): Define.
15741 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
15742 (arc_add_jli_section): New function.
15743 (jli_call_scan): Likewise.
15744 (arc_reorg): Call jli_call_scan.
15745 (arc_output_addsi): Remove 'S' from printing asm operand.
15746 (arc_is_jli_call_p): New function.
15747 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
15749 (movhi_insn): Likewise.
15750 (movsi_insn): Likewise.
15751 (movsi_set_cc_insn): Likewise.
15752 (loadqi_update): Likewise.
15753 (load_zeroextendqisi_update): Likewise.
15754 (load_signextendqisi_update): Likewise.
15755 (loadhi_update): Likewise.
15756 (load_zeroextendhisi_update): Likewise.
15757 (load_signextendhisi_update): Likewise.
15758 (loadsi_update): Likewise.
15759 (loadsf_update): Likewise.
15760 (movsicc_insn): Likewise.
15761 (bset_insn): Likewise.
15762 (bxor_insn): Likewise.
15763 (bclr_insn): Likewise.
15764 (bmsk_insn): Likewise.
15765 (bicsi3_insn): Likewise.
15766 (cmpsi_cc_c_insn): Likewise.
15767 (movsi_ne): Likewise.
15768 (movsi_cond_exec): Likewise.
15769 (clrsbsi2): Likewise.
15770 (norm_f): Likewise.
15780 (call_i): Remove 'S' asm letter, add jli instruction.
15781 (call_value_i): Likewise.
15782 * config/arc/arc.op (mjli-always): New option.
15783 * config/arc/constraints.md (Cji): New constraint.
15784 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
15786 (subsf3_fpx): Likewise.
15787 (mulsf3_fpx): Likewise.
15788 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
15790 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
15791 function attrbutes.
15792 * doc/invoke.texi (ARC): Document mjli-always option.
15794 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
15796 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
15797 avoid addition with 0 and use incw and decw where possible.
15799 2018-01-26 Richard Biener <rguenther@suse.de>
15801 PR tree-optimization/81082
15802 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
15803 association if it requires casting to unsigned.
15804 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
15805 from fold_plusminus_mult_expr to catch important cases late when
15806 range info is available.
15808 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15810 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
15811 * configure.ac (hidden_linkonce): New test.
15812 * configure: Regenerate.
15813 * config.in: Regenerate.
15815 2018-01-26 Julia Koval <julia.koval@intel.com>
15817 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
15818 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
15819 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
15820 _mm_mask_bitshuffle_epi64_mask): Fix type.
15821 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
15822 USI_FTYPE_V4DI_V4DI_USI): Remove.
15823 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
15824 __builtin_ia32_vpshufbitqmb256_mask,
15825 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
15826 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
15827 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
15829 2018-01-26 Alan Modra <amodra@gmail.com>
15832 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
15833 UNSPEC_VBPERMQ. Sort other unspecs.
15835 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
15837 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
15839 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
15841 PR middle-end/83055
15842 * predict.c (drop_profile): Do not push/pop cfun; update also
15844 (handle_missing_profiles): Fix logic looking for zero profiles.
15846 2018-01-25 Jakub Jelinek <jakub@redhat.com>
15848 PR middle-end/83977
15849 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
15850 on functions with #pragma omp declare simd or functions with simd
15852 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
15853 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
15854 Remove trailing \n from warning_at calls.
15856 2018-01-25 Tom de Vries <tom@codesourcery.com>
15859 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
15860 for neutered workers.
15862 2018-01-24 Joseph Myers <joseph@codesourcery.com>
15865 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
15866 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
15868 2018-01-24 Jeff Law <law@redhat.com>
15871 * i386.c (get_probe_interval): Move to earlier point.
15872 (ix86_compute_frame_layout): If -fstack-clash-protection and
15873 the frame is larger than the probe interval, then use pushes
15874 to save registers rather than reg->mem moves.
15875 (ix86_expand_prologue): Remove conditional for int_registers_saved
15878 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
15881 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
15882 min/max for never referenced object.
15884 2018-01-24 Jakub Jelinek <jakub@redhat.com>
15886 PR middle-end/83977
15887 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
15889 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
15890 attributes from DECL_ATTRIBUTES (decl) without affecting
15891 DECL_ATTRIBUTES (current_function_decl).
15892 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
15893 functions with non-NULL DECL_ABSTRACT_ORIGIN.
15895 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
15897 PR tree-optimization/83979
15898 * fold-const.c (fold_comparison): Use constant_boolean_node
15899 instead of boolean_{true,false}_node.
15901 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
15903 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
15906 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15908 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
15909 Simplify the clause that sets the length attribute.
15910 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
15911 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
15912 clause that sets the length attribute.
15913 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
15915 2018-01-24 Tom de Vries <tom@codesourcery.com>
15918 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
15919 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
15920 Add strict parameter.
15921 (prevent_branch_around_nothing): Insert dummy insn between branch to
15922 label and label with no ptx insn inbetween.
15923 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
15925 2018-01-24 Tom de Vries <tom@codesourcery.com>
15928 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
15929 for neutered threads in warp.
15930 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
15932 2018-01-24 Richard Biener <rguenther@suse.de>
15934 PR tree-optimization/83176
15935 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
15938 2018-01-24 Richard Biener <rguenther@suse.de>
15940 PR tree-optimization/82819
15941 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
15942 code generating pluses that are no-ops in the target precision.
15944 2018-01-24 Richard Biener <rguenther@suse.de>
15946 PR middle-end/84000
15947 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
15949 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
15951 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
15952 to merge probabilities.
15953 * predict.c (probably_never_executed): Also mark as cold functions
15954 with global 0 profile and guessed local profile.
15955 * profile-count.c (profile_probability::combine_with_count): New
15957 * profile-count.h (profile_probability::operator*,
15958 profile_probability::operator*=, profile_probability::operator/,
15959 profile_probability::operator/=): Reduce precision to adjusted
15960 and set value to guessed on contradictory divisions.
15961 (profile_probability::combine_with_freq): Remove.
15962 (profile_probability::combine_wiht_count): Declare.
15963 (profile_count::force_nonzero):: Set to adjusted.
15964 (profile_count::probability_in):: Set quality to adjusted.
15965 * tree-ssa-tail-merge.c (replace_block_by): Use
15966 combine_with_count.
15968 2018-01-23 Andrew Waterman <andrew@sifive.com>
15969 Jim Wilson <jimw@sifive.com>
15971 * config/riscv/riscv.c (riscv_stack_boundary): New.
15972 (riscv_option_override): Set riscv_stack_boundary. Handle
15973 riscv_preferred_stack_boundary_arg.
15974 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
15975 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
15976 (STACK_BOUNDARY): Set to riscv_stack_boundary.
15977 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
15978 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
15979 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
15981 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
15984 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
15985 of struct ix86_frame.
15986 (ix86_expand_epilogue): Likewise. Add a local variable for
15987 the reg_save_offset field in struct ix86_frame.
15989 2018-01-23 Bin Cheng <bin.cheng@arm.com>
15991 PR tree-optimization/82604
15992 * tree-loop-distribution.c (enum partition_kind): New enum item
15993 PKIND_PARTIAL_MEMSET.
15994 (partition_builtin_p): Support above new enum item.
15995 (generate_code_for_partition): Ditto.
15996 (compute_access_range): Differentiate cases that equality can be
15997 proven at all loops, the innermost loops or no loops.
15998 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
15999 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
16000 (finalize_partitions, distribute_loop): Don't fuse partition of
16001 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
16002 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
16003 parloop is enabled.
16005 2018-01-23 Martin Liska <mliska@suse.cz>
16007 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
16008 order to ignore the predictor.
16009 (PRED_POLYMORPHIC_CALL): Likewise.
16010 (PRED_RECURSIVE_CALL): Likewise.
16012 2018-01-23 Martin Liska <mliska@suse.cz>
16014 * tree-profile.c (tree_profiling): Print function header to
16015 aware reader which function we are working on.
16016 * value-prof.c (gimple_find_values_to_profile): Do not print
16017 not interesting value histograms.
16019 2018-01-23 Martin Liska <mliska@suse.cz>
16021 * profile-count.h (enum profile_quality): Add
16022 profile_uninitialized as the first value. Do not number values
16023 as they are zero based.
16024 (profile_count::verify): Update sanity check.
16025 (profile_probability::verify): Likewise.
16027 2018-01-23 Nathan Sidwell <nathan@acm.org>
16029 * doc/invoke.texi (ffor-scope): Deprecate.
16031 2018-01-23 David Malcolm <dmalcolm@redhat.com>
16033 PR tree-optimization/83510
16034 * domwalk.c (set_all_edges_as_executable): New function.
16035 (dom_walker::dom_walker): Convert bool param
16036 "skip_unreachable_blocks" to enum reachability. Move setup of
16037 edge flags to set_all_edges_as_executable and only do it when
16038 reachability is REACHABLE_BLOCKS.
16039 * domwalk.h (enum dom_walker::reachability): New enum.
16040 (dom_walker::dom_walker): Convert bool param
16041 "skip_unreachable_blocks" to enum reachability.
16042 (set_all_edges_as_executable): New decl.
16043 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
16044 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
16046 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
16047 but converting true to REACHABLE_BLOCKS.
16048 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
16050 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
16051 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
16052 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
16054 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
16055 if check_all_array_refs will be called.
16057 2018-01-23 David Malcolm <dmalcolm@redhat.com>
16059 * tree.c (selftest::test_location_wrappers): Add more test
16062 2018-01-23 David Malcolm <dmalcolm@redhat.com>
16064 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
16065 (selftest::test_bit_in_range): Likewise.
16067 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
16070 * doc/sourcebuild.texi (vect_float): Say that the selector
16071 only describes the situation when -funsafe-math-optimizations is on.
16072 (vect_float_strict): Document.
16074 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
16076 PR tree-optimization/83965
16077 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
16078 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
16079 instead of checking only for a reduction.
16080 (vect_recog_widen_sum_pattern): Likewise.
16082 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
16084 * predict.c (probably_never_executed): Only use precise profile info.
16085 (compute_function_frequency): Skip after inlining hack since we now
16086 have quality checking.
16088 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
16090 * profile-count.h (profile_probability::very_unlikely,
16091 profile_probability::unlikely, profile_probability::even): Set
16092 precision to guessed.
16094 2018-01-23 Richard Biener <rguenther@suse.de>
16096 PR tree-optimization/83963
16097 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
16098 Properly terminate dominator walk when crossing the exit edge not
16099 when visiting its source block.
16101 2018-01-23 Jakub Jelinek <jakub@redhat.com>
16104 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
16105 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
16107 2018-01-22 Jakub Jelinek <jakub@redhat.com>
16109 PR tree-optimization/83957
16110 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
16111 semicolon after for body surrounded by braces.
16113 PR tree-optimization/83081
16114 * profile-count.h (profile_probability::split): New method.
16115 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
16116 Use profile_probability::split.
16117 (do_compare_rtx_and_jump): Fix adjustment of probabilities
16118 when splitting a single conditional jump into 2.
16120 2018-01-22 David Malcolm <dmalcolm@redhat.com>
16122 PR tree-optimization/69452
16123 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
16126 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16128 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
16129 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
16130 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
16132 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16134 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
16136 * config/rl78/rl78.md (movdi): New define_expand.
16137 * config/rl78/rl78.c (rl78_split_movdi): New function.
16139 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
16142 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
16144 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
16145 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
16146 128-bit to produce an UNSPEC move to get the double word with the
16147 signbit and then a shift directly to do signbit.
16148 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
16149 implementation with a new version that just does either a direct
16150 move or a regular move. Move memory interface to separate insns.
16151 Move insns so they are next to the expander.
16152 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
16153 with signbit move. Split big and little endian case.
16154 (signbit<mode>2_dm_mem_le): Likewise.
16155 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
16156 (signbit<mode>2_dm2): Likewise.
16158 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16160 * config/rl78/rl78.md (anddi3): New define_expand.
16162 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16164 * config/rl78/rl78.md (umindi3): New define_expand.
16166 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16168 * config/rl78/rl78.md (smindi3): New define_expand.
16170 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16172 * config/rl78/rl78.md (smaxdi3): New define_expand.
16174 2018-01-22 Carl Love <cel@us.ibm.com>
16176 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
16177 LVX_V1TI): Add macro expansion.
16178 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
16179 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
16180 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
16181 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
16182 Change check to determine if the instruction is a byte reversing
16183 entry. Fix typo in comment.
16184 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
16185 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
16186 Add def_builtin calls for new builtins.
16187 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
16188 Add define_insn expansion.
16190 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16192 * config/rl78/rl78.md (umaxdi3): New define_expand.
16194 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16196 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
16197 for non-QImode registers.
16199 2018-01-22 Richard Biener <rguenther@suse.de>
16201 PR tree-optimization/83963
16202 * graphite-scop-detection.c (scop_detection::get_sese): Delay
16203 including the loop exit block.
16204 (scop_detection::merge_sese): Likewise.
16205 (scop_detection::add_scop): Do it here instead.
16207 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16209 * doc/sourcebuild.texi (arm_softfloat): Document.
16211 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
16214 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
16215 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
16216 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
16218 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16219 David Edelsohn <dje.gcc@gmail.com>
16222 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
16223 Change "crset eq" to "crset 2".
16224 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
16225 (*call_indirect_aix<mode>_nospec): Likewise.
16226 (*call_value_indirect_aix<mode>_nospec): Likewise.
16227 (*call_indirect_elfv2<mode>_nospec): Likewise.
16228 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
16229 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
16230 change assembly output from . to $.
16231 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
16232 (indirect_jump<mode>_nospec): Change assembly output from . to $.
16233 (*tablejump<mode>_internal1_nospec): Likewise.
16235 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
16238 * config/sh/sh_optimize_sett_clrt.cc:
16239 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
16241 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
16243 PR tree-optimization/83940
16244 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
16245 offset_dt to vect_constant_def rather than vect_unknown_def_type.
16246 (vect_check_load_store_mask): Add a mask_dt_out parameter and
16247 use it to pass back the definition type.
16248 (vect_check_store_rhs): Likewise rhs_dt_out.
16249 (vect_build_gather_load_calls): Add a mask_dt argument and use
16250 it instead of a call to vect_is_simple_use.
16251 (vectorizable_store): Update calls to vect_check_load_store_mask
16252 and vect_check_store_rhs. Use the dt returned by the latter instead
16253 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
16254 instead of calls to vect_is_simple_use. Pass the scalar rather
16255 than the vector operand to vect_is_simple_use when handling
16256 second and subsequent copies of an rhs value.
16257 (vectorizable_load): Update calls to vect_check_load_store_mask
16258 and vect_build_gather_load_calls. Use the cached mask_dt and
16259 gs_info.offset_dt instead of calls to vect_is_simple_use.
16261 2018-01-20 Jakub Jelinek <jakub@redhat.com>
16263 PR middle-end/83945
16264 * tree-emutls.c: Include gimplify.h.
16265 (lower_emutls_2): New function.
16266 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
16267 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
16268 it before further processing.
16271 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
16272 UINTVAL (trueop1) instead of INTVAL (op1).
16274 2018-01-19 Jakub Jelinek <jakub@redhat.com>
16278 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
16279 INCOMING_FRAME_SP_OFFSET if not defined.
16280 (scan_trace): Add ENTRY argument. If true and
16281 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
16282 emit a note to adjust the CFA offset.
16283 (create_cfi_notes): Adjust scan_trace callers.
16284 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
16285 INCOMING_FRAME_SP_OFFSET in the CIE.
16286 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
16287 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
16289 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
16290 * doc/tm.texi: Regenerated.
16292 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
16294 PR rtl-optimization/83147
16295 * lra-constraints.c (remove_inheritance_pseudos): Use
16296 lra_substitute_pseudo_within_insn.
16298 2018-01-19 Tom de Vries <tom@codesourcery.com>
16299 Cesar Philippidis <cesar@codesourcery.com>
16302 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
16304 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
16307 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
16308 spaces for function labels.
16310 2018-01-19 Martin Liska <mliska@suse.cz>
16312 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
16313 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
16314 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
16315 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
16316 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
16317 (PRED_CONST_RETURN): Change from 69 to 65.
16318 (PRED_NULL_RETURN): Change from 91 to 71.
16319 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
16320 (PRED_LOOP_GUARD): Change from 66 to 73.
16322 2018-01-19 Martin Liska <mliska@suse.cz>
16324 * predict.c (predict_insn_def): Add new assert.
16325 (struct branch_predictor): Change type to signed integer.
16326 (test_prediction_value_range): Amend test to cover
16327 PROB_UNINITIALIZED.
16328 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
16329 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
16330 (PRED_LOOP_ITERATIONS_MAX): Likewise.
16331 (PRED_LOOP_IV_COMPARE): Likewise.
16332 * predict.h (PROB_UNINITIALIZED): Define new constant.
16334 2018-01-19 Martin Liska <mliska@suse.cz>
16336 * predict.c (dump_prediction): Add new format for
16337 analyze_brprob.py script which is enabled with -details
16339 * profile-count.h (precise_p): New function.
16341 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
16343 PR tree-optimization/83922
16344 * tree-vect-loop.c (vect_verify_full_masking): Return false if
16345 there are no statements that need masking.
16346 (vect_active_double_reduction_p): New function.
16347 (vect_analyze_loop_operations): Use it when handling phis that
16348 are not in the loop header.
16350 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
16352 PR tree-optimization/83914
16353 * tree-vect-loop.c (vectorizable_induction): Don't convert
16354 init_expr or apply the peeling adjustment for inductions
16355 that are nested within the vectorized loop.
16357 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16359 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
16362 2018-01-18 Jakub Jelinek <jakub@redhat.com>
16366 * function.h (gimplify_parameters): Add gimple_seq * argument.
16367 * function.c: Include gimple.h and options.h.
16368 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
16369 for the added local temporaries if needed.
16370 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
16371 if there are any parameter cleanups, wrap whole body into a
16372 try/finally with the cleanups.
16374 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
16377 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
16378 Use GET_MODE_CLASS for scalar floating point.
16380 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
16384 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
16385 Fix call of call_cgraph_insertion_hooks.
16387 2018-01-18 Martin Sebor <msebor@redhat.com>
16389 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
16391 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
16394 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
16397 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
16400 * common.opt: (-ffile-prefix-map): New option.
16401 * opts.c (common_handle_option): Defer it.
16402 * opts-global.c (handle_common_deferred_options): Handle it.
16403 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
16404 * file-prefix-map.h: New file.
16405 (remap_debug_filename, add_debug_prefix_map): ...here.
16406 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
16407 * final.c (debug_prefix_map, add_debug_prefix_map
16408 remap_debug_filename): Move to...
16409 * file-prefix-map.c: New file.
16410 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
16411 generalize, get rid of alloca(), use strrchr() instead of strchr().
16412 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
16413 Implement in terms of add_prefix_map().
16414 (remap_macro_filename, remap_debug_filename): Implement in term of
16416 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
16417 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
16418 * dbxout.c: Include file-prefix-map.h.
16419 * varasm.c: Likewise.
16420 * vmsdbgout.c: Likewise.
16421 * xcoffout.c: Likewise.
16422 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
16423 * doc/cppopts.texi (-fmacro-prefix-map): Document.
16424 * doc/invoke.texi (-ffile-prefix-map): Document.
16425 (-fdebug-prefix-map): Update description.
16427 2018-01-18 Martin Liska <mliska@suse.cz>
16429 * config/i386/i386.c (indirect_thunk_name): Document that also
16431 (output_indirect_thunk): Document why both instructions
16432 (pause and lfence) are generated.
16434 2018-01-18 Richard Biener <rguenther@suse.de>
16436 PR tree-optimization/83887
16437 * graphite-scop-detection.c
16438 (scop_detection::get_nearest_dom_with_single_entry): Remove.
16439 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
16440 (scop_detection::merge_sese): Re-implement with a flood-fill
16441 algorithm that properly finds a SESE region if it exists.
16443 2018-01-18 Jakub Jelinek <jakub@redhat.com>
16446 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
16447 pointer_diff optimizations use view_convert instead of convert.
16449 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16451 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
16452 Generate different code for -mno-speculate-indirect-jumps.
16453 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
16454 (*call_indirect_aix<mode>): Disable for
16455 -mno-speculate-indirect-jumps.
16456 (*call_indirect_aix<mode>_nospec): New define_insn.
16457 (*call_value_indirect_aix<mode>): Disable for
16458 -mno-speculate-indirect-jumps.
16459 (*call_value_indirect_aix<mode>_nospec): New define_insn.
16460 (*sibcall_nonlocal_sysv<mode>): Generate different code for
16461 -mno-speculate-indirect-jumps.
16462 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
16464 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
16466 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
16467 long double type, set the flags for noting the default long double
16468 type, even if we don't pass or return a long double type.
16470 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
16473 * ipa-inline.c (flatten_function): Do not overwrite final inlining
16476 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
16478 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
16479 support for merge[hl].
16480 (fold_mergehl_helper): New helper function.
16481 (tree-vector-builder.h): New #include for tree_vector_builder usage.
16482 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
16483 (altivec_vmrglw_direct): Add xxmrglw insn.
16485 2018-01-17 Andrew Waterman <andrew@sifive.com>
16487 * config/riscv/riscv.c (riscv_conditional_register_usage): If
16488 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
16490 2018-01-17 David Malcolm <dmalcolm@redhat.com>
16493 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
16494 call the lto_location_cache before reading the
16495 DECL_SOURCE_LOCATION of the types.
16497 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
16498 Richard Sandiford <richard.sandiford@linaro.org>
16500 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
16501 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
16502 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
16503 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
16504 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
16506 * config/aarch64/constraints.md (aarch64_movti_operand):
16508 * config/aarch64/predicates.md (Uti): Add new constraint.
16510 2018-01-17 Carl Love <cel@us.ibm.com>
16512 * config/rs6000/vsx.md (define_expand xl_len_r,
16513 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
16514 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
16516 (define_expand, define_insn): Move the shift left from the
16517 define_insn to the define_expand for lxvl and stxvl instructions.
16518 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
16519 and XL_LEN_R definitions to PURE.
