1 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
4 create_tmp_var_for to create the FRAME decl.
5 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
7 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
9 * tree-inline.c (copy_edges_for_bb): Minor tweak.
10 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
11 debug statement when resetting its value.
12 (expand_call_inline): Copy the locus of the call onto the assignment
13 of the return value, if any. Use local variable in more cases.
15 2018-06-21 Martin Liska <mliska@suse.cz>
17 * ipa-pure-const.c (propagate_nothrow): Use
18 funct_state_summaries->get.
19 (dump_malloc_lattice): Likewise.
20 (propagate_malloc): Likewise.
22 2018-06-21 Richard Biener <rguenther@suse.de>
24 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
25 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
26 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
28 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
29 BLOCK_ABSTRACT_ORIGIN unconditionally.
31 2018-06-21 David Malcolm <dmalcolm@redhat.com>
33 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
35 * ipa-reference.c (ipa_reference_c_finalize): Delete
36 ipa_ref_opt_sum_summaries and set it to NULL.
38 2018-06-21 Tom de Vries <tdevries@suse.de>
40 PR tree-optimization/85859
41 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
42 test with comment from bb_no_side_effects_p.
44 2018-06-21 Richard Biener <rguenther@suse.de>
46 PR tree-optimization/86232
47 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
48 max for constant niter.
50 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
52 * config/aarch64/aarch64-simd.md
53 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
55 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
57 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
58 Make opernads of the unspec commutative.
60 2018-06-21 Richard Biener <rguenther@suse.de>
62 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
63 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
64 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
65 (vect_analyze_data_ref_dependence): Re-order checks to deal with
67 (vect_record_base_alignments): Do not record base alignment
68 for gathers or scatters.
69 (vect_compute_data_ref_alignment): Drop return value that is always
70 true. Bail out early for gathers or scatters.
71 (vect_enhance_data_refs_alignment): Bail out early for gathers
73 (vect_find_same_alignment_drs): Likewise.
74 (vect_analyze_data_refs_alignment): Remove dead code.
75 (vect_slp_analyze_and_verify_node_alignment): Likewise.
76 (vect_analyze_data_refs): For possible gathers or scatters do
77 not create an alternate DR, just check their possible validity
78 and mark them. Adjust DECL_NONALIASED handling to not rely
80 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
81 update inits of gathers or scatters.
82 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
83 Also copy gather/scatter flag to pattern vinfo.
85 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
87 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
88 behavior of vec_packsu (vector unsigned long long, vector unsigned
89 long long) to match behavior of vec_packs with same signature.
91 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
92 Thomas Schwinge <thomas@codesourcery.com>
93 Cesar Philippidis <cesar@codesourcery.com>
95 * gimplify.c (gimplify_scan_omp_clauses): Add support for
96 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
97 (gimplify_adjust_omp_clauses): Likewise.
98 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
99 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
100 (gimplify_omp_target_update): Update handling of acc update and
102 * omp-low.c (install_var_field): Remove unused parameter
103 base_pointers_restrict.
104 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
105 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
107 (omp_target_base_pointers_restrict_p): Delete.
108 (scan_omp_target): Update call to scan_sharing_clauses.
109 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
111 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
112 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
113 (convert_local_omp_clauses): Likewise.
114 * tree-pretty-print.c (dump_omp_clause): Likewise.
115 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
117 (omp_clause_code_name): Likewise.
119 2018-06-20 Jakub Jelinek <jakub@redhat.com>
122 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
125 PR tree-optimization/86231
126 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
127 anti-range don't overwrite *vr0min before using it to compute *vr0max.
129 2018-06-20 Tom de Vries <tdevries@suse.de>
131 PR tree-optimization/86097
132 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
133 iv type if signedness of iv type is not the same as that of *nit.
135 2018-06-20 Jakub Jelinek <jakub@redhat.com>
137 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
138 EDGE_EH edges, verify they are all EDGE_EH.
140 2018-06-20 Maya Rashish <coypu@sdf.org>
142 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
144 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
146 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
147 * config/aarch64/aarch64.c (xgene1_tunings): Add
148 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
149 (aarch64_mode_valid_for_sched_fusion_p):
151 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
152 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
154 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
156 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
157 * config/aarch64/iterators.md (VQ2): New mode iterator.
159 2018-06-20 Martin Liska <mliska@suse.cz>
161 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
162 Change default ratio from 10 to 8.
164 2018-06-20 Martin Liska <mliska@suse.cz>
166 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
168 (bit_test_cluster::find_bit_tests): Likewise.
169 (switch_decision_tree::analyze_switch_statement): Find clusters.
170 * tree-switch-conversion.h (struct jump_table_cluster): Document
173 2018-06-20 Martin Liska <mliska@suse.cz>
175 * tree-switch-conversion.c (switch_conversion::collect):
176 Record m_uniq property.
177 (switch_conversion::expand): Bail out for special conditions.
178 (group_cluster::~group_cluster): New.
179 (group_cluster::group_cluster): Likewise.
180 (group_cluster::dump): Likewise.
181 (jump_table_cluster::emit): New.
182 (switch_decision_tree::fix_phi_operands_for_edges): New.
183 (struct case_node): Remove struct.
184 (jump_table_cluster::can_be_handled): New.
185 (case_values_threshold): Moved to header.
186 (reset_out_edges_aux): Likewise.
187 (jump_table_cluster::is_beneficial): New.
188 (bit_test_cluster::can_be_handled): Likewise.
189 (add_case_node): Remove.
190 (bit_test_cluster::is_beneficial): New.
191 (case_bit_test::cmp): New.
192 (bit_test_cluster::emit): New.
193 (expand_switch_as_decision_tree_p): Remove.
194 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
195 (fix_phi_operands_for_edge): Likewise.
196 (switch_decision_tree::analyze_switch_statement): New.
197 (compute_cases_per_edge): Move ...
198 (switch_decision_tree::compute_cases_per_edge): ... here.
199 (try_switch_expansion): Likewise.
200 (switch_decision_tree::try_switch_expansion): Likewise.
201 (record_phi_operand_mapping): Likewise.
202 (switch_decision_tree::record_phi_operand_mapping): Likewise.
203 (emit_case_decision_tree): Likewise.
204 (switch_decision_tree::emit): Likewise.
205 (balance_case_nodes): Likewise.
206 (switch_decision_tree::balance_case_nodes): Likewise.
207 (dump_case_nodes): Likewise.
208 (switch_decision_tree::dump_case_nodes): Likewise.
209 (emit_jump): Likewise.
210 (switch_decision_tree::emit_jump): Likewise.
211 (emit_cmp_and_jump_insns): Likewise.
212 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
213 (emit_case_nodes): Likewise.
214 (switch_decision_tree::emit_case_nodes): Likewise.
215 (conditional_probability): Remove.
216 * tree-switch-conversion.h (enum cluster_type): New.