16521 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
16523 * config/i386/i386.c (indirect_thunk_name): Declare regno
16524 as unsigned int. Compare regno with INVALID_REGNUM.
16525 (output_indirect_thunk): Ditto.
16526 (output_indirect_thunk_function): Ditto.
16527 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
16528 in the call to output_indirect_thunk_function.
16530 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
16532 PR middle-end/83884
16533 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
16534 rather than the size of inner_type to determine the stack slot size
16535 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
16537 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
16540 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
16543 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
16545 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
16546 endian Linux systems to optionally enable multilibs for selecting
16547 the long double type if the user configured an explicit type.
16548 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
16549 have no long double multilibs if not defined.
16550 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
16551 warn if the user used -mabi={ieee,ibm}longdouble and we built
16552 multilibs for long double.
16553 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
16554 appropriate multilib option.
16555 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
16557 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
16558 for building long double multilibs.
16559 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
16561 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
16563 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
16566 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
16568 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
16571 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
16574 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
16577 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
16579 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
16580 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
16582 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
16584 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
16585 different rtl trees depending on TARGET_64BIT.
16586 (rs6000_gen_lvx): Likewise.
16588 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
16590 * config/visium/visium.md (nop): Tweak comment.
16591 (hazard_nop): Likewise.
16593 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16595 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
16596 -mspeculate-indirect-jumps.
16597 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
16598 for -mno-speculate-indirect-jumps.
16599 (*call_indirect_elfv2<mode>_nospec): New define_insn.
16600 (*call_value_indirect_elfv2<mode>): Disable for
16601 -mno-speculate-indirect-jumps.
16602 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
16603 (indirect_jump): Emit different RTL for
16604 -mno-speculate-indirect-jumps.
16605 (*indirect_jump<mode>): Disable for
16606 -mno-speculate-indirect-jumps.
16607 (*indirect_jump<mode>_nospec): New define_insn.
16608 (tablejump): Emit different RTL for
16609 -mno-speculate-indirect-jumps.
16610 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
16611 (tablejumpsi_nospec): New define_expand.
16612 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
16613 (tablejumpdi_nospec): New define_expand.
16614 (*tablejump<mode>_internal1): Disable for
16615 -mno-speculate-indirect-jumps.
16616 (*tablejump<mode>_internal1_nospec): New define_insn.
16617 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
16620 2018-01-16 Artyom Skrobov tyomitch@gmail.com
16622 * caller-save.c (insert_save): Drop unnecessary parameter. All
16625 2018-01-16 Jakub Jelinek <jakub@redhat.com>
16626 Richard Biener <rguenth@suse.de>
16629 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
16630 return early, inline manually is_gimple_sizepos. Make sure if we
16631 call gimplify_expr we don't end up with a gimple constant.
16632 * tree.c (variably_modified_type_p): Don't return true for
16633 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
16634 * gimplify.h (is_gimple_sizepos): Remove.
16636 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16638 PR tree-optimization/83857
16639 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
16640 vectorizable_live_operation for pure SLP statements.
16641 (vectorizable_live_operation): Handle PHIs.
16643 2018-01-16 Richard Biener <rguenther@suse.de>
16645 PR tree-optimization/83867
16646 * tree-vect-stmts.c (vect_transform_stmt): Precompute
16647 nested_in_vect_loop_p since the scalar stmt may get invalidated.
16649 2018-01-16 Jakub Jelinek <jakub@redhat.com>
16652 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
16653 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
16654 If off is not INTEGER_CST, issue a may not be aligned warning
16655 rather than isn't aligned. Use isn%'t rather than isn't.
16656 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
16658 <case MULT_EXPR>: Improve the case when bottom and one of the
16659 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
16660 operand, in that case check if the other operand is multiple of
16661 bottom divided by the INTEGER_CST operand.
16663 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16666 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
16667 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
16668 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
16669 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
16670 * config/pa/pa.c (pa_function_arg_advance): Likewise.
16671 (pa_function_arg, pa_arg_partial_bytes): Likewise.
16672 (pa_function_arg_size): New function.
16674 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16676 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
16677 in a separate statement.
16679 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16681 PR tree-optimization/83847
16682 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
16683 group gathers and scatters.
16685 2018-01-16 Jakub Jelinek <jakub@redhat.com>
16687 PR rtl-optimization/86620
16688 * params.def (max-sched-ready-insns): Bump minimum value to 1.
16690 PR rtl-optimization/83213
16691 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
16692 to last if both are JUMP_INSNs.
16694 PR tree-optimization/83843
16695 * gimple-ssa-store-merging.c
16696 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
16697 store_immediate_info for bswap/nop orig_stores.
16699 2018-01-15 Andrew Waterman <andrew@sifive.com>
16701 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
16703 <UDIV>: Increase cost if !TARGET_DIV.
16705 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
16707 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
16708 (define_attr "cr_logical_3op"): New.
16709 (cceq_ior_compare): Adjust.
16710 (cceq_ior_compare_complement): Adjust.
16711 (*cceq_rev_compare): Adjust.
16712 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
16713 (is_cracked_insn): Adjust.
16714 (insn_must_be_first_in_group): Adjust.
16715 * config/rs6000/40x.md: Adjust.
16716 * config/rs6000/440.md: Adjust.
16717 * config/rs6000/476.md: Adjust.
16718 * config/rs6000/601.md: Adjust.
16719 * config/rs6000/603.md: Adjust.
16720 * config/rs6000/6xx.md: Adjust.
16721 * config/rs6000/7450.md: Adjust.
16722 * config/rs6000/7xx.md: Adjust.
16723 * config/rs6000/8540.md: Adjust.
16724 * config/rs6000/cell.md: Adjust.
16725 * config/rs6000/e300c2c3.md: Adjust.
16726 * config/rs6000/e500mc.md: Adjust.
16727 * config/rs6000/e500mc64.md: Adjust.
16728 * config/rs6000/e5500.md: Adjust.
16729 * config/rs6000/e6500.md: Adjust.
16730 * config/rs6000/mpc.md: Adjust.
16731 * config/rs6000/power4.md: Adjust.
16732 * config/rs6000/power5.md: Adjust.
16733 * config/rs6000/power6.md: Adjust.
16734 * config/rs6000/power7.md: Adjust.
16735 * config/rs6000/power8.md: Adjust.
16736 * config/rs6000/power9.md: Adjust.
16737 * config/rs6000/rs64.md: Adjust.
16738 * config/rs6000/titan.md: Adjust.
16740 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16742 * config/i386/predicates.md (indirect_branch_operand): Rewrite
16743 ix86_indirect_branch_register logic.
16745 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16747 * config/i386/constraints.md (Bs): Update
16748 ix86_indirect_branch_register check. Don't check
16749 ix86_indirect_branch_register with GOT_memory_operand.
16751 * config/i386/predicates.md (GOT_memory_operand): Don't check
16752 ix86_indirect_branch_register here.
16753 (GOT32_symbol_operand): Likewise.
16755 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16757 * config/i386/predicates.md (constant_call_address_operand):
16758 Rewrite ix86_indirect_branch_register logic.
16759 (sibcall_insn_operand): Likewise.
16761 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16763 * config/i386/constraints.md (Bs): Replace
16764 ix86_indirect_branch_thunk_register with
16765 ix86_indirect_branch_register.
16767 * config/i386/i386.md (indirect_jump): Likewise.
16768 (tablejump): Likewise.
16769 (*sibcall_memory): Likewise.
16770 (*sibcall_value_memory): Likewise.
16771 Peepholes of indirect call and jump via memory: Likewise.
16772 * config/i386/i386.opt: Likewise.
16773 * config/i386/predicates.md (indirect_branch_operand): Likewise.
16774 (GOT_memory_operand): Likewise.
16775 (call_insn_operand): Likewise.
16776 (sibcall_insn_operand): Likewise.
16777 (GOT32_symbol_operand): Likewise.
16779 2018-01-15 Jakub Jelinek <jakub@redhat.com>
16781 PR middle-end/83837
16782 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
16783 type rather than type addr's type points to.
16784 (expand_omp_atomic_mutex): Likewise.
16785 (expand_omp_atomic): Likewise.
16787 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16790 * config/i386/i386.c (output_indirect_thunk_function): Use
16791 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
16792 for __x86_return_thunk.
16794 2018-01-15 Richard Biener <rguenther@suse.de>
16796 PR middle-end/83850
16797 * expmed.c (extract_bit_field_1): Fix typo.
16799 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16802 * config/arm/iterators.md (VF): New mode iterator.
16803 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
16804 Remove integer-related logic from pattern.
16805 (neon_vabd<mode>_3): Likewise.
16807 2018-01-15 Jakub Jelinek <jakub@redhat.com>
16809 PR middle-end/82694
16810 * common.opt (fstrict-overflow): No longer an alias.
16811 (fwrapv-pointer): New option.
16812 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
16813 also for pointer types based on flag_wrapv_pointer.
16814 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
16815 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
16816 opts->x_flag_wrapv got set.
16817 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
16818 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
16819 POINTER_TYPE_OVERFLOW_UNDEFINED.
16820 * match.pd: Likewise in address comparison pattern.
16821 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
16823 2018-01-15 Richard Biener <rguenther@suse.de>
16826 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
16827 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
16828 Reset type names to their identifier if their TYPE_DECL doesn't
16829 have linkage (and thus is used for ODR and devirt).
16830 (save_debug_info_for_decl): Remove.
16831 (save_debug_info_for_type): Likewise.
16832 (add_tree_to_fld_list): Adjust.
16833 * tree-pretty-print.c (dump_generic_node): Make dumping of
16834 type names more robust.
16836 2018-01-15 Richard Biener <rguenther@suse.de>
16838 * BASE-VER: Bump to 8.0.1.
16840 2018-01-14 Martin Sebor <msebor@redhat.com>
16843 * builtins.c (check_access): Avoid warning when the no-warning bit
16846 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
16848 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
16849 * ira-color (allocno_hard_regs_compare): Likewise.
16851 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
16854 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
16855 Use .pushsection/.popsection.
16857 2018-01-14 Martin Sebor <msebor@redhat.com>
16860 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
16862 2018-01-14 Jakub Jelinek <jakub@redhat.com>
16864 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
16865 entry from extra_headers.
16866 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
16867 extra_headers, make the list bitwise identical to the i?86-*-* one.
16869 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16871 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
16872 -mcmodel=large with -mindirect-branch=thunk,
16873 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
16874 -mfunction-return=thunk-extern.
16875 * doc/invoke.texi: Document -mcmodel=large is incompatible with
16876 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
16877 -mfunction-return=thunk and -mfunction-return=thunk-extern.
16879 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16881 * config/i386/i386.c (print_reg): Print the name of the full
16882 integer register without '%'.
16883 (ix86_print_operand): Handle 'V'.
16884 * doc/extend.texi: Document 'V' modifier.
16886 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16888 * config/i386/constraints.md (Bs): Disallow memory operand for
16889 -mindirect-branch-register.
16891 * config/i386/predicates.md (indirect_branch_operand): Likewise.
16892 (GOT_memory_operand): Likewise.
16893 (call_insn_operand): Likewise.
16894 (sibcall_insn_operand): Likewise.
16895 (GOT32_symbol_operand): Likewise.
16896 * config/i386/i386.md (indirect_jump): Call convert_memory_address
16897 for -mindirect-branch-register.
16898 (tablejump): Likewise.
16899 (*sibcall_memory): Likewise.
16900 (*sibcall_value_memory): Likewise.
16901 Disallow peepholes of indirect call and jump via memory for
16902 -mindirect-branch-register.
16903 (*call_pop): Replace m with Bw.
16904 (*call_value_pop): Likewise.
16905 (*sibcall_pop_memory): Replace m with Bs.
16906 * config/i386/i386.opt (mindirect-branch-register): New option.
16907 * doc/invoke.texi: Document -mindirect-branch-register option.
16909 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16911 * config/i386/i386-protos.h (ix86_output_function_return): New.
16912 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
16913 set function_return_type.
16914 (indirect_thunk_name): Add ret_p to indicate thunk for function
16916 (output_indirect_thunk_function): Pass false to
16917 indirect_thunk_name.
16918 (ix86_output_indirect_branch_via_reg): Likewise.
16919 (ix86_output_indirect_branch_via_push): Likewise.
16920 (output_indirect_thunk_function): Create alias for function
16921 return thunk if regno < 0.
16922 (ix86_output_function_return): New function.
16923 (ix86_handle_fndecl_attribute): Handle function_return.
16924 (ix86_attribute_table): Add function_return.
16925 * config/i386/i386.h (machine_function): Add
16926 function_return_type.
16927 * config/i386/i386.md (simple_return_internal): Use
16928 ix86_output_function_return.
16929 (simple_return_internal_long): Likewise.
16930 * config/i386/i386.opt (mfunction-return=): New option.
16931 (indirect_branch): Mention -mfunction-return=.
16932 * doc/extend.texi: Document function_return function attribute.
16933 * doc/invoke.texi: Document -mfunction-return= option.
16935 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16937 * config/i386/i386-opts.h (indirect_branch): New.
16938 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
16939 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
16940 with local indirect jump when converting indirect call and jump.
16941 (ix86_set_indirect_branch_type): New.
16942 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
16943 (indirectlabelno): New.
16944 (indirect_thunk_needed): Likewise.
16945 (indirect_thunk_bnd_needed): Likewise.
16946 (indirect_thunks_used): Likewise.
16947 (indirect_thunks_bnd_used): Likewise.
16948 (INDIRECT_LABEL): Likewise.
16949 (indirect_thunk_name): Likewise.
16950 (output_indirect_thunk): Likewise.
16951 (output_indirect_thunk_function): Likewise.
16952 (ix86_output_indirect_branch_via_reg): Likewise.
16953 (ix86_output_indirect_branch_via_push): Likewise.
16954 (ix86_output_indirect_branch): Likewise.
16955 (ix86_output_indirect_jmp): Likewise.
16956 (ix86_code_end): Call output_indirect_thunk_function if needed.
16957 (ix86_output_call_insn): Call ix86_output_indirect_branch if
16959 (ix86_handle_fndecl_attribute): Handle indirect_branch.
16960 (ix86_attribute_table): Add indirect_branch.
16961 * config/i386/i386.h (machine_function): Add indirect_branch_type
16962 and has_local_indirect_jump.
16963 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
16965 (tablejump): Likewise.
16966 (*indirect_jump): Use ix86_output_indirect_jmp.
16967 (*tablejump_1): Likewise.
16968 (simple_return_indirect_internal): Likewise.
16969 * config/i386/i386.opt (mindirect-branch=): New option.
16970 (indirect_branch): New.
16973 (thunk-inline): Likewise.
16974 (thunk-extern): Likewise.
16975 * doc/extend.texi: Document indirect_branch function attribute.
16976 * doc/invoke.texi: Document -mindirect-branch= option.
16978 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
16981 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
16983 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
16985 * ipa-inline.c (want_inline_small_function_p): Return false if
16986 inlining has already failed with CIF_FINAL_ERROR.
16987 (update_caller_keys): Call want_inline_small_function_p before
16989 (update_callee_keys): Likewise.
16991 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
16993 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
16995 (rs6000_quadword_masked_address_p): Likewise.
16996 (quad_aligned_load_p): Likewise.
16997 (quad_aligned_store_p): Likewise.
16998 (const_load_sequence_p): Add comment to describe the outer-most loop.
16999 (mimic_memory_attributes_and_flags): New function.
17000 (rs6000_gen_stvx): Likewise.
17001 (replace_swapped_aligned_store): Likewise.
17002 (rs6000_gen_lvx): Likewise.
17003 (replace_swapped_aligned_load): Likewise.
17004 (replace_swapped_load_constant): Capitalize argument name in
17005 comment describing this function.
17006 (rs6000_analyze_swaps): Add a third pass to search for vector loads
17007 and stores that access quad-word aligned addresses and replace
17008 with stvx or lvx instructions when appropriate.
17009 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
17010 New function prototype.
17011 (rs6000_quadword_masked_address_p): Likewise.
17012 (rs6000_gen_lvx): Likewise.
17013 (rs6000_gen_stvx): Likewise.
17014 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
17015 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
17016 when memory address is aligned.
17017 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
17018 this split to select lvx instruction when memory address is aligned.
17019 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
17020 instruction when memory address is aligned.
17021 (*vsx_le_perm_load_v16qi): Likewise.
17022 (four unnamed splitters): Modify to select the stvx instruction
17023 when memory is aligned.
17025 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
17027 * predict.c (determine_unlikely_bbs): Handle correctly BBs
17028 which appears in the queue multiple times.
17030 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17031 Alan Hayward <alan.hayward@arm.com>
17032 David Sherwood <david.sherwood@arm.com>
17034 * tree-vectorizer.h (vec_lower_bound): New structure.
17035 (_loop_vec_info): Add check_nonzero and lower_bounds.
17036 (LOOP_VINFO_CHECK_NONZERO): New macro.
17037 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
17038 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
17039 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
17040 fields. Make seg_len the distance travelled, not including the
17042 (dr_direction_indicator): Declare.
17043 (dr_zero_step_indicator): Likewise.
17044 (dr_known_forward_stride_p): Likewise.
17045 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
17047 (runtime_alias_check_p): Allow runtime alias checks with
17049 (operator ==): Compare access_size and align.
17050 (prune_runtime_alias_test_list): Rework for new distinction between
17051 the access_size and seg_len.
17052 (create_intersect_range_checks_index): Likewise. Cope with polynomial
17054 (get_segment_min_max): New function.
17055 (create_intersect_range_checks): Use it.
17056 (dr_step_indicator): New function.
17057 (dr_direction_indicator): Likewise.
17058 (dr_zero_step_indicator): Likewise.
17059 (dr_known_forward_stride_p): Likewise.
17060 * tree-loop-distribution.c (data_ref_segment_size): Return
17061 DR_STEP * (niters - 1).
17062 (compute_alias_check_pairs): Update call to the dr_with_seg_len
17064 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
17065 (vect_preserves_scalar_order_p): New function, split out from...
17066 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
17067 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
17068 (vect_vfa_access_size): New function.
17069 (vect_vfa_align): Likewise.
17070 (vect_compile_time_alias): Take access_size_a and access_b arguments.
17071 (dump_lower_bound): New function.
17072 (vect_check_lower_bound): Likewise.
17073 (vect_small_gap_p): Likewise.
17074 (vectorizable_with_step_bound_p): Likewise.
17075 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
17076 depencies if the vectorization factor is 1. Convert the checks
17077 for nonzero steps into checks on the bounds of DR_STEP. Try using
17078 a bunds check for variable steps if the minimum required step is
17079 relatively small. Update calls to the dr_with_seg_len
17080 constructor and to vect_compile_time_alias.
17081 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
17083 (vect_loop_versioning): Call it.
17084 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
17086 (vect_estimate_min_profitable_iters): Account for any bounds checks.
17088 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17089 Alan Hayward <alan.hayward@arm.com>
17090 David Sherwood <david.sherwood@arm.com>
17092 * doc/sourcebuild.texi (vect_scatter_store): Document.
17093 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
17095 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
17097 * genopinit.c (main): Add supports_vec_scatter_store and
17098 supports_vec_scatter_store_cached to target_optabs.
17099 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
17100 IFN_MASK_SCATTER_STORE.
17101 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
17103 * internal-fn.h (internal_store_fn_p): Declare.
17104 (internal_fn_stored_value_index): Likewise.
17105 * internal-fn.c (scatter_store_direct): New macro.
17106 (expand_scatter_store_optab_fn): New function.
17107 (direct_scatter_store_optab_supported_p): New macro.
17108 (internal_store_fn_p): New function.
17109 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
17110 IFN_MASK_SCATTER_STORE.
17111 (internal_fn_mask_index): Likewise.
17112 (internal_fn_stored_value_index): New function.
17113 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
17114 for scatter stores.
17115 * optabs-query.h (supports_vec_scatter_store_p): Declare.
17116 * optabs-query.c (supports_vec_scatter_store_p): New function.
17117 * tree-vectorizer.h (vect_get_store_rhs): Declare.
17118 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
17119 true for scatter stores.
17120 (vect_gather_scatter_fn_p): Handle scatter stores too.
17121 (vect_check_gather_scatter): Consider using scatter stores if
17122 supports_vec_scatter_store_p.
17123 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
17124 scatter stores too.
17125 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
17126 internal_fn_stored_value_index.
17127 (check_load_store_masking): Handle scatter stores too.
17128 (vect_get_store_rhs): Make public.
17129 (vectorizable_call): Use internal_store_fn_p.
17130 (vectorizable_store): Handle scatter store internal functions.
17131 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
17132 when deciding whether the end of the group has been reached.
17133 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
17134 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
17135 (mask_scatter_store<mode>): New insns.
17137 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17138 Alan Hayward <alan.hayward@arm.com>
17139 David Sherwood <david.sherwood@arm.com>
17141 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
17142 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
17143 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
17145 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
17146 Use vect_truncate_gather_scatter_offset if we can't treat the
17147 operation as a normal gather load or scatter store.
17148 (get_group_load_store_type): Take the gather_scatter_info
17149 as argument. Try using a gather load or scatter store for
17150 single-element groups.
17151 (get_load_store_type): Update calls to get_group_load_store_type
17152 and vect_use_strided_gather_scatters_p.
17154 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17155 Alan Hayward <alan.hayward@arm.com>
17156 David Sherwood <david.sherwood@arm.com>
17158 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
17159 optional tree argument.
17160 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
17162 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
17163 but continue to use the current value as a fallback.
17164 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
17165 to compare the updates.
17166 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
17167 (get_load_store_type): Use it when handling a strided access.
17168 (vect_get_strided_load_store_ops): New function.
17169 (vect_get_data_ptr_increment): Likewise.
17170 (vectorizable_load): Handle strided gather loads. Always pass
17171 a step to vect_create_data_ref_ptr and bump_vector_ptr.
17173 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17174 Alan Hayward <alan.hayward@arm.com>
17175 David Sherwood <david.sherwood@arm.com>
17177 * doc/md.texi (gather_load@var{m}): Document.
17178 (mask_gather_load@var{m}): Likewise.
17179 * genopinit.c (main): Add supports_vec_gather_load and
17180 supports_vec_gather_load_cached to target_optabs.
17181 * optabs-tree.c (init_tree_optimization_optabs): Use
17182 ggc_cleared_alloc to allocate target_optabs.
17183 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
17184 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
17186 * internal-fn.h (internal_load_fn_p): Declare.
17187 (internal_gather_scatter_fn_p): Likewise.
17188 (internal_fn_mask_index): Likewise.
17189 (internal_gather_scatter_fn_supported_p): Likewise.
17190 * internal-fn.c (gather_load_direct): New macro.
17191 (expand_gather_load_optab_fn): New function.
17192 (direct_gather_load_optab_supported_p): New macro.
17193 (direct_internal_fn_optab): New function.
17194 (internal_load_fn_p): Likewise.
17195 (internal_gather_scatter_fn_p): Likewise.
17196 (internal_fn_mask_index): Likewise.
17197 (internal_gather_scatter_fn_supported_p): Likewise.
17198 * optabs-query.c (supports_at_least_one_mode_p): New function.
17199 (supports_vec_gather_load_p): Likewise.
17200 * optabs-query.h (supports_vec_gather_load_p): Declare.
17201 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
17202 and memory_type field.
17203 (NUM_PATTERNS): Bump to 15.
17204 * tree-vect-data-refs.c: Include internal-fn.h.
17205 (vect_gather_scatter_fn_p): New function.
17206 (vect_describe_gather_scatter_call): Likewise.
17207 (vect_check_gather_scatter): Try using internal functions for
17208 gather loads. Recognize existing calls to a gather load function.
17209 (vect_analyze_data_refs): Consider using gather loads if
17210 supports_vec_gather_load_p.
17211 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
17212 (vect_get_gather_scatter_offset_type): Likewise.
17213 (vect_convert_mask_for_vectype): Likewise.
17214 (vect_add_conversion_to_patterm): Likewise.
17215 (vect_try_gather_scatter_pattern): Likewise.
17216 (vect_recog_gather_scatter_pattern): New pattern recognizer.
17217 (vect_vect_recog_func_ptrs): Add it.
17218 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
17219 internal_fn_mask_index and internal_gather_scatter_fn_p.
17220 (check_load_store_masking): Take the gather_scatter_info as an
17221 argument and handle gather loads.
17222 (vect_get_gather_scatter_ops): New function.
17223 (vectorizable_call): Check internal_load_fn_p.
17224 (vectorizable_load): Likewise. Handle gather load internal
17226 (vectorizable_store): Update call to check_load_store_masking.
17227 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
17228 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
17229 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
17230 (aarch64_gather_scale_operand_d): New predicates.
17231 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
17232 (mask_gather_load<mode>): New insns.
17234 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17235 Alan Hayward <alan.hayward@arm.com>
17236 David Sherwood <david.sherwood@arm.com>
17238 * optabs.def (fold_left_plus_optab): New optab.
17239 * doc/md.texi (fold_left_plus_@var{m}): Document.
17240 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
17241 * internal-fn.c (fold_left_direct): Define.
17242 (expand_fold_left_optab_fn): Likewise.
17243 (direct_fold_left_optab_supported_p): Likewise.
17244 * fold-const-call.c (fold_const_fold_left): New function.
17245 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
17246 * tree-parloops.c (valid_reduction_p): New function.
17247 (gather_scalar_reductions): Use it.
17248 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
17249 (vect_finish_replace_stmt): Declare.
17250 * tree-vect-loop.c (fold_left_reduction_fn): New function.
17251 (needs_fold_left_reduction_p): New function, split out from...
17252 (vect_is_simple_reduction): ...here. Accept reductions that
17253 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
17254 (vect_force_simple_reduction): Also store the reduction type in
17255 the assignment's STMT_VINFO_REDUC_TYPE.
17256 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
17257 (merge_with_identity): New function.
17258 (vect_expand_fold_left): Likewise.
17259 (vectorize_fold_left_reduction): Likewise.
17260 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
17261 scalar phi in place for it. Check for target support and reject
17262 cases that would reassociate the operation. Defer the transform
17263 phase to vectorize_fold_left_reduction.
17264 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
17265 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
17266 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
17268 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17270 * tree-if-conv.c (predicate_mem_writes): Remove redundant
17271 call to ifc_temp_var.
17273 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17274 Alan Hayward <alan.hayward@arm.com>
17275 David Sherwood <david.sherwood@arm.com>
17277 * target.def (legitimize_address_displacement): Take the original
17278 offset as a poly_int.
17279 * targhooks.h (default_legitimize_address_displacement): Update
17281 * targhooks.c (default_legitimize_address_displacement): Likewise.
17282 * doc/tm.texi: Regenerate.
17283 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
17284 as an argument, moving assert of ad->disp == ad->disp_term to...
17285 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
17286 Try calling targetm.legitimize_address_displacement before expanding
17287 the address rather than afterwards, and adjust for the new interface.
17288 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
17289 Match the new hook interface. Handle SVE addresses.
17290 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
17291 new hook interface.
17293 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17295 * Makefile.in (OBJS): Add early-remat.o.
17296 * target.def (select_early_remat_modes): New hook.
17297 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
17298 * doc/tm.texi: Regenerate.
17299 * targhooks.h (default_select_early_remat_modes): Declare.
17300 * targhooks.c (default_select_early_remat_modes): New function.
17301 * timevar.def (TV_EARLY_REMAT): New timevar.
17302 * passes.def (pass_early_remat): New pass.
17303 * tree-pass.h (make_pass_early_remat): Declare.
17304 * early-remat.c: New file.
17305 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
17307 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
17309 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17310 Alan Hayward <alan.hayward@arm.com>
17311 David Sherwood <david.sherwood@arm.com>
17313 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
17314 vfm1 with a bound_epilog parameter.
17315 (vect_do_peeling): Update calls accordingly, and move the prologue
17316 call earlier in the function. Treat the base bound_epilog as 0 for
17317 fully-masked loops and retain vf - 1 for other loops. Add 1 to
17318 this base when peeling for gaps.
17319 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
17320 with fully-masked loops.
17321 (vect_estimate_min_profitable_iters): Handle the single peeled
17322 iteration in that case.
17324 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17325 Alan Hayward <alan.hayward@arm.com>
17326 David Sherwood <david.sherwood@arm.com>
17328 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
17329 single-element interleaving even if the size is not a power of 2.
17330 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
17331 accesses for single-element interleaving if the group size is
17334 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17335 Alan Hayward <alan.hayward@arm.com>
17336 David Sherwood <david.sherwood@arm.com>
17338 * doc/md.texi (fold_extract_last_@var{m}): Document.
17339 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
17340 * optabs.def (fold_extract_last_optab): New optab.
17341 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
17342 * internal-fn.c (fold_extract_direct): New macro.
17343 (expand_fold_extract_optab_fn): Likewise.
17344 (direct_fold_extract_optab_supported_p): Likewise.
17345 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
17346 * tree-vect-loop.c (vect_model_reduction_cost): Handle
17347 EXTRACT_LAST_REDUCTION.
17348 (get_initial_def_for_reduction): Do not create an initial vector
17349 for EXTRACT_LAST_REDUCTION reductions.
17350 (vectorizable_reduction): Leave the scalar phi in place for
17351 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
17352 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
17353 epilogue code for EXTRACT_LAST_REDUCTION and defer the
17354 transform phase to vectorizable_condition.
17355 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
17357 (vect_finish_stmt_generation): ...here.
17358 (vect_finish_replace_stmt): New function.
17359 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
17360 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
17362 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
17364 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17365 Alan Hayward <alan.hayward@arm.com>
17366 David Sherwood <david.sherwood@arm.com>
17368 * doc/md.texi (extract_last_@var{m}): Document.
17369 * optabs.def (extract_last_optab): New optab.
17370 * internal-fn.def (EXTRACT_LAST): New internal function.
17371 * internal-fn.c (cond_unary_direct): New macro.
17372 (expand_cond_unary_optab_fn): Likewise.
17373 (direct_cond_unary_optab_supported_p): Likewise.
17374 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
17375 loops using EXTRACT_LAST.
17376 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
17377 (extract_last_<mode>): ...this optab.
17378 (vec_extract<mode><Vel>): Update accordingly.
17380 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17381 Alan Hayward <alan.hayward@arm.com>
17382 David Sherwood <david.sherwood@arm.com>
17384 * target.def (empty_mask_is_expensive): New hook.
17385 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
17386 * doc/tm.texi: Regenerate.
17387 * targhooks.h (default_empty_mask_is_expensive): Declare.
17388 * targhooks.c (default_empty_mask_is_expensive): New function.
17389 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
17390 if the target says that empty masks are expensive.
17391 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
17393 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
17395 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17396 Alan Hayward <alan.hayward@arm.com>
17397 David Sherwood <david.sherwood@arm.com>
17399 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
17400 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
17401 (vect_use_loop_mask_for_alignment_p): New function.
17402 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
17403 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
17404 niters_skip argument. Make sure that the first niters_skip elements
17405 of the first iteration are inactive.
17406 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
17407 Update call to vect_set_loop_masks_directly.
17408 (get_misalign_in_elems): New function, split out from...
17409 (vect_gen_prolog_loop_niters): ...here.
17410 (vect_update_init_of_dr): Take a code argument that specifies whether
17411 the adjustment should be added or subtracted.
17412 (vect_update_init_of_drs): Likewise.
17413 (vect_prepare_for_masked_peels): New function.
17414 (vect_do_peeling): Skip prologue peeling if we're using a mask
17415 instead. Update call to vect_update_inits_of_drs.
17416 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
17418 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
17419 alignment. Do not include the number of peeled iterations in
17420 the minimum threshold in that case.
17421 (vectorizable_induction): Adjust the start value down by
17422 LOOP_VINFO_MASK_SKIP_NITERS iterations.
17423 (vect_transform_loop): Call vect_prepare_for_masked_peels.
17424 Take the number of skipped iterations into account when calculating
17426 * tree-vect-stmts.c (vect_gen_while_not): New function.
17428 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17429 Alan Hayward <alan.hayward@arm.com>
17430 David Sherwood <david.sherwood@arm.com>
17432 * doc/sourcebuild.texi (vect_fully_masked): Document.
17433 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
17434 default value to 0.
17435 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
17437 (vect_analyze_loop_2): ...here. Don't check the vectorization
17438 factor against the number of loop iterations if the loop is
17441 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17442 Alan Hayward <alan.hayward@arm.com>
17443 David Sherwood <david.sherwood@arm.com>
17445 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
17446 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
17447 (dump_groups): Update accordingly.
17448 (iv_use::mem_type): New member variable.
17449 (address_p): New function.
17450 (record_use): Add a mem_type argument and initialize the new
17452 (record_group_use): Add a mem_type argument. Use address_p.
17453 Remove obsolete null checks of base_object. Update call to record_use.
17454 (find_interesting_uses_op): Update call to record_group_use.
17455 (find_interesting_uses_cond): Likewise.
17456 (find_interesting_uses_address): Likewise.
17457 (get_mem_type_for_internal_fn): New function.
17458 (find_address_like_use): Likewise.
17459 (find_interesting_uses_stmt): Try find_address_like_use before
17460 calling find_interesting_uses_op.
17461 (addr_offset_valid_p): Use the iv mem_type field as the type
17462 of the addressed memory.
17463 (add_autoinc_candidates): Likewise.
17464 (get_address_cost): Likewise.
17465 (split_small_address_groups_p): Use address_p.
17466 (split_address_groups): Likewise.
17467 (add_iv_candidate_for_use): Likewise.
17468 (autoinc_possible_for_pair): Likewise.
17469 (rewrite_groups): Likewise.
17470 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
17471 (determine_group_iv_cost): Update after split of USE_ADDRESS.
17472 (get_alias_ptr_type_for_ptr_address): New function.
17473 (rewrite_use_address): Rewrite address uses in calls that were
17474 identified by find_address_like_use.
17476 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17477 Alan Hayward <alan.hayward@arm.com>
17478 David Sherwood <david.sherwood@arm.com>
17480 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
17482 * gimple-expr.h (is_gimple_addressable: Likewise.
17483 * gimple-expr.c (is_gimple_address): Likewise.
17484 * internal-fn.c (expand_call_mem_ref): New function.
17485 (expand_mask_load_optab_fn): Use it.
17486 (expand_mask_store_optab_fn): Likewise.
17488 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17489 Alan Hayward <alan.hayward@arm.com>
17490 David Sherwood <david.sherwood@arm.com>
17492 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
17493 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
17494 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
17495 (cond_umax@var{mode}): Document.
17496 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
17497 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
17498 (cond_umin_optab, cond_umax_optab): New optabs.
17499 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
17500 (COND_IOR, COND_XOR): New internal functions.
17501 * internal-fn.h (get_conditional_internal_fn): Declare.
17502 * internal-fn.c (cond_binary_direct): New macro.
17503 (expand_cond_binary_optab_fn): Likewise.
17504 (direct_cond_binary_optab_supported_p): Likewise.
17505 (get_conditional_internal_fn): New function.
17506 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
17507 Cope with reduction statements that are vectorized as calls rather
17509 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
17510 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
17511 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
17512 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
17513 (UNSPEC_COND_EOR): New unspecs.
17514 (optab): Add mappings for them.
17515 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
17516 (sve_int_op, sve_fp_op): New int attributes.
17518 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17519 Alan Hayward <alan.hayward@arm.com>
17520 David Sherwood <david.sherwood@arm.com>
17522 * optabs.def (while_ult_optab): New optab.
17523 * doc/md.texi (while_ult@var{m}@var{n}): Document.
17524 * internal-fn.def (WHILE_ULT): New internal function.
17525 * internal-fn.h (direct_internal_fn_supported_p): New override
17526 that takes two types as argument.
17527 * internal-fn.c (while_direct): New macro.
17528 (expand_while_optab_fn): New function.
17529 (convert_optab_supported_p): Likewise.
17530 (direct_while_optab_supported_p): New macro.
17531 * wide-int.h (wi::udiv_ceil): New function.
17532 * tree-vectorizer.h (rgroup_masks): New structure.
17533 (vec_loop_masks): New typedef.
17534 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
17535 and fully_masked_p.
17536 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
17537 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
17538 (vect_max_vf): New function.
17539 (slpeel_make_loop_iterate_ntimes): Delete.
17540 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
17541 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
17542 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
17543 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
17544 internal-fn.h, stor-layout.h and optabs-query.h.
17545 (vect_set_loop_mask): New function.
17546 (add_preheader_seq): Likewise.
17547 (add_header_seq): Likewise.
17548 (interleave_supported_p): Likewise.
17549 (vect_maybe_permute_loop_masks): Likewise.
17550 (vect_set_loop_masks_directly): Likewise.
17551 (vect_set_loop_condition_masked): Likewise.
17552 (vect_set_loop_condition_unmasked): New function, split out from
17553 slpeel_make_loop_iterate_ntimes.
17554 (slpeel_make_loop_iterate_ntimes): Rename to..
17555 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
17556 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
17557 (vect_do_peeling): Update call accordingly.
17558 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
17560 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
17561 mask_compare_type, can_fully_mask_p and fully_masked_p.
17562 (release_vec_loop_masks): New function.
17563 (_loop_vec_info): Use it to free the loop masks.
17564 (can_produce_all_loop_masks_p): New function.
17565 (vect_get_max_nscalars_per_iter): Likewise.
17566 (vect_verify_full_masking): Likewise.
17567 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
17568 retries, and free the mask rgroups before retrying. Check loop-wide
17569 reasons for disallowing fully-masked loops. Make the final decision
17570 about whether use a fully-masked loop or not.
17571 (vect_estimate_min_profitable_iters): Do not assume that peeling
17572 for the number of iterations will be needed for fully-masked loops.
17573 (vectorizable_reduction): Disable fully-masked loops.
17574 (vectorizable_live_operation): Likewise.
17575 (vect_halve_mask_nunits): New function.
17576 (vect_double_mask_nunits): Likewise.
17577 (vect_record_loop_mask): Likewise.
17578 (vect_get_loop_mask): Likewise.
17579 (vect_transform_loop): Handle the case in which the final loop
17580 iteration might handle a partial vector. Call vect_set_loop_condition
17581 instead of slpeel_make_loop_iterate_ntimes.
17582 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
17583 (check_load_store_masking): New function.
17584 (prepare_load_store_mask): Likewise.
17585 (vectorizable_store): Handle fully-masked loops.
17586 (vectorizable_load): Likewise.
17587 (supportable_widening_operation): Use vect_halve_mask_nunits for
17589 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
17590 (vect_gen_while): New function.
17591 * config/aarch64/aarch64.md (umax<mode>3): New expander.
17592 (aarch64_uqdec<mode>): New insn.
17594 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17595 Alan Hayward <alan.hayward@arm.com>
17596 David Sherwood <david.sherwood@arm.com>
17598 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
17599 (reduc_xor_scal_optab): New optabs.
17600 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
17601 (reduc_xor_scal_@var{m}): Document.
17602 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
17603 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
17604 internal functions.
17605 * fold-const-call.c (fold_const_call): Handle them.
17606 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
17607 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
17608 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
17609 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
17610 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
17611 (UNSPEC_XORV): New unspecs.
17612 (optab): Add entries for them.
17613 (BITWISEV): New int iterator.
17614 (bit_reduc_op): New int attributes.
17616 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17617 Alan Hayward <alan.hayward@arm.com>
17618 David Sherwood <david.sherwood@arm.com>
17620 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
17621 * internal-fn.def (VEC_SHL_INSERT): New internal function.
17622 * optabs.def (vec_shl_insert_optab): New optab.
17623 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
17624 (duplicate_and_interleave): Likewise.
17625 * tree-vect-loop.c: Include internal-fn.h.
17626 (neutral_op_for_slp_reduction): New function, split out from
17627 get_initial_defs_for_reduction.
17628 (get_initial_def_for_reduction): Handle option 2 for variable-length
17629 vectors by loading the neutral value into a vector and then shifting
17630 the initial value into element 0.
17631 (get_initial_defs_for_reduction): Replace the code argument with
17632 the neutral value calculated by neutral_op_for_slp_reduction.
17633 Use gimple_build_vector for constant-length vectors.
17634 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
17635 but the first group_size elements have a neutral value.
17636 Use duplicate_and_interleave otherwise.
17637 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
17638 Update call to get_initial_defs_for_reduction. Handle SLP
17639 reductions for variable-length vectors by creating one vector
17640 result for each scalar result, with the elements associated
17641 with other scalar results stubbed out with the neutral value.
17642 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
17643 Require IFN_VEC_SHL_INSERT for double reductions on
17644 variable-length vectors, or SLP reductions that have
17645 a neutral value. Require can_duplicate_and_interleave_p
17646 support for variable-length unchained SLP reductions if there
17647 is no neutral value, such as for MIN/MAX reductions. Also require
17648 the number of vector elements to be a multiple of the number of
17649 SLP statements when doing variable-length unchained SLP reductions.
17650 Update call to vect_create_epilog_for_reduction.
17651 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
17652 and remove initial values.
17653 (duplicate_and_interleave): Make public.
17654 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
17655 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
17657 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17658 Alan Hayward <alan.hayward@arm.com>
17659 David Sherwood <david.sherwood@arm.com>
17661 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
17662 (can_duplicate_and_interleave_p): New function.
17663 (vect_get_and_check_slp_defs): Take the vector of statements
17664 rather than just the current one. Remove excess parentheses.
17665 Restriction rejectinon of vect_constant_def and vect_external_def
17666 for variable-length vectors to boolean types, or types for which
17667 can_duplicate_and_interleave_p is false.
17668 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
17669 (duplicate_and_interleave): New function.
17670 (vect_get_constant_vectors): Use gimple_build_vector for
17671 constant-length vectors and suitable variable-length constant
17672 vectors. Use duplicate_and_interleave for other variable-length
17673 vectors. Don't defer the update when inserting new statements.
17675 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17676 Alan Hayward <alan.hayward@arm.com>
17677 David Sherwood <david.sherwood@arm.com>
17679 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
17680 min_profitable_iters doesn't go negative.
17682 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17683 Alan Hayward <alan.hayward@arm.com>
17684 David Sherwood <david.sherwood@arm.com>
17686 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
17687 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
17688 * optabs.def (vec_mask_load_lanes_optab): New optab.
17689 (vec_mask_store_lanes_optab): Likewise.
17690 * internal-fn.def (MASK_LOAD_LANES): New internal function.
17691 (MASK_STORE_LANES): Likewise.
17692 * internal-fn.c (mask_load_lanes_direct): New macro.
17693 (mask_store_lanes_direct): Likewise.
17694 (expand_mask_load_optab_fn): Handle masked operations.
17695 (expand_mask_load_lanes_optab_fn): New macro.
17696 (expand_mask_store_optab_fn): Handle masked operations.
17697 (expand_mask_store_lanes_optab_fn): New macro.
17698 (direct_mask_load_lanes_optab_supported_p): Likewise.
17699 (direct_mask_store_lanes_optab_supported_p): Likewise.
17700 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
17702 (vect_load_lanes_supported): Likewise.
17703 * tree-vect-data-refs.c (strip_conversion): New function.
17704 (can_group_stmts_p): Likewise.
17705 (vect_analyze_data_ref_accesses): Use it instead of checking
17706 for a pair of assignments.
17707 (vect_store_lanes_supported): Take a masked_p parameter.
17708 (vect_load_lanes_supported): Likewise.
17709 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
17710 vect_store_lanes_supported and vect_load_lanes_supported.
17711 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
17712 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
17713 parameter. Don't allow gaps for masked accesses.
17714 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
17715 and vect_load_lanes_supported.
17716 (get_load_store_type): Take a masked_p parameter and update
17717 call to get_group_load_store_type.
17718 (vectorizable_store): Update call to get_load_store_type.
17719 Handle IFN_MASK_STORE_LANES.
17720 (vectorizable_load): Update call to get_load_store_type.
17721 Handle IFN_MASK_LOAD_LANES.
17723 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17724 Alan Hayward <alan.hayward@arm.com>
17725 David Sherwood <david.sherwood@arm.com>
17727 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
17729 * config/aarch64/aarch64-protos.h
17730 (aarch64_sve_struct_memory_operand_p): Declare.