218 (struct cluster): Likewise.
219 (cluster::cluster): Likewise.
220 (struct simple_cluster): Likewise.
221 (simple_cluster::simple_cluster): Likewise.
222 (struct group_cluster): Likewise.
223 (struct jump_table_cluster): Likewise.
224 (struct bit_test_cluster): Likewise.
225 (struct min_cluster_item): Likewise.
226 (struct case_tree_node): Likewise.
227 (case_tree_node::case_tree_node): Likewise.
228 (jump_table_cluster::case_values_threshold): Likewise.
229 (struct case_bit_test): Likewise.
230 (struct switch_decision_tree): Likewise.
231 (struct switch_conversion): Likewise.
232 (switch_decision_tree::reset_out_edges_aux): Likewise.
234 2018-06-20 Martin Liska <mliska@suse.cz>
236 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
237 (hoist_edge_and_branch_if_true): Likewise.
238 (expand_switch_using_bit_tests_p): Likewise.
239 (struct case_bit_test): Likewise.
240 (case_bit_test_cmp): Likewise.
241 (emit_case_bit_tests): Likewise.
242 (switch_conversion::switch_conversion): New class.
243 (struct switch_conv_info): Remove old struct.
244 (collect_switch_conv_info): More to ...
245 (switch_conversion::collect): ... this.
246 (check_range): Likewise.
247 (switch_conversion::check_range): Likewise.
248 (check_all_empty_except_final): Likewise.
249 (switch_conversion::check_all_empty_except_final): Likewise.
250 (check_final_bb): Likewise.
251 (switch_conversion::check_final_bb): Likewise.
252 (create_temp_arrays): Likewise.
253 (switch_conversion::create_temp_arrays): Likewise.
254 (free_temp_arrays): Likewise.
255 (gather_default_values): Likewise.
256 (switch_conversion::gather_default_values): Likewise.
257 (build_constructors): Likewise.
258 (switch_conversion::build_constructors): Likewise.
259 (constructor_contains_same_values_p): Likewise.
260 (switch_conversion::contains_same_values_p): Likewise.
261 (array_value_type): Likewise.
262 (switch_conversion::array_value_type): Likewise.
263 (build_one_array): Likewise.
264 (switch_conversion::build_one_array): Likewise.
265 (build_arrays): Likewise.
266 (switch_conversion::build_arrays): Likewise.
267 (gen_def_assigns): Likewise.
268 (switch_conversion::gen_def_assigns): Likewise.
269 (prune_bbs): Likewise.
270 (switch_conversion::prune_bbs): Likewise.
271 (fix_phi_nodes): Likewise.
272 (switch_conversion::fix_phi_nodes): Likewise.
273 (gen_inbound_check): Likewise.
274 (switch_conversion::gen_inbound_check): Likewise.
275 (process_switch): Use the newly created class.
276 (switch_conversion::expand): New.
277 (switch_conversion::~switch_conversion): New.
278 * tree-switch-conversion.h: New file.
280 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
282 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
283 tree-vect-patterns.c.
284 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
285 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
286 (vect_recog_sad_pattern): Likewise.
287 (vect_recog_widen_sum_pattern): Likewise.
288 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
289 (vect_recog_widen_shift_pattern): Remove the type_in argument.
290 (vect_recog_rotate_pattern): Likewise.
291 (vect_recog_mult_pattern): Likewise.
292 (vect_recog_vector_vector_shift_pattern): Likewise.
293 (vect_recog_divmod_pattern): Likewise.
294 (vect_recog_mixed_size_cond_pattern): Likewise.
295 (vect_recog_bool_pattern): Likewise.
296 (vect_recog_mask_conversion_pattern): Likewise.
297 (vect_try_gather_scatter_pattern): Likewise.
298 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
299 (vect_recog_over_widening_pattern): Likewise.
300 (vect_recog_gather_scatter_pattern): Likewise.
301 (vect_recog_func_ptr): Move from tree-vectorizer.h
302 (vect_vect_recog_func_ptrs): Move further down the file.
303 (vect_recog_func): Likewise. Remove the third argument.
304 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
305 (vect_pattern_recog_1): Expect the pattern function to do any
306 necessary target tests. Also expect it to provide a vector type.
307 Remove the type_in handling.
309 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
311 * tree-vect-patterns.c (vect_pattern_detected): New function.
312 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
313 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
314 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
315 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
316 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
317 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
318 (vect_recog_mask_conversion_pattern)
319 (vect_try_gather_scatter_pattern): Likewise.
321 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
323 * tree-vect-patterns.c (vect_get_internal_def): New function.
324 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
325 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
326 (search_type_for_mask_1): Use it.
328 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
330 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
331 redundant WIDEN_SUM_EXPR handling.
332 (vect_recog_sad_pattern): Likewise.
334 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
336 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
337 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
338 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
339 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
340 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
342 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
344 * tree-vect-stmts.c (vectorizable_call): Make sure that we
345 use the stmt_vec_info of the original bb statement for the
346 new zero assignment, even if the call is part of a pattern.
348 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
350 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
351 that the sequence is attached to the original statement rather
352 than the pattern statement.
353 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
354 PATTERN_DEF_SEQ from the original statement rather than
355 the main pattern statement.
356 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
357 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
358 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
360 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
362 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
363 definition statements before the early exit for statements that aren't
365 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
367 (vect_transform_loop): ...here. Process pattern definition
368 statements without first checking whether the main pattern
369 statement is live or relevant.
371 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
373 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
374 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
376 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
378 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
379 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
380 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
381 (expand_block_compare): Change select_block_compare_mode call.
382 (expand_strncmp_align_check): Use new functions, fix comment.
383 (emit_final_str_compare_gpr): New function.
384 (expand_strn_compare): Refactor and clean up code.
385 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
387 2018-06-19 Tony Reix <tony.reix@atos.com>
388 Damien Bergamini <damien.bergamini@atos.com>
389 David Edelsohn <dje.gcc@gmail.com>
391 * collect2.c (static_obj): New variable.
392 (static_libs): New variable.
393 (is_in_list): Uncomment declaration.
394 (main): Track AIX libraries linked statically.
395 (is_in_list): Uncomment definition.
396 (scan_prog_file): Don't add AIX shared libraries initializer
397 to constructor list if linking statically.
399 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
401 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
403 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
405 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
407 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
410 2018-06-19 Martin Liska <mliska@suse.cz>
412 * config/i386/i386.c (ix86_can_inline_p): Do not use
413 ipa_fn_summaries::get_create.
414 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
416 (devirtualization_time_bonus): Likewise.
417 (ipcp_propagate_stage): Likewise.
418 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
419 (edge_set_predicate): Likewise.
420 (evaluate_conditions_for_known_args): Likewise.
421 (evaluate_properties_for_edge): Likewise.