17731 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
17732 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
17733 (VPRED, vpred): Handle SVE structure modes.
17734 * config/aarch64/constraints.md (Utx): New constraint.
17735 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
17736 (aarch64_sve_struct_nonimmediate_operand): New predicates.
17737 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
17738 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
17739 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
17740 structure modes. Split into pieces after RA.
17741 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
17742 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
17744 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
17745 SVE structure modes.
17746 (aarch64_classify_address): Likewise.
17747 (sizetochar): Move earlier in file.
17748 (aarch64_print_operand): Handle SVE register lists.
17749 (aarch64_array_mode): New function.
17750 (aarch64_sve_struct_memory_operand_p): Likewise.
17751 (TARGET_ARRAY_MODE): Redefine.
17753 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17754 Alan Hayward <alan.hayward@arm.com>
17755 David Sherwood <david.sherwood@arm.com>
17757 * target.def (array_mode): New target hook.
17758 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
17759 * doc/tm.texi: Regenerate.
17760 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
17761 * hooks.c (hook_optmode_mode_uhwi_none): New function.
17762 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
17763 targetm.array_mode.
17764 * stor-layout.c (mode_for_array): Likewise. Support polynomial
17767 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17768 Alan Hayward <alan.hayward@arm.com>
17769 David Sherwood <david.sherwood@arm.com>
17771 * fold-const.c (fold_binary_loc): Check the argument types
17772 rather than the result type when testing for a vector operation.
17774 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17776 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
17777 * doc/tm.texi: Regenerate.
17779 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17780 Alan Hayward <alan.hayward@arm.com>
17781 David Sherwood <david.sherwood@arm.com>
17783 * doc/invoke.texi (-msve-vector-bits=): Document new option.
17784 (sve): Document new AArch64 extension.
17785 * doc/md.texi (w): Extend the description of the AArch64
17786 constraint to include SVE vectors.
17787 (Upl, Upa): Document new AArch64 predicate constraints.
17788 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
17790 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
17791 (msve-vector-bits=): New option.
17792 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
17793 SVE when these are disabled.
17794 (sve): New extension.
17795 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
17796 modes. Adjust their number of units based on aarch64_sve_vg.
17797 (MAX_BITSIZE_MODE_ANY_MODE): Define.
17798 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
17799 aarch64_addr_query_type.
17800 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
17801 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
17802 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
17803 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
17804 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
17805 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
17806 (aarch64_simd_imm_zero_p): Delete.
17807 (aarch64_check_zero_based_sve_index_immediate): Declare.
17808 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
17809 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
17810 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
17811 (aarch64_sve_float_mul_immediate_p): Likewise.
17812 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
17813 rather than an rtx.
17814 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
17815 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
17816 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
17817 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
17818 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
17819 (aarch64_regmode_natural_size): Likewise.
17820 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
17821 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
17823 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
17824 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
17825 for VG and the SVE predicate registers.
17826 (V_ALIASES): Add a "z"-prefixed alias.
17827 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
17828 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
17829 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
17830 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
17831 (REG_CLASS_NAMES): Add entries for them.
17832 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
17833 and the predicate registers.
17834 (aarch64_sve_vg): Declare.
17835 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
17836 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
17837 (REGMODE_NATURAL_SIZE): Define.
17838 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
17840 * config/aarch64/aarch64.c: Include cfgrtl.h.
17841 (simd_immediate_info): Add a constructor for series vectors,
17842 and an associated step field.
17843 (aarch64_sve_vg): New variable.
17844 (aarch64_dbx_register_number): Handle VG and the predicate registers.
17845 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
17846 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
17847 (VEC_ANY_DATA, VEC_STRUCT): New constants.
17848 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
17849 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
17850 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
17851 (aarch64_get_mask_mode): New functions.
17852 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
17853 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
17854 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
17855 predicate modes and predicate registers. Explicitly restrict
17856 GPRs to modes of 16 bytes or smaller. Only allow FP registers
17857 to store a vector mode if it is recognized by
17858 aarch64_classify_vector_mode.
17859 (aarch64_regmode_natural_size): New function.
17860 (aarch64_hard_regno_caller_save_mode): Return the original mode
17862 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
17863 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
17864 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
17865 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
17867 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
17868 does not overlap dest if the function is frame-related. Handle
17870 (aarch64_split_add_offset): New function.
17871 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
17872 them aarch64_add_offset.
17873 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
17874 and update call to aarch64_sub_sp.
17875 (aarch64_add_cfa_expression): New function.
17876 (aarch64_expand_prologue): Pass extra temporary registers to the
17877 functions above. Handle the case in which we need to emit new
17878 DW_CFA_expressions for registers that were originally saved
17879 relative to the stack pointer, but now have to be expressed
17880 relative to the frame pointer.
17881 (aarch64_output_mi_thunk): Pass extra temporary registers to the
17883 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
17884 IP0 and IP1 values for SVE frames.
17885 (aarch64_expand_vec_series): New function.
17886 (aarch64_expand_sve_widened_duplicate): Likewise.
17887 (aarch64_expand_sve_const_vector): Likewise.
17888 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
17889 Handle SVE constants. Use emit_move_insn to move a force_const_mem
17890 into the register, rather than emitting a SET directly.
17891 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
17892 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
17893 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
17894 (offset_9bit_signed_scaled_p): New functions.
17895 (aarch64_replicate_bitmask_imm): New function.
17896 (aarch64_bitmask_imm): Use it.
17897 (aarch64_cannot_force_const_mem): Reject expressions involving
17898 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
17899 (aarch64_classify_index): Handle SVE indices, by requiring
17900 a plain register index with a scale that matches the element size.
17901 (aarch64_classify_address): Handle SVE addresses. Assert that
17902 the mode of the address is VOIDmode or an integer mode.
17903 Update call to aarch64_classify_symbol.
17904 (aarch64_classify_symbolic_expression): Update call to
17905 aarch64_classify_symbol.
17906 (aarch64_const_vec_all_in_range_p): New function.
17907 (aarch64_print_vector_float_operand): Likewise.
17908 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
17909 "vN" for FP registers with SVE modes. Handle (const ...) vectors
17910 and the FP immediates 1.0 and 0.5.
17911 (aarch64_print_address_internal): Handle SVE addresses.
17912 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
17913 (aarch64_regno_regclass): Handle predicate registers.
17914 (aarch64_secondary_reload): Handle big-endian reloads of SVE
17916 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
17917 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
17918 (aarch64_convert_sve_vector_bits): New function.
17919 (aarch64_override_options): Use it to handle -msve-vector-bits=.
17920 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
17921 rather than an rtx.
17922 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
17923 Handle SVE vector and predicate modes. Accept VL-based constants
17924 that need only one temporary register, and VL offsets that require
17925 no temporary registers.
17926 (aarch64_conditional_register_usage): Mark the predicate registers
17927 as fixed if SVE isn't available.
17928 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
17929 Return true for SVE vector and predicate modes.
17930 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
17931 rather than an unsigned int. Handle SVE modes.
17932 (aarch64_preferred_simd_mode): Update call accordingly. Handle
17934 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
17936 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
17937 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
17938 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
17939 (aarch64_sve_float_mul_immediate_p): New functions.
17940 (aarch64_sve_valid_immediate): New function.
17941 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
17942 Explicitly reject structure modes. Check for INDEX constants.
17943 Handle PTRUE and PFALSE constants.
17944 (aarch64_check_zero_based_sve_index_immediate): New function.
17945 (aarch64_simd_imm_zero_p): Delete.
17946 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
17947 vector modes. Accept constants in the range of CNT[BHWD].
17948 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
17949 ask for an Advanced SIMD mode.
17950 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
17951 (aarch64_simd_vector_alignment): Handle SVE predicates.
17952 (aarch64_vectorize_preferred_vector_alignment): New function.
17953 (aarch64_simd_vector_alignment_reachable): Use it instead of
17955 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
17956 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
17958 (MAX_VECT_LEN): Delete.
17959 (expand_vec_perm_d): Add a vec_flags field.
17960 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
17961 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
17962 (aarch64_evpc_ext): Don't apply a big-endian lane correction
17964 (aarch64_evpc_rev): Rename to...
17965 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
17966 (aarch64_evpc_rev_global): New function.
17967 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
17968 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
17970 (aarch64_evpc_sve_tbl): New function.
17971 (aarch64_expand_vec_perm_const_1): Update after rename of
17972 aarch64_evpc_rev. Handle SVE permutes too, trying
17973 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
17974 than aarch64_evpc_tbl.
17975 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
17976 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
17977 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
17978 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
17979 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
17980 (aarch64_expand_sve_vcond): New functions.
17981 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
17982 of aarch64_vector_mode_p.
17983 (aarch64_dwarf_poly_indeterminate_value): New function.
17984 (aarch64_compute_pressure_classes): Likewise.
17985 (aarch64_can_change_mode_class): Likewise.
17986 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
17987 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
17988 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
17989 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
17990 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
17991 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
17992 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
17993 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
17995 (Dn, Dl, Dr): Accept const as well as const_vector.
17996 (Dz): Likewise. Compare against CONST0_RTX.
17997 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
17998 of "vector" where appropriate.
17999 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
18000 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
18001 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
18002 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
18003 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
18004 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
18005 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
18006 (v_int_equiv): Extend to SVE modes.
18007 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
18009 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
18010 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
18011 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
18012 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
18013 (SVE_COND_FP_CMP): New int iterators.
18014 (perm_hilo): Handle the new unpack unspecs.
18015 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
18017 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
18018 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
18019 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
18020 (aarch64_equality_operator, aarch64_constant_vector_operand)
18021 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
18022 (aarch64_sve_nonimmediate_operand): Likewise.
18023 (aarch64_sve_general_operand): Likewise.
18024 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
18025 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
18026 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
18027 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
18028 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
18029 (aarch64_sve_float_arith_immediate): Likewise.
18030 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
18031 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
18032 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
18033 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
18034 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
18035 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
18036 (aarch64_sve_float_arith_operand): Likewise.
18037 (aarch64_sve_float_arith_with_sub_operand): Likewise.
18038 (aarch64_sve_float_mul_operand): Likewise.
18039 (aarch64_sve_vec_perm_operand): Likewise.
18040 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
18041 (aarch64_mov_operand): Accept const_poly_int and const_vector.
18042 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
18043 as well as const_vector.
18044 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
18045 in file. Use CONST0_RTX and CONSTM1_RTX.
18046 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
18047 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
18048 Use aarch64_simd_imm_zero.
18049 * config/aarch64/aarch64-sve.md: New file.
18050 * config/aarch64/aarch64.md: Include it.
18051 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
18052 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
18053 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
18054 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
18055 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
18056 (sve): New attribute.
18057 (enabled): Disable instructions with the sve attribute unless
18059 (movqi, movhi): Pass CONST_POLY_INT operaneds through
18060 aarch64_expand_mov_immediate.
18061 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
18062 CNT[BHSD] immediates.
18063 (movti): Split CONST_POLY_INT moves into two halves.
18064 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
18065 Split additions that need a temporary here if the destination
18066 is the stack pointer.
18067 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
18068 (*add<mode>3_poly_1): New instruction.
18069 (set_clobber_cc): New expander.
18071 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18073 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
18074 parameter and use it instead of GET_MODE_SIZE (innermode). Use
18075 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
18076 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
18077 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
18078 Change innermode from fixed_mode_size to machine_mode.
18079 (simplify_subreg): Update call accordingly. Handle a constant-sized
18080 subreg of a variable-length CONST_VECTOR.
18082 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18083 Alan Hayward <alan.hayward@arm.com>
18084 David Sherwood <david.sherwood@arm.com>
18086 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
18087 (add_offset_to_base): New function, split out from...
18088 (create_mem_ref): ...here. When handling a scale other than 1,
18089 check first whether the address is valid without the offset.
18090 Add it into the base if so, leaving the index and scale as-is.
18092 2018-01-12 Jakub Jelinek <jakub@redhat.com>
18095 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
18096 fold_for_warn before checking if arg2 is INTEGER_CST.
18098 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
18100 * config/rs6000/predicates.md (load_multiple_operation): Delete.
18101 (store_multiple_operation): Delete.
18102 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
18103 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
18104 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
18105 guarded by TARGET_STRING.
18106 (rs6000_output_load_multiple): Delete.
18107 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
18108 OPTION_MASK_STRING / TARGET_STRING handling.
18109 (print_operand) <'N', 'O'>: Add comment that these are unused now.
18110 (const rs6000_opt_masks) <"string">: Change mask to 0.
18111 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
18112 (MASK_STRING): Delete.
18113 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
18115 (load_multiple): Delete.
18122 (store_multiple): Delete.
18129 (movmemsi_8reg): Delete.
18130 (corresponding unnamed define_insn): Delete.
18131 (movmemsi_6reg): Delete.
18132 (corresponding unnamed define_insn): Delete.
18133 (movmemsi_4reg): Delete.
18134 (corresponding unnamed define_insn): Delete.
18135 (movmemsi_2reg): Delete.
18136 (corresponding unnamed define_insn): Delete.
18137 (movmemsi_1reg): Delete.
18138 (corresponding unnamed define_insn): Delete.
18139 * config/rs6000/rs6000.opt (mno-string): New.
18140 (mstring): Replace by deprecation warning stub.
18141 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
18143 2018-01-12 Jakub Jelinek <jakub@redhat.com>
18145 * regrename.c (regrename_do_replace): If replacing the same
18146 reg multiple times, try to reuse last created gen_raw_REG.
18149 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
18150 main to workaround a bug in GDB.
18152 2018-01-12 Tom de Vries <tom@codesourcery.com>
18155 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
18157 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
18159 PR rtl-optimization/80481
18160 * ira-color.c (get_cap_member): New function.
18161 (allocnos_conflict_by_live_ranges_p): Use it.
18162 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
18163 (setup_slot_coalesced_allocno_live_ranges): Ditto.
18165 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
18168 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
18169 (*saddl_se_1): Ditto.
18170 (*ssubsi_1): Ditto.
18171 (*ssubl_se_1): Ditto.
18173 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
18175 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
18176 rather than wi::to_widest for DR_INITs.
18177 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
18178 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
18179 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
18181 (vect_analyze_group_access_1): Note that here.
18183 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
18185 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
18186 polynomial type sizes.
18188 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
18190 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
18191 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
18192 (gimple_add_tmp_var): Likewise.
18194 2018-01-12 Martin Liska <mliska@suse.cz>
18196 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
18197 (gimple_alloc_sizes): Likewise.
18198 (dump_gimple_statistics): Use PRIu64 in printf format.
18199 * gimple.h: Change uint64_t to int.
18201 2018-01-12 Martin Liska <mliska@suse.cz>
18203 * tree-core.h: Use uint64_t instead of int.
18204 * tree.c (tree_node_counts): Likewise.
18205 (tree_node_sizes): Likewise.
18206 (dump_tree_statistics): Use PRIu64 in printf format.
18208 2018-01-12 Martin Liska <mliska@suse.cz>
18210 * Makefile.in: As qsort_chk is implemented in vec.c, add
18211 vec.o to linkage of gencfn-macros.
18212 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
18213 passing the info to record_node_allocation_statistics.
18214 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
18216 * ggc-common.c (struct ggc_usage): Add operator== and use
18217 it in operator< and compare function.
18218 * mem-stats.h (struct mem_usage): Likewise.
18219 * vec.c (struct vec_usage): Remove operator< and compare
18220 function. Can be simply inherited.
18222 2018-01-12 Martin Jambor <mjambor@suse.cz>
18225 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
18226 * tree-ssa-math-opts.c: Include domwalk.h.
18227 (convert_mult_to_fma_1): New function.
18228 (fma_transformation_info): New type.
18229 (fma_deferring_state): Likewise.
18230 (cancel_fma_deferring): New function.
18231 (result_of_phi): Likewise.
18232 (last_fma_candidate_feeds_initial_phi): Likewise.
18233 (convert_mult_to_fma): Added deferring logic, split actual
18234 transformation to convert_mult_to_fma_1.
18235 (math_opts_dom_walker): New type.
18236 (math_opts_dom_walker::after_dom_children): New method, body moved
18237 here from pass_optimize_widening_mul::execute, added deferring logic
18239 (pass_optimize_widening_mul::execute): Moved most of code to
18240 math_opts_dom_walker::after_dom_children.
18241 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
18242 * config/i386/i386.c (ix86_option_override_internal): Added
18243 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
18245 2018-01-12 Richard Biener <rguenther@suse.de>
18248 * dwarf2out.c (gen_variable_die): Do not reset old_die for
18249 inline instance vars.
18251 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
18254 * config/rx/rx.c (rx_is_restricted_memory_address):
18255 Handle SUBREG case.
18257 2018-01-12 Richard Biener <rguenther@suse.de>
18259 PR tree-optimization/80846
18260 * target.def (split_reduction): New target hook.
18261 * targhooks.c (default_split_reduction): New function.
18262 * targhooks.h (default_split_reduction): Declare.
18263 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
18264 target requests first reduce vectors by combining low and high
18266 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
18267 (get_vectype_for_scalar_type_and_size): Export.
18268 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
18269 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
18270 * doc/tm.texi: Regenerate.
18271 * config/i386/i386.c (ix86_split_reduction): Implement
18272 TARGET_VECTORIZE_SPLIT_REDUCTION.
18274 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
18277 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
18278 in PIC mode except for TARGET_VXWORKS_RTP.
18279 * config/sparc/sparc.c: Include cfgrtl.h.
18280 (TARGET_INIT_PIC_REG): Define.
18281 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
18282 (sparc_pic_register_p): New predicate.
18283 (sparc_legitimate_address_p): Use it.
18284 (sparc_legitimize_pic_address): Likewise.
18285 (sparc_delegitimize_address): Likewise.
18286 (sparc_mode_dependent_address_p): Likewise.
18287 (gen_load_pcrel_sym): Remove 4th parameter.
18288 (load_got_register): Adjust call to above. Remove obsolete stuff.
18289 (sparc_expand_prologue): Do not call load_got_register here.
18290 (sparc_flat_expand_prologue): Likewise.
18291 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
18292 (sparc_use_pseudo_pic_reg): New function.
18293 (sparc_init_pic_reg): Likewise.
18294 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
18295 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
18297 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
18299 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
18300 Add item for branch_cost.
18302 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
18304 PR rtl-optimization/83565
18305 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
18306 not extend the result to a larger mode for rotate operations.
18307 (num_sign_bit_copies1): Likewise.
18309 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18312 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
18314 Use values-Xc.o for -pedantic.
18315 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
18317 2018-01-12 Martin Liska <mliska@suse.cz>
18320 * ipa-devirt.c (final_warning_record::grow_type_warnings):
18322 (possible_polymorphic_call_targets): Use it.
18323 (ipa_devirt): Likewise.
18325 2018-01-12 Martin Liska <mliska@suse.cz>
18327 * profile-count.h (enum profile_quality): Use 0 as invalid
18328 enum value of profile_quality.
18330 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
18332 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
18333 -mext-string options.
18335 2018-01-12 Richard Biener <rguenther@suse.de>
18337 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
18338 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
18339 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
18341 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
18343 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
18345 * configure.ac (--with-long-double-format): Add support for the
18346 configuration option to change the default long double format on
18348 * config.gcc (powerpc*-linux*-*): Likewise.
18349 * configure: Regenerate.
18350 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
18351 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
18352 used without modification.
18354 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18356 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
18357 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
18358 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
18359 MISC_BUILTIN_SPEC_BARRIER.
18360 (rs6000_init_builtins): Likewise.
18361 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
18363 (speculation_barrier): New define_insn.
18364 * doc/extend.texi: Document __builtin_speculation_barrier.
18366 2018-01-11 Jakub Jelinek <jakub@redhat.com>
18369 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
18370 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
18371 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
18373 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
18374 integral modes instead of "ss" and "sd".
18375 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
18376 vectors with 32-bit and 64-bit elements.
18377 (vecdupssescalarmodesuffix): New mode attribute.
18378 (vec_dup<mode>): Use it.
18380 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
18383 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
18384 frame if argument is passed on stack.
18386 2018-01-11 Jakub Jelinek <jakub@redhat.com>
18389 * ree.c (combine_reaching_defs): Optimize also
18390 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
18391 reg2=any_extend(exp); reg1=reg2;, formatting fix.
18393 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
18395 PR middle-end/83189
18396 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
18398 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
18400 PR middle-end/83718
18401 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
18402 after they are computed.
18404 2018-01-11 Bin Cheng <bin.cheng@arm.com>
18406 PR tree-optimization/83695
18407 * gimple-loop-linterchange.cc
18408 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
18409 reset cached scev information after interchange.
18410 (pass_linterchange::execute): Remove call to scev_reset_htab.
18412 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18414 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
18415 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
18416 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
18417 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
18418 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
18419 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
18420 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
18421 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
18422 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
18423 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
18424 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
18425 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
18426 (V_lane_reg): Likewise.
18427 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
18429 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
18430 (vfmal_lane_low<mode>_intrinsic,
18431 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
18432 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
18433 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
18434 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
18435 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
18436 vfmsl_lane_high<mode>_intrinsic): New define_insns.
18438 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18440 * config/arm/arm-cpus.in (fp16fml): New feature.
18441 (ALL_SIMD): Add fp16fml.
18442 (armv8.2-a): Add fp16fml as an option.
18443 (armv8.3-a): Likewise.
18444 (armv8.4-a): Add fp16fml as part of fp16.
18445 * config/arm/arm.h (TARGET_FP16FML): Define.
18446 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
18448 * config/arm/arm-modes.def (V2HF): Define.
18449 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
18450 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
18451 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
18452 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
18453 vfmsl_low, vfmsl_high): New set of builtins.
18454 * config/arm/iterators.md (PLUSMINUS): New code iterator.
18455 (vfml_op): New code attribute.
18456 (VFMLHALVES): New int iterator.
18457 (VFML, VFMLSEL): New mode attributes.
18458 (V_reg): Define mapping for V2HF.
18459 (V_hi, V_lo): New mode attributes.
18460 (VF_constraint): Likewise.