422 (ipa_call_summary::reset): Tranform to ...
423 (ipa_call_summary::~ipa_call_summary): ... this.
424 (ipa_fn_summary::reset): Transform to ...
425 (ipa_fn_summary::~ipa_fn_summary): ... this.
426 (ipa_fn_summary_t::remove): Rename to ...
427 (ipa_fn_summary_t::remove_callees): ... this.
428 (ipa_fn_summary_t::duplicate): Use placement new
429 instead of memory copy.
430 (ipa_call_summary_t::duplicate): Likewise.
431 (ipa_call_summary_t::remove): Remove.
432 (dump_ipa_call_summary): Change get_create to get.
433 (ipa_dump_fn_summary): Dump only when summary exists.
434 (analyze_function_body): Use symbol_summary::get instead
436 (compute_fn_summary): Likewise.
437 (estimate_edge_devirt_benefit): Likewise.
438 (estimate_edge_size_and_time): Likewise.
439 (inline_update_callee_summaries): Likewise.
440 (remap_edge_change_prob): Likewise.
441 (remap_edge_summaries): Likewise.
442 (ipa_merge_fn_summary_after_inlining): Likewise.
443 (write_ipa_call_summary): Likewise.
444 (ipa_fn_summary_write): Likewise.
445 (ipa_free_fn_summary): Likewise.
446 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
447 (struct ipa_call_summary): Likewise.
448 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
450 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
451 (estimate_size_after_inlining): Likewise.
452 (estimate_growth): Likewise.
453 (growth_likely_positive): Likewise.
454 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
455 (inline_call): Likewise.
456 * ipa-inline.c (caller_growth_limits): Likewise.
457 (can_inline_edge_p): Likewise.
458 (can_inline_edge_by_limits_p): Likewise.
459 (compute_uninlined_call_time): Likewise.
460 (compute_inlined_call_time): Likewise.
461 (want_inline_small_function_p): Likewise.
462 (edge_badness): Likewise.
463 (update_caller_keys): Likewise.
464 (update_callee_keys): Likewise.
465 (inline_small_functions): Likewise.
466 (inline_to_all_callers_1): Likewise.
467 (dump_overall_stats): Likewise.
468 (early_inline_small_functions): Likewise.
469 (early_inliner): Likewise.
470 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
471 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
472 * ipa-pure-const.c (malloc_candidate_p): Likewise.
473 * ipa-split.c (execute_split_functions): Likewise.
474 * symbol-summary.h: Likewise.
475 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
477 2018-06-19 Richard Biener <rguenther@suse.de>
479 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
480 (vectorize_loops): ... here. Fix dbgcnt handling.
481 (try_vectorize_loop): Wrap try_vectorize_loop_1.
483 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
486 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
487 ieee128 argument takes up only one (vector) register, not two (floating
490 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
492 * gimplify.c (gimplify_init_constructor): Really never clear for an
493 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
495 2018-06-19 Richard Biener <rguenther@suse.de>
497 PR tree-optimization/86179
498 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
499 after failed recognition.
501 2018-06-18 Martin Sebor <msebor@redhat.com>
504 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
505 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
506 Handle integer subtraction.
507 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
508 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
510 2018-06-18 David Malcolm <dmalcolm@redhat.com>
512 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
513 param from rtx to rtx_insn *.
514 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
516 (frv_ifcvt_modify_insn): Likwise.
517 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
518 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
519 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
520 as_a <rtx_insn *> cast to local "unprotected_region" once
521 it's been established that it's not NULL or pc_rtx.
522 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
523 param "sethi" from rtx to rtx_insn *.
524 (nds32_group_float_insns): Likewise for param "insn".
525 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
527 (vax_output_int_subtract): Likewise.
528 * config/vax/vax.c (vax_output_int_add): Likewise for param
530 (vax_output_int_subtract): Likewise.
531 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
532 (emit_pattern_after): Likewise for param "after".
533 (emit_insn_after): Likewise.
534 (emit_jump_insn_after): Likewise.
535 (emit_call_insn_after): Likewise.
536 (emit_debug_insn_after): Likewise.
537 (emit_pattern_before): Likewise for param "before".
538 (emit_insn_before): Likewise.
539 (emit_jump_insn_before): Likewise.
540 * final.c (get_insn_template): Likewise for param "insn", removing
542 * output.h (get_insn_template): Likewise for 2nd param.
543 * rtl.h (emit_insn_before): Likewise.
544 (emit_jump_insn_before): Likewise.
545 (emit_debug_insn_before_noloc): Likewise.
546 (emit_insn_after): Likewise.
547 (emit_jump_insn_after): Likewise.
548 (emit_call_insn_after): Likewise.
549 (emit_debug_insn_after): Likewise.
550 (set_insn_deleted): Likewise for param.
552 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
555 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
556 floating point modes, so that IFmode is numerically greater than
557 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
558 to declare the ordering. This prevents IFmode from being
559 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
560 machine. Include rs6000-modes.h to share the fractional values
561 between genmodes* and the rest of the compiler.
565 * config/rs6000/rs6000-modes.h: New file.
566 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
567 meaning of rs6000_long_double_size so that 126..128 selects an
568 appropriate 128-bit floating point type.
569 (rs6000_option_override_internal): Likewise.
570 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
571 (TARGET_LONG_DOUBLE_128): Change the meaning of
572 rs6000_long_double_size so that 126..128 selects an appropriate
573 128-bit floating point type.
574 (LONG_DOUBLE_TYPE_SIZE): Update comment.
575 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
576 source and destination to match the standard usage.
577 (truncifkf2): Likewise.
578 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
579 ISA 2.07 to use an explicit clobber, instead of passing in a
581 (copysign<mode>3_soft): Likewise.
583 2018-06-18 David Malcolm <dmalcolm@redhat.com>
585 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
586 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
587 (vect_slp_analyze_instance_dependence): Likewise.
588 (vect_enhance_data_refs_alignment): Likewise.
589 (vect_analyze_data_refs_alignment): Likewise.
590 (vect_slp_analyze_and_verify_instance_alignment
591 (vect_analyze_data_ref_accesses): Likewise.
592 (vect_prune_runtime_alias_test_list): Likewise.
593 (vect_analyze_data_refs): Likewise.
594 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
595 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
596 (vect_analyze_scalar_cycles_1): Likewise.
597 (vect_get_loop_niters): Likewise.
598 (vect_analyze_loop_form_1): Likewise.
599 (vect_update_vf_for_slp): Likewise.
600 (vect_analyze_loop_operations): Likewise.
601 (vect_analyze_loop): Likewise.
602 (vectorizable_induction): Likewise.
603 (vect_transform_loop): Likewise.
604 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
605 * tree-vect-slp.c (vect_analyze_slp): Likewise.
606 (vect_make_slp_decision): Likewise.
607 (vect_detect_hybrid_slp): Likewise.