18461 (vfml_half, vfml_half_selector): New int attributes.
18462 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
18464 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
18465 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
18467 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
18468 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
18469 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
18470 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
18472 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
18473 Document new effective target and option set.
18475 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18477 * config/arm/arm-cpus.in (armv8_4): New feature.
18478 (ARMv8_4a): New fgroup.
18479 (armv8.4-a): New arch.
18480 * config/arm/arm-tables.opt: Regenerate.
18481 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
18482 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
18483 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
18484 Add matching rules for -march=armv8.4-a and extensions.
18485 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
18487 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
18490 * config/rx/rx.md (BW): New mode attribute.
18491 (sync_lock_test_and_setsi): Add mode suffix to insn output.
18493 2018-01-11 Richard Biener <rguenther@suse.de>
18495 PR tree-optimization/83435
18496 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
18497 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
18498 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
18500 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18501 Alan Hayward <alan.hayward@arm.com>
18502 David Sherwood <david.sherwood@arm.com>
18504 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
18506 (aarch64_classify_address): Initialize it. Track polynomial offsets.
18507 (aarch64_print_address_internal): Use it to check for a zero offset.
18509 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18510 Alan Hayward <alan.hayward@arm.com>
18511 David Sherwood <david.sherwood@arm.com>
18513 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
18514 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
18515 Return a poly_int64 rather than a HOST_WIDE_INT.
18516 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
18517 rather than a HOST_WIDE_INT.
18518 * config/aarch64/aarch64.h (aarch64_frame): Protect with
18519 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
18520 hard_fp_offset, frame_size, initial_adjust, callee_offset and
18521 final_offset from HOST_WIDE_INT to poly_int64.
18522 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
18523 to_constant when getting the number of units in an Advanced SIMD
18525 (aarch64_builtin_vectorized_function): Check for a constant number
18527 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
18529 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
18530 attribute instead of GET_MODE_NUNITS.
18531 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
18532 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
18533 GET_MODE_SIZE for fixed-size registers.
18534 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
18535 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
18536 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
18537 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
18538 (aarch64_print_operand, aarch64_print_address_internal)
18539 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
18540 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
18541 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
18542 Handle polynomial GET_MODE_SIZE.
18543 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
18544 wider than SImode without modification.
18545 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
18546 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
18547 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
18548 passing and returning SVE modes.
18549 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
18550 rather than GEN_INT.
18551 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
18552 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
18553 (aarch64_allocate_and_probe_stack_space): Likewise.
18554 (aarch64_layout_frame): Cope with polynomial offsets.
18555 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
18556 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
18557 polynomial offsets.
18558 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
18559 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
18560 poly_int64 rather than a HOST_WIDE_INT.
18561 (aarch64_get_separate_components, aarch64_process_components)
18562 (aarch64_expand_prologue, aarch64_expand_epilogue)
18563 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
18564 (aarch64_anchor_offset): New function, split out from...
18565 (aarch64_legitimize_address): ...here.
18566 (aarch64_builtin_vectorization_cost): Handle polynomial
18567 TYPE_VECTOR_SUBPARTS.
18568 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
18570 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
18571 number of elements from the PARALLEL rather than the mode.
18572 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
18573 rather than GET_MODE_BITSIZE.
18574 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
18575 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
18576 (aarch64_expand_vec_perm_const_1): Handle polynomial
18577 d->perm.length () and d->perm elements.
18578 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
18579 Apply to_constant to d->perm elements.
18580 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
18581 polynomial CONST_VECTOR_NUNITS.
18582 (aarch64_move_pointer): Take amount as a poly_int64 rather
18584 (aarch64_progress_pointer): Avoid temporary variable.
18585 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
18586 the mode attribute instead of GET_MODE.
18588 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18589 Alan Hayward <alan.hayward@arm.com>
18590 David Sherwood <david.sherwood@arm.com>
18592 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
18593 x exists before using it.
18594 (aarch64_add_constant_internal): Rename to...
18595 (aarch64_add_offset_1): ...this. Replace regnum with separate
18596 src and dest rtxes. Handle the case in which they're different,
18597 including when the offset is zero. Replace scratchreg with an rtx.
18598 Use 2 additions if there is no spare register into which we can
18599 move a 16-bit constant.
18600 (aarch64_add_constant): Delete.
18601 (aarch64_add_offset): Replace reg with separate src and dest
18602 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
18603 Use aarch64_add_offset_1.
18604 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
18605 an rtx rather than an int. Take the delta as a poly_int64
18606 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
18607 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
18608 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
18609 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
18610 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
18611 and aarch64_add_sp.
18612 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
18613 aarch64_add_constant.
18615 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18617 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
18618 Use scalar_float_mode.
18620 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18622 * config/aarch64/aarch64-simd.md
18623 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
18624 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
18625 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
18626 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
18627 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
18628 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
18629 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
18630 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
18631 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
18632 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
18634 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18637 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
18638 targ_options->x_arm_arch_string is non NULL.
18640 2018-01-11 Tamar Christina <tamar.christina@arm.com>
18642 * config/aarch64/aarch64.h
18643 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
18645 2018-01-11 Sudakshina Das <sudi.das@arm.com>
18648 * expmed.c (emit_store_flag_force): Swap if const op0
18649 and change VOIDmode to mode of op0.
18651 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18653 PR rtl-optimization/83761
18654 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
18655 than bytes to mode_for_size.
18657 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
18659 PR middle-end/83189
18660 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
18661 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
18664 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
18666 PR middle-end/83575
18667 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
18668 when in layout mode.
18669 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
18670 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
18673 2018-01-10 Michael Collison <michael.collison@arm.com>
18675 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
18676 * config/aarch64/aarch64-option-extension.def: Add
18677 AARCH64_OPT_EXTENSION of 'fp16fml'.
18678 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18679 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
18680 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
18681 * config/aarch64/constraints.md (Ui7): New constraint.
18682 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
18683 (VFMLA_SEL_W): Ditto.
18686 (VFMLA16_LOW): New int iterator.
18687 (VFMLA16_HIGH): Ditto.
18688 (UNSPEC_FMLAL): New unspec.
18689 (UNSPEC_FMLSL): Ditto.
18690 (UNSPEC_FMLAL2): Ditto.
18691 (UNSPEC_FMLSL2): Ditto.
18692 (f16mac): New code attribute.
18693 * config/aarch64/aarch64-simd-builtins.def
18694 (aarch64_fmlal_lowv2sf): Ditto.
18695 (aarch64_fmlsl_lowv2sf): Ditto.
18696 (aarch64_fmlalq_lowv4sf): Ditto.
18697 (aarch64_fmlslq_lowv4sf): Ditto.
18698 (aarch64_fmlal_highv2sf): Ditto.
18699 (aarch64_fmlsl_highv2sf): Ditto.
18700 (aarch64_fmlalq_highv4sf): Ditto.
18701 (aarch64_fmlslq_highv4sf): Ditto.
18702 (aarch64_fmlal_lane_lowv2sf): Ditto.
18703 (aarch64_fmlsl_lane_lowv2sf): Ditto.
18704 (aarch64_fmlal_laneq_lowv2sf): Ditto.
18705 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
18706 (aarch64_fmlalq_lane_lowv4sf): Ditto.
18707 (aarch64_fmlsl_lane_lowv4sf): Ditto.
18708 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
18709 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
18710 (aarch64_fmlal_lane_highv2sf): Ditto.
18711 (aarch64_fmlsl_lane_highv2sf): Ditto.
18712 (aarch64_fmlal_laneq_highv2sf): Ditto.
18713 (aarch64_fmlsl_laneq_highv2sf): Ditto.
18714 (aarch64_fmlalq_lane_highv4sf): Ditto.
18715 (aarch64_fmlsl_lane_highv4sf): Ditto.
18716 (aarch64_fmlalq_laneq_highv4sf): Ditto.
18717 (aarch64_fmlsl_laneq_highv4sf): Ditto.
18718 * config/aarch64/aarch64-simd.md:
18719 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
18720 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
18721 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
18722 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
18723 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
18724 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
18725 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
18726 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
18727 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
18728 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
18729 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
18730 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
18731 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
18732 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
18733 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
18734 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
18735 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
18736 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
18737 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
18738 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
18739 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
18740 (vfmlsl_low_u32): Ditto.
18741 (vfmlalq_low_u32): Ditto.
18742 (vfmlslq_low_u32): Ditto.
18743 (vfmlal_high_u32): Ditto.
18744 (vfmlsl_high_u32): Ditto.
18745 (vfmlalq_high_u32): Ditto.
18746 (vfmlslq_high_u32): Ditto.
18747 (vfmlal_lane_low_u32): Ditto.
18748 (vfmlsl_lane_low_u32): Ditto.
18749 (vfmlal_laneq_low_u32): Ditto.
18750 (vfmlsl_laneq_low_u32): Ditto.
18751 (vfmlalq_lane_low_u32): Ditto.
18752 (vfmlslq_lane_low_u32): Ditto.
18753 (vfmlalq_laneq_low_u32): Ditto.
18754 (vfmlslq_laneq_low_u32): Ditto.
18755 (vfmlal_lane_high_u32): Ditto.
18756 (vfmlsl_lane_high_u32): Ditto.
18757 (vfmlal_laneq_high_u32): Ditto.
18758 (vfmlsl_laneq_high_u32): Ditto.
18759 (vfmlalq_lane_high_u32): Ditto.
18760 (vfmlslq_lane_high_u32): Ditto.
18761 (vfmlalq_laneq_high_u32): Ditto.
18762 (vfmlslq_laneq_high_u32): Ditto.
18763 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
18764 (AARCH64_FL_FOR_ARCH8_4): New.
18765 (AARCH64_ISA_F16FML): New ISA flag.
18766 (TARGET_F16FML): New feature flag for fp16fml.
18767 (doc/invoke.texi): Document new fp16fml option.
18769 2018-01-10 Michael Collison <michael.collison@arm.com>
18771 * config/aarch64/aarch64-builtins.c:
18772 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
18773 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18774 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
18775 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
18776 (AARCH64_ISA_SHA3): New ISA flag.
18777 (TARGET_SHA3): New feature flag for sha3.
18778 * config/aarch64/iterators.md (sha512_op): New int attribute.
18779 (CRYPTO_SHA512): New int iterator.
18780 (UNSPEC_SHA512H): New unspec.
18781 (UNSPEC_SHA512H2): Ditto.
18782 (UNSPEC_SHA512SU0): Ditto.
18783 (UNSPEC_SHA512SU1): Ditto.
18784 * config/aarch64/aarch64-simd-builtins.def
18785 (aarch64_crypto_sha512hqv2di): New builtin.
18786 (aarch64_crypto_sha512h2qv2di): Ditto.
18787 (aarch64_crypto_sha512su0qv2di): Ditto.
18788 (aarch64_crypto_sha512su1qv2di): Ditto.
18789 (aarch64_eor3qv8hi): Ditto.
18790 (aarch64_rax1qv2di): Ditto.
18791 (aarch64_xarqv2di): Ditto.
18792 (aarch64_bcaxqv8hi): Ditto.
18793 * config/aarch64/aarch64-simd.md:
18794 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
18795 (aarch64_crypto_sha512su0qv2di): Ditto.
18796 (aarch64_crypto_sha512su1qv2di): Ditto.
18797 (aarch64_eor3qv8hi): Ditto.
18798 (aarch64_rax1qv2di): Ditto.
18799 (aarch64_xarqv2di): Ditto.
18800 (aarch64_bcaxqv8hi): Ditto.
18801 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
18802 (vsha512h2q_u64): Ditto.
18803 (vsha512su0q_u64): Ditto.
18804 (vsha512su1q_u64): Ditto.
18805 (veor3q_u16): Ditto.
18806 (vrax1q_u64): Ditto.
18807 (vxarq_u64): Ditto.
18808 (vbcaxq_u16): Ditto.
18809 * config/arm/types.md (crypto_sha512): New type attribute.
18810 (crypto_sha3): Ditto.
18811 (doc/invoke.texi): Document new sha3 option.
18813 2018-01-10 Michael Collison <michael.collison@arm.com>
18815 * config/aarch64/aarch64-builtins.c:
18816 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
18817 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18818 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
18819 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
18820 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
18821 (AARCH64_ISA_SM4): New ISA flag.
18822 (TARGET_SM4): New feature flag for sm4.
18823 * config/aarch64/aarch64-simd-builtins.def
18824 (aarch64_sm3ss1qv4si): Ditto.
18825 (aarch64_sm3tt1aq4si): Ditto.
18826 (aarch64_sm3tt1bq4si): Ditto.
18827 (aarch64_sm3tt2aq4si): Ditto.
18828 (aarch64_sm3tt2bq4si): Ditto.
18829 (aarch64_sm3partw1qv4si): Ditto.
18830 (aarch64_sm3partw2qv4si): Ditto.
18831 (aarch64_sm4eqv4si): Ditto.
18832 (aarch64_sm4ekeyqv4si): Ditto.
18833 * config/aarch64/aarch64-simd.md:
18834 (aarch64_sm3ss1qv4si): Ditto.
18835 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
18836 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
18837 (aarch64_sm4eqv4si): Ditto.
18838 (aarch64_sm4ekeyqv4si): Ditto.
18839 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
18840 (sm3part_op): Ditto.
18841 (CRYPTO_SM3TT): Ditto.
18842 (CRYPTO_SM3PART): Ditto.
18843 (UNSPEC_SM3SS1): New unspec.
18844 (UNSPEC_SM3TT1A): Ditto.
18845 (UNSPEC_SM3TT1B): Ditto.
18846 (UNSPEC_SM3TT2A): Ditto.
18847 (UNSPEC_SM3TT2B): Ditto.
18848 (UNSPEC_SM3PARTW1): Ditto.
18849 (UNSPEC_SM3PARTW2): Ditto.
18850 (UNSPEC_SM4E): Ditto.
18851 (UNSPEC_SM4EKEY): Ditto.
18852 * config/aarch64/constraints.md (Ui2): New constraint.
18853 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
18854 * config/arm/types.md (crypto_sm3): New type attribute.
18855 (crypto_sm4): Ditto.
18856 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
18857 (vsm3tt1aq_u32): Ditto.
18858 (vsm3tt1bq_u32): Ditto.
18859 (vsm3tt2aq_u32): Ditto.
18860 (vsm3tt2bq_u32): Ditto.
18861 (vsm3partw1q_u32): Ditto.
18862 (vsm3partw2q_u32): Ditto.
18863 (vsm4eq_u32): Ditto.
18864 (vsm4ekeyq_u32): Ditto.
18865 (doc/invoke.texi): Document new sm4 option.
18867 2018-01-10 Michael Collison <michael.collison@arm.com>
18869 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
18870 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
18871 (AARCH64_FL_FOR_ARCH8_4): New.
18872 (AARCH64_FL_V8_4): New flag.
18873 (doc/invoke.texi): Document new armv8.4-a option.
18875 2018-01-10 Michael Collison <michael.collison@arm.com>
18877 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18878 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
18879 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
18880 * config/aarch64/aarch64-option-extension.def: Add
18881 AARCH64_OPT_EXTENSION of 'sha2'.
18882 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
18883 (crypto): Disable sha2 and aes if crypto disabled.
18884 (crypto): Enable aes and sha2 if enabled.
18885 (simd): Disable sha2 and aes if simd disabled.
18886 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
18888 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
18889 (TARGET_SHA2): New feature flag for sha2.
18890 (TARGET_AES): New feature flag for aes.
18891 * config/aarch64/aarch64-simd.md:
18892 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
18893 conditional on TARGET_AES.
18894 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
18895 (aarch64_crypto_sha1hsi): Make pattern conditional
18897 (aarch64_crypto_sha1hv4si): Ditto.
18898 (aarch64_be_crypto_sha1hv4si): Ditto.
18899 (aarch64_crypto_sha1su1v4si): Ditto.
18900 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
18901 (aarch64_crypto_sha1su0v4si): Ditto.
18902 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
18903 (aarch64_crypto_sha256su0v4si): Ditto.
18904 (aarch64_crypto_sha256su1v4si): Ditto.
18905 (doc/invoke.texi): Document new aes and sha2 options.
18907 2018-01-10 Martin Sebor <msebor@redhat.com>
18909 PR tree-optimization/83781
18910 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
18913 2018-01-11 Martin Sebor <msebor@gmail.com>
18914 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18916 PR tree-optimization/83501
18917 PR tree-optimization/81703
18919 * tree-ssa-strlen.c (get_string_cst): Rename...
18920 (get_string_len): ...to this. Handle global constants.
18921 (handle_char_store): Adjust.
18923 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
18924 Jim Wilson <jimw@sifive.com>
18926 * config/riscv/riscv-protos.h (riscv_output_return): New.
18927 * config/riscv/riscv.c (struct machine_function): New naked_p field.
18928 (riscv_attribute_table, riscv_output_return),
18929 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
18930 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
18931 (riscv_compute_frame_info): Only compute frame->mask if not a naked
18933 (riscv_expand_prologue): Add early return for naked function.
18934 (riscv_expand_epilogue): Likewise.
18935 (riscv_function_ok_for_sibcall): Return false for naked function.
18936 (riscv_set_current_function): New.
18937 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
18938 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
18939 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
18940 * doc/extend.texi (RISC-V Function Attributes): New.
18942 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
18944 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
18945 check for 128-bit long double before checking TCmode.
18946 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
18947 128-bit long doubles before checking TFmode or TCmode.
18948 (FLOAT128_IBM_P): Likewise.
18950 2018-01-10 Martin Sebor <msebor@redhat.com>
18952 PR tree-optimization/83671
18953 * builtins.c (c_strlen): Unconditionally return zero for the empty
18955 Use -Warray-bounds for warnings.
18956 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
18957 for non-constant array indices with COMPONENT_REF, arrays of
18958 arrays, and pointers to arrays.
18959 (gimple_fold_builtin_strlen): Determine and set length range for
18960 non-constant character arrays.
18962 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
18964 PR middle-end/81897
18965 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
18968 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
18970 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
18972 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
18975 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
18976 VECTOR_MEM_ALTIVEC_OR_VSX_P.
18977 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
18978 indexed_or_indirect_operand predicate.
18979 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
18980 (*vsx_le_perm_load_v8hi): Likewise.
18981 (*vsx_le_perm_load_v16qi): Likewise.
18982 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
18983 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
18984 (*vsx_le_perm_store_v8hi): Likewise.
18985 (*vsx_le_perm_store_v16qi): Likewise.
18986 (eight unnamed splitters): Likewise.
18988 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
18990 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
18991 * config/rs6000/emmintrin.h: Likewise.
18992 * config/rs6000/mmintrin.h: Likewise.
18993 * config/rs6000/xmmintrin.h: Likewise.
18995 2018-01-10 David Malcolm <dmalcolm@redhat.com>
18998 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
19000 * tree.c (tree_nop_conversion): Return true for location wrapper
19002 (maybe_wrap_with_location): New function.
19003 (selftest::check_strip_nops): New function.
19004 (selftest::test_location_wrappers): New function.
19005 (selftest::tree_c_tests): Call it.
19006 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
19007 (maybe_wrap_with_location): New decl.
19008 (EXPR_LOCATION_WRAPPER_P): New macro.
19009 (location_wrapper_p): New inline function.
19010 (tree_strip_any_location_wrapper): New inline function.
19012 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
19015 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
19016 stack_realign_offset for the largest alignment of stack slot
19018 (ix86_find_max_used_stack_alignment): New function.
19019 (ix86_finalize_stack_frame_flags): Use it. Set
19020 max_used_stack_alignment if we don't realign stack.
19021 * config/i386/i386.h (machine_function): Add
19022 max_used_stack_alignment.
19024 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
19026 * config/arm/arm.opt (-mbranch-cost): New option.
19027 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
19030 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
19033 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
19034 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
19036 2018-01-10 Richard Biener <rguenther@suse.de>
19039 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
19040 early out so it also covers the case where we have a non-NULL
19043 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
19045 PR tree-optimization/83753
19046 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
19047 for non-strided grouped accesses if the number of elements is 1.
19049 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
19052 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
19053 * i386.h (TARGET_USE_GATHER): Define.
19054 * x86-tune.def (X86_TUNE_USE_GATHER): New.
19056 2018-01-10 Martin Liska <mliska@suse.cz>
19059 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
19060 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
19062 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
19063 CLEANUP_NO_PARTITIONING is not set.
19065 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
19067 * doc/rtl.texi: Remove documentation of (const ...) wrappers
19068 for vectors, as a partial revert of r254296.
19069 * rtl.h (const_vec_p): Delete.
19070 (const_vec_duplicate_p): Don't test for vector CONSTs.
19071 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
19072 * expmed.c (make_tree): Likewise.
19075 * common.md (E, F): Use CONSTANT_P instead of checking for
19077 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
19078 checking for CONST_VECTOR.
19080 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
19082 PR middle-end/83575
19083 * predict.c (force_edge_cold): Handle in more sane way edges
19084 with no prediction.
19086 2018-01-09 Carl Love <cel@us.ibm.com>
19088 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
19090 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
19091 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
19092 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
19093 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
19094 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
19095 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
19096 * config/rs6000/rs6000-protos.h: Add extern defition for
19097 rs6000_generate_float2_double_code.
19098 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
19100 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
19101 (float2_v2df): Add define_expand.
19103 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
19106 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
19107 op_mode in the force_to_mode call.
19109 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
19111 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
19112 instead of checking each element individually.
19113 (aarch64_evpc_uzp): Likewise.
19114 (aarch64_evpc_zip): Likewise.
19115 (aarch64_evpc_ext): Likewise.
19116 (aarch64_evpc_rev): Likewise.
19117 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
19118 instead of checking each element individually. Return true without
19120 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
19121 whether all selected elements come from the same input, instead of
19122 checking each element individually. Remove calls to gen_rtx_REG,
19123 start_sequence and end_sequence and instead assert that no rtl is
19126 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
19128 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
19129 order of HIGH and CONST checks.
19131 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
19133 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
19134 if the destination isn't an SSA_NAME.
19136 2018-01-09 Richard Biener <rguenther@suse.de>
19138 PR tree-optimization/83668
19139 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
19141 (canonicalize_loop_form): ... here, renamed from ...
19142 (canonicalize_loop_closed_ssa_form): ... this and amended to
19143 swap successor edges for loop exit blocks to make us use
19144 the RPO order we need for initial schedule generation.
19146 2018-01-09 Joseph Myers <joseph@codesourcery.com>
19148 PR tree-optimization/64811
19149 * match.pd: When optimizing comparisons with Inf, avoid
19150 introducing or losing exceptions from comparisons with NaN.
19152 2018-01-09 Martin Liska <mliska@suse.cz>
19155 * asan.c (shadow_mem_size): Add gcc_assert.
19157 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
19159 Don't save registers in main().
19162 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
19163 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
19164 * config/avr/avr.c (avr_set_current_function): Don't error if
19165 naked, OS_task or OS_main are specified at the same time.
19166 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
19168 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
19170 * common/config/avr/avr-common.c (avr_option_optimization_table):
19171 Switch on -mmain-is-OS_task for optimizing compilations.
19173 2018-01-09 Richard Biener <rguenther@suse.de>
19175 PR tree-optimization/83572
19176 * graphite.c: Include cfganal.h.
19177 (graphite_transform_loops): Connect infinite loops to exit
19178 and remove fake edges at the end.
19180 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
19182 * ipa-inline.c (edge_badness): Revert accidental checkin.
19184 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
19187 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
19188 symbols; not inline clones.
19190 2018-01-09 Jakub Jelinek <jakub@redhat.com>
19193 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
19194 hard registers. Formatting fixes.
19196 PR preprocessor/83722
19197 * gcc.c (try_generate_repro): Pass
19198 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
19199 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
19202 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
19203 Kito Cheng <kito.cheng@gmail.com>
19205 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
19206 (riscv_leaf_function_p): Delete.
19207 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
19209 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19211 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
19213 (do_ifelse): New function.
19214 (do_isel): New function.
19215 (do_sub3): New function.
19216 (do_add3): New function.
19217 (do_load_mask_compare): New function.
19218 (do_overlap_load_compare): New function.
19219 (expand_compare_loop): New function.
19220 (expand_block_compare): Call expand_compare_loop() when appropriate.
19221 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
19222 option description.
19223 (-mblock-compare-inline-loop-limit): New option.
19225 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19228 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
19229 Reverse order of second and third operands in first alternative.
19230 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
19231 of first and second elements in UNSPEC_VPERMR vector.
19232 (altivec_expand_vec_perm_le): Likewise.
19234 2018-01-08 Jeff Law <law@redhat.com>
19236 PR rtl-optimizatin/81308
19237 * tree-switch-conversion.c (cfg_altered): New file scoped static.
19238 (process_switch): If group_case_labels makes a change, then set
19240 (pass_convert_switch::execute): If a switch is converted, then
19241 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
19243 PR rtl-optimization/81308
19244 * recog.c (split_all_insns): Conditionally cleanup the CFG after
19247 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
19249 PR target/83663 - Revert r255946
19250 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
19251 generation for cases where splatting a value is not useful.
19252 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
19253 across a vec_duplicate and a paradoxical subreg forming a vector
19254 mode to a vec_concat.
19256 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19258 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
19259 -march=armv8.3-a variants.
19260 * config/arm/t-multilib: Likewise.
19261 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
19263 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19265 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
19267 (cceq_ior_compare_complement): Give it a name so I can use it, and
19268 change boolean_or_operator predicate to boolean_operator so it can
19269 be used to generate a crand.
19270 (eqne): New code iterator.
19271 (bd/bd_neg): New code_attrs.
19272 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
19273 a single define_insn.
19274 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
19275 decrement (bdnzt/bdnzf/bdzt/bdzf).
19276 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
19277 with the new names of the branch decrement patterns, and added the
19278 names of the branch decrement conditional patterns.
19280 2018-01-08 Richard Biener <rguenther@suse.de>
19282 PR tree-optimization/83563
19283 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
19286 2018-01-08 Richard Biener <rguenther@suse.de>
19288 PR middle-end/83713
19289 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
19291 2018-01-08 Richard Biener <rguenther@suse.de>
19293 PR tree-optimization/83685
19294 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
19295 references to abnormals.
19297 2018-01-08 Richard Biener <rguenther@suse.de>
19300 * dwarf2out.c (output_indirect_strings): Handle empty
19301 skeleton_debug_str_hash.
19302 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
19304 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
19306 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
19307 (emit_store_direct): Likewise.
19308 (arc_trampoline_adjust_address): Likewise.
19309 (arc_asm_trampoline_template): New function.
19310 (arc_initialize_trampoline): Use asm_trampoline_template.
19311 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
19312 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
19313 * config/arc/arc.md (flush_icache): Delete pattern.
19315 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
19317 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
19318 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
19321 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
19324 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
19325 by not USED_FOR_TARGET.
19326 (make_pass_resolve_sw_modes): Likewise.
19328 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
19330 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
19333 2018-01-08 Richard Biener <rguenther@suse.de>
19335 PR middle-end/83580
19336 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
19338 2018-01-08 Richard Biener <rguenther@suse.de>
19340 PR middle-end/83517
19341 * match.pd ((t * 2) / 2) -> t): Add missing :c.
19343 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
19345 PR middle-end/81897
19346 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
19347 basic blocks with a small number of successors.
19348 (convert_control_dep_chain_into_preds): Improve handling of
19350 (dump_predicates): Split apart into...
19351 (dump_pred_chain): ...here...
19352 (dump_pred_info): ...and here.
19353 (can_one_predicate_be_invalidated_p): Add debugging printfs.
19354 (can_chain_union_be_invalidated_p): Improve check for invalidation
19356 (uninit_uses_cannot_happen): Avoid unnecessary if
19357 convert_control_dep_chain_into_preds yielded nothing.
19359 2018-01-06 Martin Sebor <msebor@redhat.com>
19361 PR tree-optimization/83640
19362 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
19363 subtracting negative offset from size.
19364 (builtin_access::overlap): Adjust offset bounds of the access to fall
19365 within the size of the object if possible.
19367 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
19369 PR rtl-optimization/83699
19370 * expmed.c (extract_bit_field_1): Restrict the vector usage of
19371 extract_bit_field_as_subreg to cases in which the extracted
19372 value is also a vector.
19374 * lra-constraints.c (process_alt_operands): Test for the equivalence
19375 substitutions when detecting a possible reload cycle.
19377 2018-01-06 Jakub Jelinek <jakub@redhat.com>
19380 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
19381 by default if flag_selective_schedling{,2}. Formatting fixes.
19383 PR rtl-optimization/83682
19384 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
19385 if it has non-VECTOR_MODE element mode.
19386 (vec_duplicate_p): Likewise.
19388 PR middle-end/83694
19389 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
19390 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
19392 2018-01-05 Jakub Jelinek <jakub@redhat.com>
19395 * config/i386/i386-builtin.def
19396 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
19397 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
19398 Require also OPTION_MASK_ISA_AVX512F in addition to
19399 OPTION_MASK_ISA_GFNI.
19400 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
19401 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
19402 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
19403 to OPTION_MASK_ISA_GFNI.
19404 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
19405 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
19406 OPTION_MASK_ISA_AVX512BW.
19407 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
19408 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
19409 addition to OPTION_MASK_ISA_GFNI.
19410 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
19411 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
19412 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
19413 to OPTION_MASK_ISA_GFNI.
19414 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
19415 a requirement for all ISAs rather than any of them with a few
19417 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
19419 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
19420 bitmasks to be enabled with 3 exceptions, instead of requiring any
19421 enabled ISA with lots of exceptions.
19422 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
19423 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
19424 Change avx512bw in isa attribute to avx512f.
19425 * config/i386/sgxintrin.h: Add license boilerplate.
19426 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
19427 to __AVX512F__ and __AVX512VL to __AVX512VL__.
19428 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
19429 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
19431 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
19432 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
19433 temporarily sse2 rather than sse if not enabled already.
19436 * config/i386/sse.md (VI248_VLBW): Rename to ...
19437 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
19438 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
19439 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
19440 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
19441 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
19442 mode iterator instead of VI248_VLBW.
19444 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
19446 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
19447 (record_modified): Skip clobbers; add debug output.
19448 (param_change_prob): Use sreal frequencies.
19450 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
19452 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
19453 punt for user-aligned variables.
19455 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
19457 * tree-chrec.c (chrec_contains_symbols): Return true for
19460 2018-01-05 Sudakshina Das <sudi.das@arm.com>
19463 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
19464 of (x|y) == x for BICS pattern.
19466 2018-01-05 Jakub Jelinek <jakub@redhat.com>
19468 PR tree-optimization/83605
19469 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
19470 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
19473 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
19475 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
19476 * config/epiphany/rtems.h: New file.
19478 2018-01-04 Jakub Jelinek <jakub@redhat.com>
19479 Uros Bizjak <ubizjak@gmail.com>
19482 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
19483 QIreg_operand instead of register_operand predicate.
19484 * config/i386/i386.c (ix86_rop_should_change_byte_p,
19485 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
19486 comments instead of -fmitigate[-_]rop.
19488 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19491 * cgraphunit.c (symbol_table::compile): Switch to text_section
19492 before calling assembly_start debug hook.
19493 * run-rtl-passes.c (run_rtl_passes): Likewise.
19496 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19498 * tree-vrp.c (extract_range_from_binary_expr_1): Check
19499 range_int_cst_p rather than !symbolic_range_p before calling
19500 extract_range_from_multiplicative_op_1.
19502 2018-01-04 Jeff Law <law@redhat.com>
19504 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
19505 redundant test in assertion.
19507 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19509 * doc/rtl.texi: Document machine_mode wrapper classes.
19511 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19513 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
19514 using tree_to_uhwi.
19516 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19518 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
19519 the VEC_PERM_EXPR fold to fail.
19521 2018-01-04 Jakub Jelinek <jakub@redhat.com>
19524 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
19525 to switched_sections.
19527 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19530 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
19531 test for d.testing.
19533 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
19536 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
19537 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
19539 2018-01-04 Jakub Jelinek <jakub@redhat.com>
19542 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
19543 is BLKmode and bitpos not zero or mode change is needed.
19545 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19548 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
19551 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
19554 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
19555 instead of MULT rtx. Update all corresponding splitters.
19556 (*saddl_se): Ditto.
19557 (*ssub<modesuffix>): Ditto.
19558 (*ssubl_se): Ditto.
19559 (*cmp_sadd_di): Update split patterns.
19560 (*cmp_sadd_si): Ditto.
19561 (*cmp_sadd_sidi): Ditto.
19562 (*cmp_ssub_di): Ditto.
19563 (*cmp_ssub_si): Ditto.
19564 (*cmp_ssub_sidi): Ditto.
19565 * config/alpha/predicates.md (const23_operand): New predicate.
19566 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
19567 Look for ASHIFT, not MULT inner operand.
19568 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
19570 2018-01-04 Martin Liska <mliska@suse.cz>
19572 PR gcov-profile/83669
19573 * gcov.c (output_intermediate_file): Add version to intermediate
19575 * doc/gcov.texi: Document new field 'version' in intermediate
19576 file format. Fix location of '-k' option of gcov command.
19578 2018-01-04 Martin Liska <mliska@suse.cz>
19581 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
19583 2018-01-04 Jakub Jelinek <jakub@redhat.com>
19585 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
19587 2018-01-03 Martin Sebor <msebor@redhat.com>
19589 PR tree-optimization/83655
19590 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
19591 checking calls with invalid arguments.
19593 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19595 * tree-vect-stmts.c (vect_get_store_rhs): New function.
19596 (vectorizable_mask_load_store): Delete.
19597 (vectorizable_call): Return false for masked loads and stores.
19598 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
19599 instead of gimple_assign_rhs1.
19600 (vectorizable_load): Handle IFN_MASK_LOAD.
19601 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
19603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19605 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
19607 (vectorizable_mask_load_store): ...here.
19608 (vectorizable_load): ...and here.
19610 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19612 * tree-vect-stmts.c (vect_build_all_ones_mask)
19613 (vect_build_zero_merge_argument): New functions, split out from...
19614 (vectorizable_load): ...here.
19616 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19618 * tree-vect-stmts.c (vect_check_store_rhs): New function,
19620 (vectorizable_mask_load_store): ...here.
19621 (vectorizable_store): ...and here.
19623 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19625 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
19627 (vectorizable_mask_load_store): ...here.
19629 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19631 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
19632 (vect_model_store_cost): Take a vec_load_store_type instead of a
19634 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
19635 (vect_model_store_cost): Take a vec_load_store_type instead of a
19637 (vectorizable_mask_load_store): Update accordingly.
19638 (vectorizable_store): Likewise.
19639 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
19641 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19643 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
19644 IFN_MASK_LOAD calls here rather than...
19645 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
19647 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19648 Alan Hayward <alan.hayward@arm.com>
19649 David Sherwood <david.sherwood@arm.com>
19651 * expmed.c (extract_bit_field_1): For vector extracts,
19652 fall back to extract_bit_field_as_subreg if vec_extract
19655 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19656 Alan Hayward <alan.hayward@arm.com>
19657 David Sherwood <david.sherwood@arm.com>
19659 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
19660 they are variable or constant sized.
19661 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
19662 slots for constant-sized data.
19664 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19665 Alan Hayward <alan.hayward@arm.com>
19666 David Sherwood <david.sherwood@arm.com>
19668 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
19669 handling COND_EXPRs with boolean comparisons, try to find a better
19670 basis for the mask type than the boolean itself.
19672 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19674 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
19675 is calculated and how it can be overridden.
19676 * genmodes.c (max_bitsize_mode_any_mode): New variable.
19677 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
19679 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
19682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19683 Alan Hayward <alan.hayward@arm.com>
19684 David Sherwood <david.sherwood@arm.com>
19686 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
19687 Remove the mode argument.
19688 (aarch64_simd_valid_immediate): Remove the mode and inverse
19690 * config/aarch64/iterators.md (bitsize): New iterator.
19691 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
19692 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
19693 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
19694 aarch64_simd_valid_immediate.
19695 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
19696 (aarch64_reg_or_bic_imm): Likewise.
19697 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
19698 with an insn_type enum and msl with a modifier_type enum.
19699 Replace element_width with a scalar_mode. Change the shift
19700 to unsigned int. Add constructors for scalar_float_mode and
19701 scalar_int_mode elements.
19702 (aarch64_vect_float_const_representable_p): Delete.
19703 (aarch64_can_const_movi_rtx_p)
19704 (aarch64_simd_scalar_immediate_valid_for_move)
19705 (aarch64_simd_make_constant): Update call to
19706 aarch64_simd_valid_immediate.
19707 (aarch64_advsimd_valid_immediate_hs): New function.
19708 (aarch64_advsimd_valid_immediate): Likewise.
19709 (aarch64_simd_valid_immediate): Remove mode and inverse
19710 arguments. Rewrite to use the above. Use const_vec_duplicate_p
19711 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
19712 and aarch64_float_const_representable_p on the result.
19713 (aarch64_output_simd_mov_immediate): Remove mode argument.
19714 Update call to aarch64_simd_valid_immediate and use of
19715 simd_immediate_info.
19716 (aarch64_output_scalar_simd_mov_immediate): Update call
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 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
19724 (mode_nunits): Likewise CONST_MODE_NUNITS.
19725 * machmode.def (ADJUST_NUNITS): Document.
19726 * genmodes.c (mode_data::need_nunits_adj): New field.
19727 (blank_mode): Update accordingly.
19728 (adj_nunits): New variable.
19729 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
19731 (emit_mode_size_inline): Set need_bytesize_adj for all modes
19732 listed in adj_nunits.
19733 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
19734 listed in adj_nunits. Don't emit case statements for such modes.
19735 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
19736 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
19737 nothing if adj_nunits is nonnull.
19738 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
19739 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
19740 (emit_mode_fbit): Update use of print_maybe_const_decl.
19741 (emit_move_size): Likewise. Treat the array as non-const
19743 (emit_mode_adjustments): Handle adj_nunits.
19745 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19747 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
19748 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
19749 (VECTOR_MODES): Use it.
19750 (make_vector_modes): Take the prefix as an argument.
19752 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19753 Alan Hayward <alan.hayward@arm.com>
19754 David Sherwood <david.sherwood@arm.com>
19756 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
19757 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
19758 for MODE_VECTOR_BOOL.
19759 * machmode.def (VECTOR_BOOL_MODE): Document.
19760 * genmodes.c (VECTOR_BOOL_MODE): New macro.
19761 (make_vector_bool_mode): New function.
19762 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
19764 * lto-streamer-in.c (lto_input_mode_table): Likewise.
19765 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
19767 * stor-layout.c (int_mode_for_mode): Likewise.
19768 * tree.c (build_vector_type_for_mode): Likewise.
19769 * varasm.c (output_constant_pool_2): Likewise.
19770 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
19771 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
19772 for MODE_VECTOR_BOOL.
19773 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
19774 of mode class checks.
19775 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
19776 instead of a list of mode class checks.
19777 (expand_vector_scalar_condition): Likewise.
19778 (type_for_widest_vector_mode): Handle BImode as an inner mode.
19780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19781 Alan Hayward <alan.hayward@arm.com>
19782 David Sherwood <david.sherwood@arm.com>
19784 * machmode.h (mode_size): Change from unsigned short to
19786 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
19787 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
19788 or if measurement_type is not polynomial.
19789 (fixed_size_mode::includes_p): Check for constant-sized modes.
19790 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
19791 return a poly_uint16 rather than an unsigned short.
19792 (emit_mode_size): Change the type of mode_size from unsigned short
19793 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
19794 (emit_mode_adjustments): Cope with polynomial vector sizes.
19795 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
19797 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
19799 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
19800 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
19801 * caller-save.c (setup_save_areas): Likewise.
19802 (replace_reg_with_saved_mem): Likewise.
19803 * calls.c (emit_library_call_value_1): Likewise.
19804 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
19805 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
19806 (gen_lowpart_for_combine): Likewise.
19807 * convert.c (convert_to_integer_1): Likewise.
19808 * cse.c (equiv_constant, cse_insn): Likewise.
19809 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
19810 (cselib_subst_to_values): Likewise.
19811 * dce.c (word_dce_process_block): Likewise.
19812 * df-problems.c (df_word_lr_mark_ref): Likewise.
19813 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
19814 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
19815 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
19816 (rtl_for_decl_location): Likewise.
19817 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
19818 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
19819 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
19820 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
19821 (expand_expr_real_1): Likewise.
19822 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
19823 (pad_below): Likewise.
19824 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
19825 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
19826 * ira.c (get_subreg_tracking_sizes): Likewise.
19827 * ira-build.c (ira_create_allocno_objects): Likewise.
19828 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
19829 (ira_sort_regnos_for_alter_reg): Likewise.
19830 * ira-costs.c (record_operand_costs): Likewise.
19831 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
19832 (resolve_simple_move): Likewise.
19833 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
19834 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
19835 (lra_constraints): Likewise.
19836 (CONST_POOL_OK_P): Reject variable-sized modes.
19837 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
19838 (add_pseudo_to_slot, lra_spill): Likewise.
19839 * omp-low.c (omp_clause_aligned_alignment): Likewise.
19840 * optabs-query.c (get_best_extraction_insn): Likewise.
19841 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
19842 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
19843 (expand_mult_highpart, valid_multiword_target_p): Likewise.
19844 * recog.c (offsettable_address_addr_space_p): Likewise.
19845 * regcprop.c (maybe_mode_change): Likewise.
19846 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
19847 * regrename.c (build_def_use): Likewise.
19848 * regstat.c (dump_reg_info): Likewise.
19849 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
19850 (find_reloads, find_reloads_subreg_address): Likewise.
19851 * reload1.c (eliminate_regs_1): Likewise.
19852 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
19853 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
19854 (simplify_binary_operation_1, simplify_subreg): Likewise.
19855 * targhooks.c (default_function_arg_padding): Likewise.
19856 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
19857 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
19858 (verify_gimple_assign_ternary): Likewise.
19859 * tree-inline.c (estimate_move_cost): Likewise.
19860 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19861 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
19862 (get_address_cost_ainc): Likewise.
19863 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
19864 (vect_supportable_dr_alignment): Likewise.
19865 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
19866 (vectorizable_reduction): Likewise.
19867 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
19868 (vectorizable_operation, vectorizable_load): Likewise.
19869 * tree.c (build_same_sized_truth_vector_type): Likewise.
19870 * valtrack.c (cleanup_auto_inc_dec): Likewise.
19871 * var-tracking.c (emit_note_insn_var_location): Likewise.
19872 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
19873 (ADDR_VEC_ALIGN): Likewise.
19875 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19876 Alan Hayward <alan.hayward@arm.com>
19877 David Sherwood <david.sherwood@arm.com>
19879 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
19881 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
19882 or if measurement_type is polynomial.
19883 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
19884 * combine.c (make_extraction): Likewise.
19885 * dse.c (find_shift_sequence): Likewise.
19886 * dwarf2out.c (mem_loc_descriptor): Likewise.
19887 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
19888 (extract_bit_field, extract_low_bits): Likewise.
19889 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
19890 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
19891 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
19892 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
19893 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
19894 * reload.c (find_reloads): Likewise.
19895 * reload1.c (alter_reg): Likewise.
19896 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
19897 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
19898 * tree-if-conv.c (predicate_mem_writes): Likewise.
19899 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
19900 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
19901 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
19902 * valtrack.c (dead_debug_insert_temp): Likewise.
19903 * varasm.c (mergeable_constant_section): Likewise.
19904 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
19906 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19907 Alan Hayward <alan.hayward@arm.com>
19908 David Sherwood <david.sherwood@arm.com>
19910 * expr.c (expand_assignment): Cope with polynomial mode sizes
19911 when assigning to a CONCAT.
19913 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19914 Alan Hayward <alan.hayward@arm.com>
19915 David Sherwood <david.sherwood@arm.com>
19917 * machmode.h (mode_precision): Change from unsigned short to
19919 (mode_to_precision): Return a poly_uint16 rather than an unsigned
19921 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
19922 or if measurement_type is not polynomial.