608 (vect_slp_analyze_operations): Likewise.
609 (vect_slp_bb): Likewise.
610 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
611 (vectorizable_bswap): Likewise.
612 (vectorizable_call): Likewise.
613 (vectorizable_simd_clone_call): Likewise.
614 (vectorizable_conversion): Likewise.
615 (vectorizable_assignment): Likewise.
616 (vectorizable_shift): Likewise.
617 (vectorizable_operation): Likewise.
618 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
620 2018-06-18 Martin Sebor <msebor@redhat.com>
622 PR tree-optimization/81384
623 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
624 * builtins.c (expand_builtin_strnlen): New function.
625 (expand_builtin): Call it.
626 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
627 * builtins.def (BUILT_IN_STRNLEN): New.
628 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
629 Warn for bounds in excess of maximum object size.
630 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
631 single-value ranges. Handle strnlen.
632 (handle_builtin_strlen): Handle strnlen.
633 (strlen_check_and_optimize_stmt): Same.
634 * doc/extend.texi (Other Builtins): Document strnlen.
636 2018-06-18 Maya Rashish <coypu@sdf.org>
638 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
639 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
640 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
642 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
644 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
646 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
648 * tree.c (escaped_string::escape): Replace cast to char * by
649 const_cast<char *> (unescaped).
651 2018-06-18 Nick Clifton <nickc@redhat.com>
654 * tree.c (escaped_string): New class. Converts an unescaped
655 string into its escaped equivalent.
656 (warn_deprecated_use): Use the new class to convert the
657 deprecation message, if present.
658 (test_escaped_strings): New self test.
659 (test_c_tests): Add test_escaped_strings.
660 * doc/extend.texi (deprecated): Add a note that the
661 deprecation message is affected by the -fmessage-length
662 option, and that control characters will be escaped.
663 (#pragma GCC error): Document this pragma.
664 (#pragma GCC warning): Likewise.
665 * doc/invoke.texi (-fmessage-length): Document this option's
666 effect on the #warning and #error preprocessor directives and
667 the deprecated attribute.
669 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
671 * tree.c (decl_value_expr_lookup): Revert latest change.
672 (decl_value_expr_insert): Likewise.
674 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
676 * gimplify.c (nonlocal_vlas): Delete.
677 (nonlocal_vla_vars): Likewise.
678 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
680 (gimplify_body): Do not create and destroy nonlocal_vlas.
681 * tree-nested.c: Include diagnostic.h.
682 (use_pointer_in_frame): Tweak.
683 (lookup_field_for_decl): Add assertion and declare the transformation.
684 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
685 internal error when the reference is in a wrong context. Do not
686 create a debug decl by default.
687 (note_nonlocal_block_vlas): Delete.
688 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
689 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
690 create a debug decl by default.
691 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
692 call is in a wrong context.
693 (fixup_vla_decls): New function.
694 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
695 debug variables were created.
696 * tree.c (decl_value_expr_lookup): Add checking assertion.
697 (decl_value_expr_insert): Likewise.
699 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
702 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
703 * tree-scalar-evolution.c (interpret_expr): Likewise.
704 (expression_expensive_p): Likewise.
705 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
706 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
707 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
708 (ssa_defined_by_minus_one_stmt_p): New.
710 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
713 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
714 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
715 * dojump.c (do_jump): Likewise.
716 * expr.c (expand_expr_real_2): Check operand type's sign.
717 * fold-const.c (const_unop): Handle ABSU_EXPR.
718 (fold_abs_const): Likewise.
719 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
720 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
721 (strip_sign_op_1): Likesise.
722 * match.pd: Add new pattern to generate ABSU_EXPR.
723 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
724 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
725 * tree-eh.c (operation_could_trap_helper_p): Likewise.
726 * tree-inline.c (estimate_operator_cost): Likewise.
727 * tree-pretty-print.c (dump_generic_node): Likewise.
728 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
729 * tree.def (ABSU_EXPR): New.
731 2018-06-16 Jakub Jelinek <jakub@redhat.com>
734 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
735 documented as preserved for backward compatibility only.
736 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
738 PR rtl-optimization/86108
739 * bb-reorder.c (create_forwarder_block): Renamed to ...
740 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
741 jump from new landing pad to the second part.
742 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
745 2018-06-15 Jakub Jelinek <jakub@redhat.com>
748 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
749 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
750 Only call store_expr for halves if the mode is the same.
753 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
754 Fix up comment formatting.
756 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
758 * typed-splay-tree.h (typed_splay_tree::remove): New function.
759 (typed_splay_tree::closure,
760 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
761 (typed_splay_tree::typed_splay_tree,
762 typed_splay_tree::operator =): Declared private.
763 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
764 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
765 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
766 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
767 typed_splay_tree::splay_tree_splay,
768 typed_splay_tree::splay_tree_foreach_helper,
769 typed_splay_tree::splay_tree_insert,
770 typed_splay_tree::splay_tree_remove,
771 typed_splay_tree::splay_tree_lookup,
772 typed_splay_tree::splay_tree_predecessor,
773 typed_splay_tree::splay_tree_successor,
774 typed_splay_tree::splay_tree_min,
775 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
776 (typed_splay_tree::root, typed_splay_tree::comp,
777 typed_splay_tree::delete_key,
778 typed_splay_tree::delete_value): New data members.
779 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
780 typed_splay_tree::remove.
782 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
784 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
785 -mginv and -mno-ginv to the assembler.
786 * config/mips/mips.opt (-mcrc): New option.
788 * doc/invoke.text (-mcrc): Document.
791 2018-06-15 Nick Clifton <nickc@redhat.com>
794 * tree.c (escaped_string): New class. Converts an unescaped
795 string into its escaped equivalent.
796 (warn_deprecated_use): Use the new class to convert the
797 deprecation message, if present.
798 (test_escaped_strings): New self test.
799 (test_c_tests): Add test_escaped_strings.
800 * doc/extend.texi (deprecated): Add a note that the
801 deprecation message is affected by the -fmessage-length
802 option, and that control characters will be escaped.
803 (#pragma GCC error): Document this pragma.
804 (#pragma GCC warning): Likewise.
805 * doc/invoke.texi (-fmessage-length): Document this option's
806 effect on the #warning and #error preprocessor directives and
807 the deprecated attribute.
809 2018-06-15 Richard Biener <rguenther@suse.de>
811 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
812 here, also noting vector size used.
813 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
814 size used in MSG_OPTIMIZED_LOCATIONS dump.
815 (pass_slp_vectorize::execute): Adjust.
817 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
820 * config/arc/arc.c (arc_return_address_register): Fix
823 2018-06-15 Richard Biener <rguenther@suse.de>
826 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
827 leave useless conversion stripping to force_gimple_operand_gsi.
828 (gimplify_build2): Likewise.
829 (gimplify_build1): Likewise.
831 2018-06-15 Richard Biener <rguenther@suse.de>
834 * tree-cfg.c (move_stmt_op): unshare invariant addresses
835 before adjusting their block.
837 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
839 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
840 multilibs for *-*-rtems*.
841 * config/riscv/t-rtems: New file.
843 2018-06-14 Jakub Jelinek <jakub@redhat.com>
846 * match.pd ((A +- CST1) +- CST2): Punt if last resort
847 unsigned_type_for returns NULL.
850 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
851 subregs of multi-word pseudos unless the float mode has word size.
853 2018-06-14 Richard Biener <rguenther@suse.de>
856 * tree-vect-generic.c (build_word_mode_vector_type): Remove
857 duplicate and harmful type_hash_canon.
858 * tree.c (type_hash_canon): Assert we didn't find ourselves.
860 2018-06-14 Richard Biener <rguenther@suse.de>
863 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
866 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
868 * config/rtems.h (STDINT_LONG32): Define.
870 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
871 Prachi Godbole <prachi.godbole@imgtec.com>
873 * config/mips/mips-cpus.def: Define P6600.
874 * config/mips/mips-tables.opt: Regenerate.
875 * config/mips/mips.c (mips_ucbranch_type): New enum.
876 (mips_rtx_cost_data): Add support for P6600.
877 (mips_issue_rate): Likewise.
878 (mips_multipass_dfa_lookahead): Likewise.
879 (mips_avoid_hazard): Likewise.
880 (mips_reorg_process_insns): Likewise.
881 (mips_classify_branch_p6600): New function.
882 * config/mips/mips.h (TUNE_P6600): New define.
883 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
884 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
885 * config/mips/mips.md: Include p6600.md.
886 (processor): Add p6600.
887 * config/mips/p6600.md: New file.
888 * doc/invoke.texi: Add p6600 to supported architectures.
890 2018-06-13 Martin Sebor <msebor@redhat.com>
892 PR tree-optimization/86114
893 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
895 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
897 2018-06-13 Richard Biener <rguenther@suse.de>
899 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
900 Properly set vector type of the intermediate stmt.
901 * tree-vect-stmts.c (vectorizable_operation): The destination
902 var always has vectype_out type.
904 2018-06-13 Jeff Law <law@redhat.com>
906 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
907 integer 0 for argument to print_rtl_with_bb.
908 (rl78_reorg): Likewise.
910 2018-06-13 David Malcolm <dmalcolm@redhat.com>
912 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
913 from rtx to rtx_insn *.
914 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
916 (add_sched_insns_for_speculation): Likewise for local "target",
917 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
918 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
919 from rtx_insn ** to rtx_code_label **.
920 (reorg_emit_nops): Likewise.
921 (c6x_reorg): Likewise for local "call_labels".
922 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
924 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
925 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
926 the loops over LABEL_REFS.
927 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
929 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
930 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
931 (split_branches): Strengthen local "olabel" from rtx to
932 rtx_insn *, adding a safe_as_a cast.
933 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
935 (add_insn_after): Likewise for first two params.
936 (add_insn_before): Likewise.
937 (remove_insn): Likewise for param.
938 (emit_pattern_before_noloc): Likewise for second and third params.
939 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
940 (emit_call_insn_before_noloc): Likewise.
941 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
943 (emit_barrier_before): Likewise.
944 (emit_label_before): Strengthen "label" param from "rtx" to
945 "rtx_code_label *". Strengthen "before" param from "rtx" to
947 (emit_insn_after_1): Strengthen "after" param from "rtx" to
949 (emit_pattern_after_noloc): Likewise.
950 (emit_insn_after_noloc): Likewise.
951 (emit_jump_insn_after_noloc): Likewise.
952 (emit_call_insn_after_noloc): Likewise.
953 (emit_debug_insn_after_noloc): Likewise.
954 (emit_barrier_after): Likewise.
955 (emit_label_after): Likewise for both params.
956 (emit_pattern_after_setloc): Likewise for "after" param. Convert
957 "loc" param from "int" to "location_t".
958 (emit_insn_after_setloc): Likewise.
959 (emit_jump_insn_after_setloc): Likewise.
960 (emit_call_insn_after_setloc): Likewise.
961 (emit_debug_insn_after_setloc): Likewise.
962 (emit_pattern_before_setloc): Likewise for "before" param. Convert
963 "loc" param from "int" to "location_t".
964 (emit_pattern_before): Convert NULL_RTX to NULL.
965 (emit_insn_before_setloc): Convert "loc" param from "int" to
967 (emit_jump_insn_before_setloc): Likewise.
968 (emit_call_insn_before_setloc): Likewise.
969 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
970 rtx_insn *. Convert "loc" param from "int" to "location_t".
971 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
972 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
973 Convert 3rd param from "int" to "location_t".
974 (emit_barrier_before, emit_barrier_after, next_real_insn):
975 Strengthen param from rtx to rtx_insn *.
976 (emit_label_before): Strengthen 1st param from "rtx" to
977 "rtx_code_label *". Strengthen 2nd param from "rtx" to
979 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
980 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
981 Strengthen 2nd param from "rtx" to "rtx_insn *".
982 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
983 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
984 Likewise. Convert 3rd param from "int" to "location_t".
985 (emit_label_after): Strengthen 1st param from "rtx" to
987 (next_real_insn, remove_insn): Strengthen param from "rtx" to
989 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
990 from "rtx" to "rtx_insn *".
992 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
994 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
995 bodies streamed in with -Q.
996 * dumpfile.c (dump_files): Add lto-stream-out dump file.
997 * dumpfile.h (tree_dump_index): Add lto_stream_out.
998 * gimple-streamer-out.c: Include gimple-pretty-print.h
999 (output_bb): Dump stmts streamed.
1000 * lto-section-out.c: Include print-tree.h
1001 (lto_begin_section): Dump sections created.
1002 (lto_output_decl_index): Dump decl encoded.
1003 * lto-streamer-out.c: Include print-tree.h
1004 (create_output_block): Dump output block created.
1005 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
1006 (output_function): Dump function output.
1007 (output_constructor): Dump constructor streamed.
1008 (write_global_stream): Output indexes encoded.
1009 (produce_asm_for_decls): Dump streams encoded.
1010 * lto-streamer.c (streamer_dump_file): New global var.
1011 * lto-streamer.h (streamer_dump_file): Declare.
1012 * passes.c (ipa_write_summaries): Initialize streamer dump.
1013 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
1016 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
1019 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
1020 offsets for register save directives. Emit a second batch of save
1021 directives, if need be, when the function accesses prior frames.
1023 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1025 * config/arc/fpu.md (fmasf4): Force operand to register.
1026 (fnmasf4): Likewise.
1028 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1030 * config/arc/arc-protos.h (arc_pad_return): Remove.