19923 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
19924 in which the mode is already known to be a scalar_int_mode.
19925 * genmodes.c (emit_mode_precision): Change the type of mode_precision
19926 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
19928 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
19929 for GET_MODE_PRECISION.
19930 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
19931 for GET_MODE_PRECISION.
19932 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
19934 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
19935 (expand_field_assignment, make_extraction): Likewise.
19936 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
19937 (get_last_value): Likewise.
19938 * convert.c (convert_to_integer_1): Likewise.
19939 * cse.c (cse_insn): Likewise.
19940 * expr.c (expand_expr_real_1): Likewise.
19941 * lra-constraints.c (simplify_operand_subreg): Likewise.
19942 * optabs-query.c (can_atomic_load_p): Likewise.
19943 * optabs.c (expand_atomic_load): Likewise.
19944 (expand_atomic_store): Likewise.
19945 * ree.c (combine_reaching_defs): Likewise.
19946 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
19947 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
19948 * tree.h (type_has_mode_precision_p): Likewise.
19949 * ubsan.c (instrument_si_overflow): Likewise.
19951 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19952 Alan Hayward <alan.hayward@arm.com>
19953 David Sherwood <david.sherwood@arm.com>
19955 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
19956 polynomial numbers of units.
19957 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
19958 (valid_vector_subparts_p): New function.
19959 (build_vector_type): Remove temporary shim and take the number
19960 of units as a poly_uint64 rather than an int.
19961 (build_opaque_vector_type): Take the number of units as a
19962 poly_uint64 rather than an int.
19963 * tree.c (build_vector_from_ctor): Handle polynomial
19964 TYPE_VECTOR_SUBPARTS.
19965 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
19966 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
19967 (build_vector_from_val): If the number of units is variable,
19968 use build_vec_duplicate_cst for constant operands and
19969 VEC_DUPLICATE_EXPR otherwise.
19970 (make_vector_type): Remove temporary is_constant ().
19971 (build_vector_type, build_opaque_vector_type): Take the number of
19972 units as a poly_uint64 rather than an int.
19973 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
19975 * cfgexpand.c (expand_debug_expr): Likewise.
19976 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
19977 (store_constructor, expand_expr_real_1): Likewise.
19978 (const_scalar_mask_from_tree): Likewise.
19979 * fold-const-call.c (fold_const_reduction): Likewise.
19980 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
19981 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
19982 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
19983 (fold_relational_const): Likewise.
19984 (native_interpret_vector): Likewise. Change the size from an
19985 int to an unsigned int.
19986 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
19987 TYPE_VECTOR_SUBPARTS.
19988 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
19989 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
19990 duplicating a non-constant operand into a variable-length vector.
19991 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
19992 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
19993 * ipa-icf.c (sem_variable::equals): Likewise.
19994 * match.pd: Likewise.
19995 * omp-simd-clone.c (simd_clone_subparts): Likewise.
19996 * print-tree.c (print_node): Likewise.
19997 * stor-layout.c (layout_type): Likewise.
19998 * targhooks.c (default_builtin_vectorization_cost): Likewise.
19999 * tree-cfg.c (verify_gimple_comparison): Likewise.
20000 (verify_gimple_assign_binary): Likewise.
20001 (verify_gimple_assign_ternary): Likewise.
20002 (verify_gimple_assign_single): Likewise.
20003 * tree-pretty-print.c (dump_generic_node): Likewise.
20004 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
20005 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
20006 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
20007 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
20008 (vect_shift_permute_load_chain): Likewise.
20009 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
20010 (expand_vector_condition, optimize_vector_constructor): Likewise.
20011 (lower_vec_perm, get_compute_type): Likewise.
20012 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
20013 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
20014 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
20015 (vect_recog_mask_conversion_pattern): Likewise.
20016 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
20017 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
20018 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20019 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
20020 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
20021 (vectorizable_shift, vectorizable_operation, vectorizable_store)
20022 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
20023 (supportable_widening_operation): Likewise.
20024 (supportable_narrowing_operation): Likewise.
20025 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
20027 * varasm.c (output_constant): Likewise.
20029 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20030 Alan Hayward <alan.hayward@arm.com>
20031 David Sherwood <david.sherwood@arm.com>
20033 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
20034 so that both the length == 3 and length != 3 cases set up their
20035 own permute vectors. Add comments explaining why we know the
20036 number of elements is constant.
20037 (vect_permute_load_chain): Likewise.
20039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20040 Alan Hayward <alan.hayward@arm.com>
20041 David Sherwood <david.sherwood@arm.com>
20043 * machmode.h (mode_nunits): Change from unsigned char to
20045 (ONLY_FIXED_SIZE_MODES): New macro.
20046 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
20047 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
20048 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
20050 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
20051 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
20052 or if measurement_type is not polynomial.
20053 * genmodes.c (ZERO_COEFFS): New macro.
20054 (emit_mode_nunits_inline): Make mode_nunits_inline return a
20056 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
20057 Use ZERO_COEFFS when emitting initializers.
20058 * data-streamer.h (bp_pack_poly_value): New function.
20059 (bp_unpack_poly_value): Likewise.
20060 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
20061 for GET_MODE_NUNITS.
20062 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
20063 for GET_MODE_NUNITS.
20064 * tree.c (make_vector_type): Remove temporary shim and make
20065 the real function take the number of units as a poly_uint64
20066 rather than an int.
20067 (build_vector_type_for_mode): Handle polynomial nunits.
20068 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
20069 * emit-rtl.c (const_vec_series_p_1): Likewise.
20070 (gen_rtx_CONST_VECTOR): Likewise.
20071 * fold-const.c (test_vec_duplicate_folding): Likewise.
20072 * genrecog.c (validate_pattern): Likewise.
20073 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
20074 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
20075 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
20076 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
20077 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
20078 * rtlanal.c (subreg_get_info): Likewise.
20079 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
20080 (vect_grouped_load_supported): Likewise.
20081 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
20082 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
20083 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
20084 (simplify_const_unary_operation, simplify_binary_operation_1)
20085 (simplify_const_binary_operation, simplify_ternary_operation)
20086 (test_vector_ops_duplicate, test_vector_ops): Likewise.
20087 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
20088 instead of CONST_VECTOR_NUNITS.
20089 * varasm.c (output_constant_pool_2): Likewise.
20090 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
20091 explicit-encoded elements in the XVEC for variable-length vectors.
20093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20095 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
20097 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20098 Alan Hayward <alan.hayward@arm.com>
20099 David Sherwood <david.sherwood@arm.com>
20101 * coretypes.h (fixed_size_mode): Declare.
20102 (fixed_size_mode_pod): New typedef.
20103 * builtins.h (target_builtins::x_apply_args_mode)
20104 (target_builtins::x_apply_result_mode): Change type to
20105 fixed_size_mode_pod.
20106 * builtins.c (apply_args_size, apply_result_size, result_vector)
20107 (expand_builtin_apply_args_1, expand_builtin_apply)
20108 (expand_builtin_return): Update accordingly.
20110 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20112 * cse.c (hash_rtx_cb): Hash only the encoded elements.
20113 * cselib.c (cselib_hash_rtx): Likewise.
20114 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
20115 CONST_VECTOR encoding.
20117 2018-01-03 Jakub Jelinek <jakub@redhat.com>
20118 Jeff Law <law@redhat.com>
20121 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
20122 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
20123 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
20124 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
20127 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
20128 explicitly probe *sp in a noreturn function if there were any callee
20129 register saves or frame pointer is needed.
20131 2018-01-03 Jakub Jelinek <jakub@redhat.com>
20134 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
20135 BLKmode for ternary, binary or unary expressions.
20138 * var-tracking.c (delete_vta_debug_insn): New inline function.
20139 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
20140 insns from get_insns () to NULL instead of each bb separately.
20141 Use delete_vta_debug_insn. No longer static.
20142 (vt_debug_insns_local, variable_tracking_main_1): Adjust
20143 delete_vta_debug_insns callers.
20144 * rtl.h (delete_vta_debug_insns): Declare.
20145 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
20146 instead of variable_tracking_main.
20148 2018-01-03 Martin Sebor <msebor@redhat.com>
20150 PR tree-optimization/83603
20151 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
20152 arguments past the endof the argument list in functions declared
20153 without a prototype.
20154 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
20155 Avoid checking when arguments are null.
20157 2018-01-03 Martin Sebor <msebor@redhat.com>
20160 * doc/extend.texi (attribute const): Fix a typo.
20161 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
20162 issuing -Wsuggest-attribute for void functions.
20164 2018-01-03 Martin Sebor <msebor@redhat.com>
20166 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
20167 offset_int::from instead of wide_int::to_shwi.
20168 (maybe_diag_overlap): Remove assertion.
20169 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
20170 * gimple-ssa-sprintf.c (format_directive): Same.
20171 (parse_directive): Same.
20172 (sprintf_dom_walker::compute_format_length): Same.
20173 (try_substitute_return_value): Same.
20175 2018-01-03 Jeff Law <law@redhat.com>
20177 PR middle-end/83654
20178 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
20179 non-constant residual for zero at runtime and avoid probing in
20180 that case. Reorganize code for trailing problem to mirror handling
20183 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
20185 PR tree-optimization/83501
20186 * tree-ssa-strlen.c (get_string_cst): New.
20187 (handle_char_store): Call get_string_cst.
20189 2018-01-03 Martin Liska <mliska@suse.cz>
20191 PR tree-optimization/83593
20192 * tree-ssa-strlen.c: Include tree-cfg.h.
20193 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
20194 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
20195 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
20197 (strlen_dom_walker::before_dom_children): Call
20198 gimple_purge_dead_eh_edges. Dump tranformation with details
20200 (strlen_dom_walker::before_dom_children): Update call by adding
20201 new argument cleanup_eh.
20202 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
20204 2018-01-03 Martin Liska <mliska@suse.cz>
20207 * cif-code.def (VARIADIC_THUNK): New enum value.
20208 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
20211 2018-01-03 Jan Beulich <jbeulich@suse.com>
20213 * sse.md (mov<mode>_internal): Tighten condition for when to use
20214 vmovdqu<ssescalarsize> for TI and OI modes.
20216 2018-01-03 Jakub Jelinek <jakub@redhat.com>
20218 Update copyright years.
20220 2018-01-03 Martin Liska <mliska@suse.cz>
20223 * ipa-visibility.c (function_and_variable_visibility): Skip
20224 functions with noipa attribure.
20226 2018-01-03 Jakub Jelinek <jakub@redhat.com>
20228 * gcc.c (process_command): Update copyright notice dates.
20229 * gcov-dump.c (print_version): Ditto.
20230 * gcov.c (print_version): Ditto.
20231 * gcov-tool.c (print_version): Ditto.
20232 * gengtype.c (create_file): Ditto.
20233 * doc/cpp.texi: Bump @copying's copyright year.
20234 * doc/cppinternals.texi: Ditto.
20235 * doc/gcc.texi: Ditto.
20236 * doc/gccint.texi: Ditto.
20237 * doc/gcov.texi: Ditto.
20238 * doc/install.texi: Ditto.
20239 * doc/invoke.texi: Ditto.
20241 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20243 * vector-builder.h (vector_builder::m_full_nelts): Change from
20244 unsigned int to poly_uint64.
20245 (vector_builder::full_nelts): Update prototype accordingly.
20246 (vector_builder::new_vector): Likewise.
20247 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
20248 (vector_builder::operator ==): Likewise.
20249 (vector_builder::finalize): Likewise.
20250 * int-vector-builder.h (int_vector_builder::int_vector_builder):
20251 Take the number of elements as a poly_uint64 rather than an
20253 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
20254 from unsigned int to poly_uint64.
20255 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
20256 (vec_perm_indices::new_vector): Likewise.
20257 (vec_perm_indices::length): Likewise.
20258 (vec_perm_indices::nelts_per_input): Likewise.
20259 (vec_perm_indices::input_nelts): Likewise.
20260 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
20261 number of elements per input as a poly_uint64 rather than an
20262 unsigned int. Use the original encoding for variable-length
20263 vectors, rather than clamping each individual element.
20264 For the second and subsequent elements in each pattern,
20265 clamp the step and base before clamping their sum.
20266 (vec_perm_indices::series_p): Handle polynomial element counts.
20267 (vec_perm_indices::all_in_range_p): Likewise.
20268 (vec_perm_indices_to_tree): Likewise.
20269 (vec_perm_indices_to_rtx): Likewise.
20270 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
20271 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
20272 (tree_vector_builder::new_binary_operation): Handle polynomial
20273 element counts. Return false if we need to know the number
20274 of elements at compile time.
20275 * fold-const.c (fold_vec_perm): Punt if the number of elements
20276 isn't known at compile time.
20278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20280 * vec-perm-indices.h (vec_perm_builder): Change element type
20281 from HOST_WIDE_INT to poly_int64.
20282 (vec_perm_indices::element_type): Update accordingly.
20283 (vec_perm_indices::clamp): Handle polynomial element_types.
20284 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
20285 (vec_perm_indices::all_in_range_p): Likewise.
20286 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
20288 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
20289 polynomial vec_perm_indices element types.
20290 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
20291 * fold-const.c (fold_vec_perm): Likewise.
20292 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
20293 * tree-vect-generic.c (lower_vec_perm): Likewise.
20294 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
20295 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
20296 element type to HOST_WIDE_INT.
20298 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20299 Alan Hayward <alan.hayward@arm.com>
20300 David Sherwood <david.sherwood@arm.com>
20302 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
20303 rather than an int. Use plus_constant.
20304 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
20305 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
20307 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20308 Alan Hayward <alan.hayward@arm.com>
20309 David Sherwood <david.sherwood@arm.com>
20311 * calls.c (emit_call_1, expand_call): Change struct_value_size from
20312 a HOST_WIDE_INT to a poly_int64.
20314 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20315 Alan Hayward <alan.hayward@arm.com>
20316 David Sherwood <david.sherwood@arm.com>
20318 * calls.c (load_register_parameters): Cope with polynomial
20319 mode sizes. Require a constant size for BLKmode parameters
20320 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
20321 forces a parameter to be padded at the lsb end in order to
20322 fill a complete number of words, require the parameter size
20323 to be ordered wrt UNITS_PER_WORD.
20325 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20326 Alan Hayward <alan.hayward@arm.com>
20327 David Sherwood <david.sherwood@arm.com>
20329 * reload1.c (spill_stack_slot_width): Change element type
20330 from unsigned int to poly_uint64_pod.
20331 (alter_reg): Treat mode sizes as polynomial.
20333 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20334 Alan Hayward <alan.hayward@arm.com>
20335 David Sherwood <david.sherwood@arm.com>
20337 * reload.c (complex_word_subreg_p): New function.
20338 (reload_inner_reg_of_subreg, push_reload): Use it.
20340 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20341 Alan Hayward <alan.hayward@arm.com>
20342 David Sherwood <david.sherwood@arm.com>
20344 * lra-constraints.c (process_alt_operands): Reject matched
20345 operands whose sizes aren't ordered.
20346 (match_reload): Refer to this check here.
20348 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20349 Alan Hayward <alan.hayward@arm.com>
20350 David Sherwood <david.sherwood@arm.com>
20352 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
20353 that the mode size is in the set {1, 2, 4, 8, 16}.
20355 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20356 Alan Hayward <alan.hayward@arm.com>
20357 David Sherwood <david.sherwood@arm.com>
20359 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
20360 Use plus_constant instead of gen_rtx_PLUS.
20362 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20363 Alan Hayward <alan.hayward@arm.com>
20364 David Sherwood <david.sherwood@arm.com>
20366 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
20367 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
20368 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
20369 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
20370 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
20371 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
20372 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
20373 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
20374 * config/i386/i386.c (ix86_push_rounding): ...this new function.
20375 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
20377 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
20378 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
20379 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
20380 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
20381 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
20382 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
20383 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
20384 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
20385 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
20386 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
20388 * expr.c (emit_move_resolve_push): Treat the input and result
20389 of PUSH_ROUNDING as a poly_int64.
20390 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
20391 (emit_push_insn): Likewise.
20392 * lra-eliminations.c (mark_not_eliminable): Likewise.
20393 * recog.c (push_operand): Likewise.
20394 * reload1.c (elimination_effects): Likewise.
20395 * rtlanal.c (nonzero_bits1): Likewise.
20396 * calls.c (store_one_arg): Likewise. Require the padding to be
20397 known at compile time.
20399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20400 Alan Hayward <alan.hayward@arm.com>
20401 David Sherwood <david.sherwood@arm.com>
20403 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
20404 Use plus_constant instead of gen_rtx_PLUS.
20406 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20407 Alan Hayward <alan.hayward@arm.com>
20408 David Sherwood <david.sherwood@arm.com>
20410 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
20411 rather than an int.
20413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20414 Alan Hayward <alan.hayward@arm.com>
20415 David Sherwood <david.sherwood@arm.com>
20417 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
20418 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
20419 via stack temporaries. Treat the mode size as polynomial too.
20421 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20422 Alan Hayward <alan.hayward@arm.com>
20423 David Sherwood <david.sherwood@arm.com>
20425 * expr.c (expand_expr_real_2): When handling conversions involving
20426 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
20427 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
20428 as a poly_uint64 too.
20430 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20431 Alan Hayward <alan.hayward@arm.com>
20432 David Sherwood <david.sherwood@arm.com>
20434 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
20436 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20437 Alan Hayward <alan.hayward@arm.com>
20438 David Sherwood <david.sherwood@arm.com>
20440 * combine.c (can_change_dest_mode): Handle polynomial
20441 REGMODE_NATURAL_SIZE.
20442 * expmed.c (store_bit_field_1): Likewise.
20443 * expr.c (store_constructor): Likewise.
20444 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
20445 and polynomial REGMODE_NATURAL_SIZE.
20446 (gen_lowpart_common): Likewise.
20447 * reginfo.c (record_subregs_of_mode): Likewise.
20448 * rtlanal.c (read_modify_subreg_p): Likewise.
20450 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20451 Alan Hayward <alan.hayward@arm.com>
20452 David Sherwood <david.sherwood@arm.com>
20454 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
20455 numbers of elements.
20457 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20458 Alan Hayward <alan.hayward@arm.com>
20459 David Sherwood <david.sherwood@arm.com>
20461 * match.pd: Cope with polynomial numbers of vector elements.
20463 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20464 Alan Hayward <alan.hayward@arm.com>
20465 David Sherwood <david.sherwood@arm.com>
20467 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
20468 in a POINTER_PLUS_EXPR.
20470 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20471 Alan Hayward <alan.hayward@arm.com>
20472 David Sherwood <david.sherwood@arm.com>
20474 * omp-simd-clone.c (simd_clone_subparts): New function.
20475 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
20476 (ipa_simd_modify_function_body): Likewise.
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 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
20483 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
20484 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
20485 (expand_vector_condition, vector_element): Likewise.
20486 (subparts_gt): New function.
20487 (get_compute_type): Use subparts_gt.
20488 (count_type_subparts): Delete.
20489 (expand_vector_operations_1): Use subparts_gt instead of
20490 count_type_subparts.
20492 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20493 Alan Hayward <alan.hayward@arm.com>
20494 David Sherwood <david.sherwood@arm.com>
20496 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
20497 (vect_compile_time_alias): ...this new function. Do the calculation
20498 on poly_ints rather than trees.
20499 (vect_prune_runtime_alias_test_list): Update call accordingly.
20501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20502 Alan Hayward <alan.hayward@arm.com>
20503 David Sherwood <david.sherwood@arm.com>
20505 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
20507 (vect_schedule_slp_instance): Likewise.
20509 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20510 Alan Hayward <alan.hayward@arm.com>
20511 David Sherwood <david.sherwood@arm.com>
20513 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
20514 constant and extern definitions for variable-length vectors.
20515 (vect_get_constant_vectors): Note that the number of units
20516 is known to be constant.
20518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20519 Alan Hayward <alan.hayward@arm.com>
20520 David Sherwood <david.sherwood@arm.com>
20522 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
20523 of units as polynomial. Choose between WIDE and NARROW based
20526 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20527 Alan Hayward <alan.hayward@arm.com>
20528 David Sherwood <david.sherwood@arm.com>
20530 * tree-vect-stmts.c (simd_clone_subparts): New function.
20531 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
20533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20534 Alan Hayward <alan.hayward@arm.com>
20535 David Sherwood <david.sherwood@arm.com>
20537 * tree-vect-stmts.c (vectorizable_call): Treat the number of
20538 vectors as polynomial. Use build_index_vector for
20539 IFN_GOMP_SIMD_LANE.
20541 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20542 Alan Hayward <alan.hayward@arm.com>
20543 David Sherwood <david.sherwood@arm.com>
20545 * tree-vect-stmts.c (get_load_store_type): Treat the number of
20546 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
20547 for variable-length vectors.
20548 (vectorizable_mask_load_store): Treat the number of units as
20549 polynomial, asserting that it is constant if the condition has
20550 already been enforced.
20551 (vectorizable_store, vectorizable_load): Likewise.
20553 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20554 Alan Hayward <alan.hayward@arm.com>
20555 David Sherwood <david.sherwood@arm.com>
20557 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
20558 of units as polynomial. Punt if we can't tell at compile time
20559 which vector contains the final result.
20561 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20562 Alan Hayward <alan.hayward@arm.com>
20563 David Sherwood <david.sherwood@arm.com>
20565 * tree-vect-loop.c (vectorizable_induction): Treat the number
20566 of units as polynomial. Punt on SLP inductions. Use an integer
20567 VEC_SERIES_EXPR for variable-length integer reductions. Use a
20568 cast of such a series for variable-length floating-point
20571 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20572 Alan Hayward <alan.hayward@arm.com>
20573 David Sherwood <david.sherwood@arm.com>
20575 * tree.h (build_index_vector): Declare.
20576 * tree.c (build_index_vector): New function.
20577 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
20578 of units as polynomial, forcibly converting it to a constant if
20579 vectorizable_reduction has already enforced the condition.
20580 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
20581 to create a {1,2,3,...} vector.
20582 (vectorizable_reduction): Treat the number of units as polynomial.