1031 * config/arc/arc.c (machine_function): Remove force_short_suffix
1033 (arc_print_operand): Adjust printing of '&'.
1034 (arc_verify_short): Remove conditional printing of short suffix.
1035 (arc_final_prescan_insn): Remove reference to size_reason.
1036 (pad_return): New function.
1037 (arc_reorg): Call pad_return.
1038 (arc_pad_return): Remove.
1039 (arc_init_machine_status): Remove reference to force_short_suffix.
1040 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
1041 (attr length): When attribute iscompact is true force to 2
1042 regardless; in the case of maybe check if we want to force the
1043 instruction to have 4 bytes length.
1044 (nopv): Change it to generate 4 byte long nop as well.
1045 (blockage): New pattern.
1046 (simple_return): Remove call to arc_pad_return.
1047 (p_return_i): Likewise.
1049 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1051 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
1053 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1055 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
1058 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1060 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
1061 for ARC700 and ARCv2.
1063 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
1066 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
1067 operands[2] instead of operands[1].
1070 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1072 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
1073 case, check whether the outer register overlaps an unallocatable
1074 register, not just whether it fits the required class.
1076 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1078 * poly-int.h (can_div_trunc_p): Add new overload in which all values
1080 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
1081 (memrefs_conflict_p): Likewise.
1082 (init_alias_analysis): Likewise.
1083 * cfgexpand.c (expand_debug_expr): Likewise.
1084 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
1085 * cse.c (fold_rtx): Likewise.
1086 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
1087 * expr.c (emit_block_move_hints): Likewise.
1088 (clear_storage_hints, push_block, emit_push_insn): Likewise.
1089 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
1090 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
1091 (emit_group_store): Likewise.
1092 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
1093 to read the PRE/POST_MODIFY increment.
1094 * calls.c (store_one_arg): Use strip_offset.
1095 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
1097 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
1099 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
1100 (simplify_binary_operation_1): Extend CONST_INT handling to
1102 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
1103 than a HOST_WIDE_INT.
1104 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
1106 (adjust_mems, add_stores): Update accodingly.
1107 (vt_canonicalize_addr): Track polynomial offsets.
1108 (emit_note_insn_var_location): Likewise.
1109 (vt_add_function_parameter): Likewise.
1110 (vt_initialize): Likewise.
1112 2018-06-12 Jeff Law <law@redhat.com>
1114 * config.gcc (alpha*-*-freebsd*): Remove.
1115 * config/alpha/freebsd.h: Remove.
1117 2018-06-12 David Malcolm <dmalcolm@redhat.com>
1120 * spellcheck-tree.c (levenshtein_distance): Rename to...
1121 (get_edit_distance): ...this, and update for underlying renaming.
1122 * spellcheck-tree.h (levenshtein_distance): Rename to...
1123 (get_edit_distance): ...this.
1124 * spellcheck.c (levenshtein_distance): Rename to...
1125 (get_edit_distance): ...this. Convert from Levenshtein distance
1126 to Damerau-Levenshtein distance by supporting transpositions of
1127 adjacent characters. Rename "v1" to "v_next" and "v0" to
1129 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
1130 (selftest::test_edit_distance_unit_test_oneway): ...this, and
1131 update for underlying renaming.
1132 (selftest::levenshtein_distance_unit_test): Rename to...
1133 (selftest::test_get_edit_distance_unit): ...this, and update for
1134 underlying renaming.
1135 (selftest::test_find_closest_string): Add example from PR 69968
1136 where transposition helps
1137 (selftest::test_metric_conditions): Update for renaming.
1138 (selftest::test_metric_conditions): Likewise.
1139 (selftest::spellcheck_c_tests): Likewise.
1140 * spellcheck.h (levenshtein_distance): Rename both overloads to...
1141 (get_edit_distance): ...this.
1142 (best_match::consider): Update for renaming.
1144 2018-06-12 Martin Sebor <msebor@redhat.com>
1146 PR tree-optimization/85259
1147 * builtins.c (compute_objsize): Handle constant offsets.
1148 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
1149 true iff a warning has been issued.
1150 * gimple.h (gimple_nonartificial_location): New function.
1151 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
1152 gimple_nonartificial_location and handle -Wno-system-headers.
1153 (handle_builtin_stxncpy): Same.
1155 2018-06-12 Martin Sebor <msebor@redhat.com>
1158 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
1160 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
1162 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
1163 BUILTIN_VEC_XST entries for pointer to double and long long.
1165 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
1168 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1170 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1173 2018-06-12 Martin Liska <mliska@suse.cz>
1175 * doc/options.texi: Document IntegerRange.
1177 2018-06-12 Martin Liska <mliska@suse.cz>
1179 * config/i386/i386.opt: Make MPX-related options as Deprecated.
1180 * opt-functions.awk: Handle Deprecated flag.
1181 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
1183 (read_cmdline_option): Report warning for a deprecated option.
1184 * opts.h (struct cl_option): Add new field cl_deprecated.
1185 (CL_ERR_DEPRECATED): New.
1187 2018-06-12 Martin Liska <mliska@suse.cz>
1189 * doc/options.texi: Document Deprecated option flag.
1191 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1193 * config/arc/arc-arch.h (arc_extras): New enum.
1194 (arc_cpu_t):Add field extra.
1195 (arc_cpu_types): Consider the extras.
1196 * config/arc/arc-cpus.def: Add extras info.
1197 * config/arc/arc-opts.h (processor_type): Consider extra field.
1198 * config/arc/arc.c (arc_override_options): Handle extra field.
1200 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1202 * config/arc/arc-arch.h: Update ARC_OPTX macro.
1203 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
1205 * config/arc/arc.c (arc_init): Update pic warning.
1206 (irq_range): Update irq range parsing warnings.
1207 (arc_override_options): Update various warning messages.
1208 (arc_handle_aux_attribute): Likewise.
1210 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
1212 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
1214 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1216 * doc/sourcebuild.texi: Document usage of line number 0 in verify
1217 compiler messages directives.
1219 2018-06-12 Matthew Fortune <mfortune@gmail.com>
1221 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
1222 * config/mips/mips-tables.opt: Regenerate.
1223 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
1225 * doc/invoke.texi: Document -march=i6500.
1227 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
1229 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
1230 (i6400_gpmul): Add cpu_unit.
1231 (i6400_gpdiv): Likewise.
1232 (i6400_msa_add_d): Update reservations.
1233 (i6400_msa_int_add) Likewise.
1234 (i6400_msa_short_logic3) Likewise.
1235 (i6400_msa_short_logic2) Likewise.
1236 (i6400_msa_short_logic) Likewise.
1237 (i6400_msa_move) Likewise.
1238 (i6400_msa_cmp) Likewise.
1239 (i6400_msa_short_float2) Likewise.
1240 (i6400_msa_div_d) Likewise.
1241 (i6400_msa_long_logic1) Likewise.
1242 (i6400_msa_long_logic2) Likewise.
1243 (i6400_msa_mult) Likewise.
1244 (i6400_msa_long_float2) Likewise.
1245 (i6400_msa_long_float4) Likewise.
1246 (i6400_msa_long_float5) Likewise.
1247 (i6400_msa_long_float8) Likewise.
1248 (i6400_fpu_fadd): Include frint type.
1249 (i6400_fpu_store): New define_insn_reservation.
1250 (i6400_fpu_load): Likewise.
1251 (i6400_fpu_move): Likewise.
1252 (i6400_fpu_fcmp): Likewise.
1253 (i6400_fpu_fmadd): Likewise.
1254 (i6400_int_mult): Include imul3nc type and update reservation.
1255 (i6400_int_div): Include idiv3 type and update reservation.
1256 (i6400_int_load): Update to check type not move_type.
1257 (i6400_int_store): Likewise.
1258 (i6400_int_prefetch): Set zero latency.
1260 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
1262 * gcc.c: Document new %@{...} sequence.
1263 (LINK_COMMAND_SPEC): Use it for the -L switches.
1264 (cpp_unique_options): Use it for the -I switches.
1265 (at_file_argbuf): New global variable.
1266 (in_at_file): Likewise.
1267 (alloc_args): Create at_file_argbuf.
1268 (clear_args): Truncate at_file_argbuf.
1269 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
1270 (open_at_file): New function.
1271 (close_at_file): Likewise.
1272 (create_at_file): Delete.
1273 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
1276 (validate_switches_from_spec): Deal with %@{...} sequence.
1277 (validate_switches): Likewise.
1278 (driver::finalize): Call clear_args.
1280 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
1282 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
1284 2018-06-11 Martin Sebor <msebor@redhat.com>
1286 * doc/invoke.texi (-Wall): List -Wc++17-compat.
1287 (Wno-class-memaccess): Add @opindex.
1288 (Wno-templates, Wno-multiple-inheritance): Same.
1289 (Wno-virtual-inheritance, Wno-namespaces): Same.
1290 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
1291 (Wno-format-overflow, Wno-format-truncation): Same.
1292 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
1293 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
1294 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
1295 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
1296 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
1297 (Wno-misspelled-isr): Same.
1299 2018-06-11 Martin Sebor <msebor@redhat.com>
1301 * PR tree-optimization/86083
1302 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
1304 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
1306 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
1308 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
1311 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
1312 on the correct operand.
1313 (*movdi_internal64): Ditto.
1315 2018-06-11 Martin Liska <mliska@suse.cz>
1317 PR tree-optimization/86089
1318 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
1320 2018-06-11 Julia Koval <julia.koval@intel.com>
1322 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
1323 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
1324 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
1326 2018-06-11 Olivier Hainque <hainque@adacore.com>
1328 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
1329 for Ada with strict dwarf2.
1331 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
1334 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
1337 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
1339 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
1341 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
1343 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
1346 2018-06-08 Martin Liska <mliska@suse.cz>
1348 * tree-cfg.h (debug_function): Fix argument type to match
1351 2018-06-08 Martin Liska <mliska@suse.cz>
1353 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
1354 Remove usage of MPX-related (and removed) fields.
1355 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
1357 2018-06-08 David Malcolm <dmalcolm@redhat.com>
1359 * cfg.c (debug): Use TDF_NONE rather than 0.
1360 * cfghooks.c (debug): Likewise.
1361 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
1362 (struct dump_option_value_info): Convert to...
1363 (struct kv_pair): ...this template type.
1364 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
1366 (optinfo_verbosity_options): Likewise.
1367 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
1369 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
1370 than int for "optgroup_flags" param.
1371 (dump_generic_expr_loc): Use dump_flags_t rather than int for
1373 (dump_dec): Likewise.
1374 (dump_finish): Use TDF_NONE rather than 0.
1375 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
1376 rather than int for "optgroup_flags" param. Use TDF_NONE rather
1377 than 0. Update for change to option_ptr.
1378 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
1379 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
1380 0. Update for changes to optinfo_verbosity_options and
1382 (opt_info_switch_p): Convert optgroup_flags from int to
1384 (dump_basic_block): Use dump_flags_t rather than int
1385 for "dump_kind" param.
1386 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
1387 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
1388 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
1389 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
1390 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
1391 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
1392 TDF_NONE): Convert from macros to...
1393 (enum dump_flag): ...this new enum.
1394 (dump_flags_t): Update to use enum.
1395 (operator|, operator&, operator~, operator|=, operator&=):
1396 Implement for dump_flags_t.
1397 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
1398 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
1399 Convert from macros to...
1400 (enum optgroup_flag): ...this new enum.
1401 (optgroup_flags_t): New typedef.
1402 (operator|, operator|=): Implement for optgroup_flags_t.
1403 (struct dump_file_info): Convert field "alt_flags" to
1404 dump_flags_t. Convert field "optgroup_flags" to
1406 (dump_basic_block): Use dump_flags_t rather than int for param.
1407 (dump_generic_expr_loc): Likewise.
1408 (dump_dec): Likewise.
1409 (dump_register): Convert param "optgroup_flags" to
1411 (opt_info_enable_passes): Likewise.
1412 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
1414 * gimple-pretty-print.c (debug): Likewise.
1415 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
1416 (merged_store_group::apply_stores): Likewise.
1417 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
1418 * gimple.c (verify_gimple_pp): Likewise.
1419 * graphite-poly.c (print_pbb_body): Likewise.
1420 * passes.c (pass_manager::register_one_dump_file): Convert
1421 local "optgroup_flags" to optgroup_flags_t.
1422 * print-tree.c (print_node): Use TDF_NONE rather than 0.
1424 (debug_body): Likewise.
1425 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
1426 to optgroup_flags_t.
1427 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
1429 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
1430 (convert_mult_to_fma): Likewise.
1431 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
1432 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
1433 * tree-vect-data-refs.c (dump_lower_bound): Convert param
1434 "dump_kind" to dump_flags_t.
1436 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1438 * config/rs6000/rs6000.c (min, max): Delete.
1440 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1442 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
1443 -mabi=spe and -mabi=no-spe.
1445 2018-06-08 Martin Liska <mliska@suse.cz>
1447 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
1448 where we expect an existing summary.
1450 2018-06-08 Martin Liska <mliska@suse.cz>
1452 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
1453 * ipa-inline.h (estimate_edge_growth): Likewise.
1455 2018-06-08 Martin Liska <mliska@suse.cz>
1457 * cgraph.c (function_version_hasher::hash): Use
1458 cgraph_node::get_uid ().
1459 (function_version_hasher::equal):
1460 * cgraph.h (cgraph_node::get_uid): New method.
1461 * ipa-inline.c (update_caller_keys): Use
1462 cgraph_node::get_uid ().