20583 Choose vectype_in based on the largest scalar element size rather
20584 than the smallest number of units. Enforce the restrictions
20587 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20588 Alan Hayward <alan.hayward@arm.com>
20589 David Sherwood <david.sherwood@arm.com>
20591 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
20592 number of units as polynomial.
20594 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20595 Alan Hayward <alan.hayward@arm.com>
20596 David Sherwood <david.sherwood@arm.com>
20598 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
20599 * target.def (autovectorize_vector_sizes): Return the vector sizes
20600 by pointer, using vector_sizes rather than a bitmask.
20601 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
20602 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
20603 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
20605 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
20606 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
20607 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
20608 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
20609 * omp-general.c (omp_max_vf): Likewise.
20610 * omp-low.c (omp_clause_aligned_alignment): Likewise.
20611 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
20612 * tree-vect-loop.c (vect_analyze_loop): Likewise.
20613 * tree-vect-slp.c (vect_slp_bb): Likewise.
20614 * doc/tm.texi: Regenerate.
20615 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
20617 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
20618 the vector size as a poly_uint64 rather than an unsigned int.
20619 (current_vector_size): Change from an unsigned int to a poly_uint64.
20620 (get_vectype_for_scalar_type): Update accordingly.
20621 * tree.h (build_truth_vector_type): Take the size and number of
20622 units as a poly_uint64 rather than an unsigned int.
20623 (build_vector_type): Add a temporary overload that takes
20624 the number of units as a poly_uint64 rather than an unsigned int.
20625 * tree.c (make_vector_type): Likewise.
20626 (build_truth_vector_type): Take the number of units as a poly_uint64
20627 rather than an unsigned int.
20629 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20630 Alan Hayward <alan.hayward@arm.com>
20631 David Sherwood <david.sherwood@arm.com>
20633 * target.def (get_mask_mode): Take the number of units and length
20634 as poly_uint64s rather than unsigned ints.
20635 * targhooks.h (default_get_mask_mode): Update accordingly.
20636 * targhooks.c (default_get_mask_mode): Likewise.
20637 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
20638 * doc/tm.texi: Regenerate.
20640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20641 Alan Hayward <alan.hayward@arm.com>
20642 David Sherwood <david.sherwood@arm.com>
20644 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
20645 * omp-general.c (omp_max_vf): Likewise.
20646 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
20647 (expand_omp_simd): Handle polynomial safelen.
20648 * omp-low.c (omplow_simd_context): Add a default constructor.
20649 (omplow_simd_context::max_vf): Change from int to poly_uint64.
20650 (lower_rec_simd_input_clauses): Update accordingly.
20651 (lower_rec_input_clauses): Likewise.
20653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20654 Alan Hayward <alan.hayward@arm.com>
20655 David Sherwood <david.sherwood@arm.com>
20657 * tree-vectorizer.h (vect_nunits_for_cost): New function.
20658 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
20659 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
20660 (vect_analyze_slp_cost): Likewise.
20661 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
20662 (vect_model_load_cost): Likewise.
20664 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20665 Alan Hayward <alan.hayward@arm.com>
20666 David Sherwood <david.sherwood@arm.com>
20668 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
20669 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
20670 from an unsigned int * to a poly_uint64_pod *.
20671 (calculate_unrolling_factor): New function.
20672 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
20674 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20675 Alan Hayward <alan.hayward@arm.com>
20676 David Sherwood <david.sherwood@arm.com>
20678 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
20679 from an unsigned int to a poly_uint64.
20680 (_loop_vec_info::slp_unrolling_factor): Likewise.
20681 (_loop_vec_info::vectorization_factor): Change from an int
20683 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
20684 (vect_get_num_vectors): New function.
20685 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
20686 (vect_get_num_copies): Use vect_get_num_vectors.
20687 (vect_analyze_data_ref_dependences): Change max_vf from an int *
20688 to an unsigned int *.
20689 (vect_analyze_data_refs): Change min_vf from an int * to a
20691 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
20692 than an unsigned HOST_WIDE_INT.
20693 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
20694 (vect_analyze_data_ref_dependence): Change max_vf from an int *
20695 to an unsigned int *.
20696 (vect_analyze_data_ref_dependences): Likewise.
20697 (vect_compute_data_ref_alignment): Handle polynomial vf.
20698 (vect_enhance_data_refs_alignment): Likewise.
20699 (vect_prune_runtime_alias_test_list): Likewise.
20700 (vect_shift_permute_load_chain): Likewise.
20701 (vect_supportable_dr_alignment): Likewise.
20702 (dependence_distance_ge_vf): Take the vectorization factor as a
20703 poly_uint64 rather than an unsigned HOST_WIDE_INT.
20704 (vect_analyze_data_refs): Change min_vf from an int * to a
20706 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
20707 vfm1 as a poly_uint64 rather than an int. Make the same change
20708 for the returned bound_scalar.
20709 (vect_gen_vector_loop_niters): Handle polynomial vf.
20710 (vect_do_peeling): Likewise. Update call to
20711 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
20712 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
20714 * tree-vect-loop.c (vect_determine_vectorization_factor)
20715 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
20716 (vect_get_known_peeling_cost): Likewise.
20717 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
20718 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
20719 (vect_transform_loop): Likewise. Use the lowest possible VF when
20720 updating the upper bounds of the loop.
20721 (vect_min_worthwhile_factor): Make static. Return an unsigned int
20722 rather than an int.
20723 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
20724 polynomial unroll factors.
20725 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
20726 (vect_make_slp_decision): Likewise.
20727 (vect_supported_load_permutation_p): Likewise, and polynomial
20729 (vect_analyze_slp_cost): Handle polynomial vf.
20730 (vect_slp_analyze_node_operations): Likewise.
20731 (vect_slp_analyze_bb_1): Likewise.
20732 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
20733 than an unsigned HOST_WIDE_INT.
20734 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
20735 (vectorizable_load): Handle polynomial vf.
20736 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
20738 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
20740 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20741 Alan Hayward <alan.hayward@arm.com>
20742 David Sherwood <david.sherwood@arm.com>
20744 * match.pd: Handle bit operations involving three constants
20745 and try to fold one pair.
20747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20749 * tree-vect-loop-manip.c: Include gimple-fold.h.
20750 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
20751 niters_maybe_zero parameters. Handle other cases besides a step of 1.
20752 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
20753 Add a path that uses a step of VF instead of 1, but disable it
20755 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
20756 and niters_no_overflow parameters. Update calls to
20757 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
20758 Create a new SSA name if the latter choses to use a ste other
20759 than zero, and return it via niters_vector_mult_vf_var.
20760 * tree-vect-loop.c (vect_transform_loop): Update calls to
20761 vect_do_peeling, vect_gen_vector_loop_niters and
20762 slpeel_make_loop_iterate_ntimes.
20763 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
20764 (vect_gen_vector_loop_niters): Update declarations after above changes.
20766 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
20768 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
20769 128-bit round to integer instructions.
20770 (ceil<mode>2): Likewise.
20771 (btrunc<mode>2): Likewise.
20772 (round<mode>2): Likewise.
20774 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20776 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
20777 unaligned VSX load/store on P8/P9.
20778 (expand_block_clear): Allow the use of unaligned VSX
20779 load/store on P8/P9.
20781 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20783 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
20785 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
20786 swap associated with both a load and a store.
20788 2018-01-02 Andrew Waterman <andrew@sifive.com>
20790 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
20791 * config/riscv/riscv.md (clear_cache): Use it.
20793 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
20795 * web.c: Remove out-of-date comment.
20797 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20799 * expr.c (fixup_args_size_notes): Check that any existing
20800 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
20801 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
20802 (emit_single_push_insn): ...here.
20804 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20806 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
20807 (const_vector_encoded_nelts): New function.
20808 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
20809 (const_vector_int_elt, const_vector_elt): Declare.
20810 * emit-rtl.c (const_vector_int_elt_1): New function.
20811 (const_vector_elt): Likewise.
20812 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
20813 of CONST_VECTOR_ELT.
20815 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20817 * expr.c: Include rtx-vector-builder.h.
20818 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
20819 directly on the tree encoding.
20820 (const_vector_from_tree): Likewise.
20821 * optabs.c: Include rtx-vector-builder.h.
20822 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
20823 sequence of "u" values.
20824 * vec-perm-indices.c: Include rtx-vector-builder.h.
20825 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
20826 directly on the vec_perm_indices encoding.
20828 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20830 * doc/rtl.texi (const_vector): Describe new encoding scheme.
20831 * Makefile.in (OBJS): Add rtx-vector-builder.o.
20832 * rtx-vector-builder.h: New file.
20833 * rtx-vector-builder.c: Likewise.
20834 * rtl.h (rtx_def::u2): Add a const_vector field.
20835 (CONST_VECTOR_NPATTERNS): New macro.
20836 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
20837 (CONST_VECTOR_DUPLICATE_P): Likewise.
20838 (CONST_VECTOR_STEPPED_P): Likewise.
20839 (CONST_VECTOR_ENCODED_ELT): Likewise.
20840 (const_vec_duplicate_p): Check for a duplicated vector encoding.
20841 (unwrap_const_vec_duplicate): Likewise.
20842 (const_vec_series_p): Check for a non-duplicated vector encoding.
20843 Say that the function only returns true for integer vectors.
20844 * emit-rtl.c: Include rtx-vector-builder.h.
20845 (gen_const_vec_duplicate_1): Delete.
20846 (gen_const_vector): Call gen_const_vec_duplicate instead of
20847 gen_const_vec_duplicate_1.
20848 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
20849 (gen_const_vec_duplicate): Use rtx_vector_builder.
20850 (gen_const_vec_series): Likewise.
20851 (gen_rtx_CONST_VECTOR): Likewise.
20852 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
20853 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
20854 Build a new vector rather than modifying a CONST_VECTOR in-place.
20855 (handle_special_swappables): Update call accordingly.
20856 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
20857 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
20858 Build a new vector rather than modifying a CONST_VECTOR in-place.
20859 (handle_special_swappables): Update call accordingly.
20861 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20863 * simplify-rtx.c (simplify_const_binary_operation): Use
20864 CONST_VECTOR_ELT instead of XVECEXP.
20866 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20868 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
20869 the selector elements to be different from the data elements
20870 if the selector is a VECTOR_CST.
20871 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
20872 ssizetype for the selector.
20874 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20876 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
20877 before testing each element individually.
20878 * tree-vect-generic.c (lower_vec_perm): Likewise.
20880 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20882 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
20883 * selftest-run-tests.c (selftest::run_tests): Call it.
20884 * vector-builder.h (vector_builder::operator ==): New function.
20885 (vector_builder::operator !=): Likewise.
20886 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
20887 (vec_perm_indices::all_from_input_p): New function.
20888 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
20889 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
20890 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
20891 instead of reading the VECTOR_CST directly. Detect whether both
20892 vector inputs are the same before constructing the vec_perm_indices,
20893 and update the number of inputs argument accordingly. Use the
20894 utility functions added above. Only construct sel2 if we need to.
20896 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20898 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
20899 the broadcast of the low byte.
20900 (expand_mult_highpart): Use an explicit encoding for the permutes.
20901 * optabs-query.c (can_mult_highpart_p): Likewise.
20902 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
20903 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20904 (vectorizable_bswap): Likewise.
20905 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
20906 explicit encoding for the power-of-2 permutes.
20907 (vect_permute_store_chain): Likewise.
20908 (vect_grouped_load_supported): Likewise.
20909 (vect_permute_load_chain): Likewise.
20911 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20913 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
20914 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
20915 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
20916 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
20917 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
20918 (vect_gen_perm_mask_any): Likewise.
20920 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20922 * int-vector-builder.h: New file.
20923 * vec-perm-indices.h: Include int-vector-builder.h.
20924 (vec_perm_indices): Redefine as an int_vector_builder.
20925 (auto_vec_perm_indices): Delete.
20926 (vec_perm_builder): Redefine as a stand-alone class.
20927 (vec_perm_indices::vec_perm_indices): New function.
20928 (vec_perm_indices::clamp): Likewise.
20929 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
20930 (vec_perm_indices::new_vector): New function.
20931 (vec_perm_indices::new_expanded_vector): Update for new
20932 vec_perm_indices class.
20933 (vec_perm_indices::rotate_inputs): New function.
20934 (vec_perm_indices::all_in_range_p): Operate directly on the
20935 encoded form, without computing elided elements.
20936 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
20937 encoding. Update for new vec_perm_indices class.
20938 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
20939 the given vec_perm_builder.
20940 (expand_vec_perm_var): Update vec_perm_builder constructor.
20941 (expand_mult_highpart): Use vec_perm_builder instead of
20942 auto_vec_perm_indices.
20943 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
20944 vec_perm_indices instead of auto_vec_perm_indices. Use a single
20945 or double series encoding as appropriate.
20946 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
20947 vec_perm_indices instead of auto_vec_perm_indices.
20948 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
20949 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
20950 (vect_permute_store_chain): Likewise.
20951 (vect_grouped_load_supported): Likewise.
20952 (vect_permute_load_chain): Likewise.
20953 (vect_shift_permute_load_chain): Likewise.
20954 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
20955 (vect_transform_slp_perm_load): Likewise.
20956 (vect_schedule_slp_instance): Likewise.
20957 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20958 (vectorizable_mask_load_store): Likewise.
20959 (vectorizable_bswap): Likewise.
20960 (vectorizable_store): Likewise.
20961 (vectorizable_load): Likewise.
20962 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
20963 vec_perm_indices instead of auto_vec_perm_indices. Use
20964 tree_to_vec_perm_builder to read the vector from a tree.
20965 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
20966 vec_perm_builder instead of a vec_perm_indices.
20967 (have_whole_vector_shift): Use vec_perm_builder and
20968 vec_perm_indices instead of auto_vec_perm_indices. Leave the
20969 truncation to calc_vec_perm_mask_for_shift.
20970 (vect_create_epilog_for_reduction): Likewise.
20971 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
20972 from auto_vec_perm_indices to vec_perm_indices.
20973 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
20974 instead of changing individual elements.
20975 (aarch64_vectorize_vec_perm_const): Use new_vector to install
20976 the vector in d.perm.
20977 * config/arm/arm.c (expand_vec_perm_d::perm): Change
20978 from auto_vec_perm_indices to vec_perm_indices.
20979 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
20980 instead of changing individual elements.
20981 (arm_vectorize_vec_perm_const): Use new_vector to install
20982 the vector in d.perm.
20983 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
20984 Update vec_perm_builder constructor.
20985 (rs6000_expand_interleave): Likewise.
20986 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
20987 (rs6000_expand_interleave): Likewise.
20989 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20991 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
20992 to qimode could truncate the indices.
20993 * optabs.c (expand_vec_perm_var): Likewise.
20995 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20997 * Makefile.in (OBJS): Add vec-perm-indices.o.
20998 * vec-perm-indices.h: New file.
20999 * vec-perm-indices.c: Likewise.
21000 * target.h (vec_perm_indices): Replace with a forward class
21002 (auto_vec_perm_indices): Move to vec-perm-indices.h.
21003 * optabs.h: Include vec-perm-indices.h.
21004 (expand_vec_perm): Delete.
21005 (selector_fits_mode_p, expand_vec_perm_var): Declare.
21006 (expand_vec_perm_const): Declare.
21007 * target.def (vec_perm_const_ok): Replace with...
21008 (vec_perm_const): ...this new hook.
21009 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
21010 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
21011 * doc/tm.texi: Regenerate.
21012 * optabs.def (vec_perm_const): Delete.
21013 * doc/md.texi (vec_perm_const): Likewise.
21014 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
21015 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
21016 expand_vec_perm for constant permutation vectors. Assert that
21017 the mode of variable permutation vectors is the integer equivalent
21018 of the mode that is being permuted.
21019 * optabs-query.h (selector_fits_mode_p): Declare.
21020 * optabs-query.c: Include vec-perm-indices.h.
21021 (selector_fits_mode_p): New function.
21022 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
21023 is defined, instead of checking whether the vec_perm_const_optab
21024 exists. Use targetm.vectorize.vec_perm_const instead of
21025 targetm.vectorize.vec_perm_const_ok. Check whether the indices
21026 fit in the vector mode before using a variable permute.
21027 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
21028 vec_perm_indices instead of an rtx.
21029 (expand_vec_perm): Replace with...
21030 (expand_vec_perm_const): ...this new function. Take the selector
21031 as a vec_perm_indices rather than an rtx. Also take the mode of
21032 the selector. Update call to shift_amt_for_vec_perm_mask.
21033 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
21034 Use vec_perm_indices::new_expanded_vector to expand the original
21035 selector into bytes. Check whether the indices fit in the vector
21036 mode before using a variable permute.
21037 (expand_vec_perm_var): Make global.
21038 (expand_mult_highpart): Use expand_vec_perm_const.
21039 * fold-const.c: Includes vec-perm-indices.h.
21040 * tree-ssa-forwprop.c: Likewise.
21041 * tree-vect-data-refs.c: Likewise.
21042 * tree-vect-generic.c: Likewise.
21043 * tree-vect-loop.c: Likewise.
21044 * tree-vect-slp.c: Likewise.
21045 * tree-vect-stmts.c: Likewise.
21046 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
21048 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
21049 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
21050 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
21051 (aarch64_vectorize_vec_perm_const): ...this new function.
21052 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
21053 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21054 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
21055 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
21056 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
21057 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21058 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
21060 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
21061 check for NEON modes.
21062 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
21063 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
21064 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
21065 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
21067 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
21068 the old VEC_PERM_CONST conditions.
21069 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
21070 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
21071 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
21072 (ia64_vectorize_vec_perm_const_ok): Merge into...
21073 (ia64_vectorize_vec_perm_const): ...this new function.
21074 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
21075 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
21076 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
21077 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
21078 * config/mips/mips.c (mips_expand_vec_perm_const)
21079 (mips_vectorize_vec_perm_const_ok): Merge into...
21080 (mips_vectorize_vec_perm_const): ...this new function.
21081 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
21082 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
21083 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
21084 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
21085 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
21086 (rs6000_expand_vec_perm_const): Delete.
21087 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
21089 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21090 (altivec_expand_vec_perm_const_le): Take each operand individually.
21091 Operate on constant selectors rather than rtxes.
21092 (altivec_expand_vec_perm_const): Likewise. Update call to
21093 altivec_expand_vec_perm_const_le.
21094 (rs6000_expand_vec_perm_const): Delete.
21095 (rs6000_vectorize_vec_perm_const_ok): Delete.
21096 (rs6000_vectorize_vec_perm_const): New function.
21097 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
21098 an element count and rtx array.
21099 (rs6000_expand_extract_even): Update call accordingly.
21100 (rs6000_expand_interleave): Likewise.
21101 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
21102 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
21103 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
21104 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
21105 (rs6000_expand_vec_perm_const): Delete.
21106 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
21107 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21108 (altivec_expand_vec_perm_const_le): Take each operand individually.
21109 Operate on constant selectors rather than rtxes.
21110 (altivec_expand_vec_perm_const): Likewise. Update call to
21111 altivec_expand_vec_perm_const_le.
21112 (rs6000_expand_vec_perm_const): Delete.
21113 (rs6000_vectorize_vec_perm_const_ok): Delete.
21114 (rs6000_vectorize_vec_perm_const): New function. Remove stray
21115 reference to the SPE evmerge intructions.
21116 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
21117 an element count and rtx array.
21118 (rs6000_expand_extract_even): Update call accordingly.
21119 (rs6000_expand_interleave): Likewise.
21120 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
21121 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
21123 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21125 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21127 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
21128 vector mode and that that mode matches the mode of the data
21130 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
21131 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
21132 directly using expand_vec_perm_1 when forcing selectors into
21134 (expand_vec_perm_var): New function, split out from expand_vec_perm.
21136 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21138 * optabs-query.h (can_vec_perm_p): Delete.
21139 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
21140 * optabs-query.c (can_vec_perm_p): Split into...
21141 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
21142 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
21143 particular selector is valid.
21144 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
21145 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
21146 (vect_grouped_load_supported): Likewise.
21147 (vect_shift_permute_load_chain): Likewise.
21148 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
21149 (vect_transform_slp_perm_load): Likewise.
21150 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
21151 (vectorizable_bswap): Likewise.
21152 (vect_gen_perm_mask_checked): Likewise.
21153 * fold-const.c (fold_ternary_loc): Likewise. Don't take
21154 implementations of variable permutation vectors into account
21155 when deciding which selector to use.
21156 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
21157 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
21158 with a false third argument.
21159 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
21160 to test whether the constant selector is valid and can_vec_perm_var_p
21161 to test whether a variable selector is valid.
21163 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21165 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
21166 * optabs-query.c (can_vec_perm_p): Likewise.
21167 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
21168 instead of vec_perm_indices.
21169 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
21170 (vect_gen_perm_mask_checked): Likewise,
21171 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
21172 (vect_gen_perm_mask_checked): Likewise,
21174 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21176 * optabs-query.h (qimode_for_vec_perm): Declare.
21177 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
21178 (qimode_for_vec_perm): ...this new function.
21179 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
21181 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21183 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
21184 does not have a conditional at the top.
21186 2018-01-02 Richard Biener <rguenther@suse.de>
21188 * ipa-inline.c (big_speedup_p): Fix expression.
21190 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
21193 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
21196 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
21200 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
21201 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
21202 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
21203 cond_taken_branch_cost 3->4.
21205 2018-01-01 Jakub Jelinek <jakub@redhat.com>
21207 PR tree-optimization/83581
21208 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
21209 TODO_cleanup_cfg if any changes have been made.
21211 PR middle-end/83608
21212 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
21213 convert_modes if target mode has the right side, but different mode
21216 PR middle-end/83609
21217 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
21218 last argument when extracting from CONCAT. If either from_real or
21219 from_imag is NULL, use expansion through memory. If result is not
21220 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
21221 the parts directly to inner mode, if even that fails, use expansion
21224 PR middle-end/83623
21225 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
21226 check for bswap in mode rather than HImode and use that in expand_unop
21229 Copyright (C) 2018 Free Software Foundation, Inc.
21231 Copying and distribution of this file, with or without modification,
21232 are permitted in any medium without royalty provided the copyright
21233 notice and this notice are preserved.