1463 (update_callee_keys): Likewise.
1464 * ipa-utils.c (searchc): Likewise.
1465 (ipa_reduced_postorder): Likewise.
1466 * lto-cgraph.c (input_node): Likewise.
1467 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
1468 * symbol-summary.h (symtab_insertion): Likewise.
1469 (symtab_removal): Likewise.
1470 (symtab_duplication): Likewise.
1471 * tree-pretty-print.c (dump_function_header): Likewise.
1472 * tree-sra.c (convert_callers_for_node): Likewise.
1474 2018-06-08 Martin Liska <mliska@suse.cz>
1476 * cgraph.c (symbol_table::create_edge): Always assign a new
1478 (symbol_table::free_edge): Do not recycle numbers.
1479 * cgraph.h (cgraph_edge::get): New method.
1480 * symbol-summary.h (symtab_removal): Use it.
1481 (symtab_duplication): Likewise.
1482 (call_summary::hashable_uid): Remove.
1484 2018-06-08 Martin Liska <mliska@suse.cz>
1486 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
1487 (initialize_growth_caches): Remove.
1488 (free_growth_caches): Likewise.
1489 (do_estimate_edge_time): Use edge_growth_cache.
1490 (do_estimate_edge_size): Likewise.
1491 (do_estimate_edge_hints): Likewise.
1492 * ipa-inline.c (reset_edge_caches): Likewise.
1493 (recursive_inlining): Likewise.
1494 (inline_small_functions): Likewise.
1495 * ipa-inline.h (initialize_growth_caches): Remove.
1496 (estimate_edge_size): Likewise.
1497 (estimate_edge_time): Likewise.
1498 (estimate_edge_hints): Likewise.
1499 (reset_edge_growth_cache): Likewise.
1500 * symbol-summary.h (call_summary::remove): New method.
1502 2018-06-08 Martin Liska <mliska@suse.cz>
1504 * ipa-cp.c (class edge_clone_summary): New summary.
1505 (grow_edge_clone_vectors): Remove.
1506 (ipcp_edge_duplication_hook): Remove.
1507 (class edge_clone_summary_t): New call_summary class.
1508 (ipcp_edge_removal_hook): Remove.
1509 (edge_clone_summary_t::duplicate): New function.
1510 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
1511 (create_specialized_node): Likewise.
1512 (ipcp_driver): Initialize edge_clone_summaries and do not
1515 2018-06-08 Martin Liska <mliska@suse.cz>
1517 * symbol-summary.h (get): New function.
1518 (call_summary::m_initialize_when_cloning): New class member.
1520 2018-06-08 Martin Liska <mliska@suse.cz>
1522 * cgraph.c (cgraph_node::remove): Do not recycle uid.
1523 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
1524 (symbol_table::allocate_cgraph_symbol): Do not set uid.
1525 * passes.c (uid_hash_t): Record removed_nodes by their uids.
1526 (remove_cgraph_node_from_order): Use the removed_nodes set.
1527 (do_per_function_toporder): Likwise.
1528 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
1529 instead of summary_uid.
1530 (symtab_removal): Likewise.
1531 (symtab_duplication): Likewise.
1533 2018-06-08 Martin Liska <mliska@suse.cz>
1535 * ipa-cp.c (ipcp_store_bits_results): Use
1536 ipcp_transformation_sum.
1537 (ipcp_store_vr_results): Likewise.
1538 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
1540 (ipcp_transformation_initialize): ... this.
1541 (ipa_set_node_agg_value_chain):
1542 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
1543 (write_ipcp_transformation_info): Likewise.
1544 (read_ipcp_transformation_info): Likewise.
1545 (ipcp_update_bits): Likewise.
1546 (ipcp_update_vr): Likewise.
1547 (ipcp_transform_function): Likewise.
1548 * ipa-prop.h: Rename ipcp_transformation_summary to
1549 ipcp_transformation.
1550 (class ipcp_transformation_t): New function summary.
1551 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
1552 (ipa_get_agg_replacements_for_node): Likewise.
1554 2018-06-08 Martin Liska <mliska@suse.cz>
1556 * ipa-pure-const.c (struct funct_state_d): Do it class instead
1558 (class funct_state_summary_t): New function_summary class.
1559 (has_function_state): Remove.
1560 (get_function_state): Likewise.
1561 (set_function_state): Likewise.
1562 (add_new_function): Likewise.
1563 (funct_state_summary_t::insert): New function.
1564 (duplicate_node_data): Remove.
1565 (remove_node_data): Remove.
1566 (funct_state_summary_t::duplicate): New function.
1567 (register_hooks): Create new funct_state_summaries.
1568 (pure_const_generate_summary): Use it.
1569 (pure_const_write_summary): Likewise.
1570 (pure_const_read_summary): Likewise.
1571 (propagate_pure_const): Likewise.
1572 (propagate_nothrow): Likewise.
1573 (dump_malloc_lattice): Likewise.
1574 (propagate_malloc): Likewise.
1575 (execute): Do not register hooks, just remove summary
1577 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
1580 2018-06-08 Martin Liska <mliska@suse.cz>
1582 * ipa-reference.c (remove_node_data): Remove.
1583 (duplicate_node_data): Likewise.
1584 (class ipa_ref_var_info_summary_t): New class.
1585 (class ipa_ref_opt_summary_t): Likewise.
1586 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
1587 (get_reference_optimization_summary): Use
1588 ipa_ref_opt_sum_summaries.
1589 (set_reference_vars_info): Remove.
1590 (set_reference_optimization_summary): Likewise.
1591 (ipa_init): Create summaries.
1592 (init_function_info): Use function summary.
1593 (ipa_ref_opt_summary_t::duplicate): New function.
1594 (ipa_ref_opt_summary_t::remove): New function.
1595 (get_read_write_all_from_node): Fix GNU coding style.
1596 (propagate): Use function summary.
1597 (write_node_summary_p): Fix GNU coding style.
1598 (stream_out_bitmap): Likewise.
1599 (ipa_reference_read_optimization_summary): Use function summary.
1600 (ipa_reference_c_finalize): Do not release hooks.
1602 2018-06-08 Martin Liska <mliska@suse.cz>
1604 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
1605 (analyze_function_body): Extract multiple calls of get_create.
1606 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1607 * ipa-inline.c (recursive_inlining): Use ::get method.
1608 * ipa-inline.h (estimate_edge_growth): Likewise.
1610 2018-06-08 Martin Liska <mliska@suse.cz>
1612 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
1614 (hsa_function_summary::hsa_function_summary): Use the new enum
1616 (hsa_gpu_implementation_p): Use hsa_summaries::get.
1617 * hsa-gen.c (hsa_get_host_function): Likewise.
1618 (get_brig_function_name): Likewise.
1619 * ipa-hsa.c (process_hsa_functions): Likewise.
1620 (ipa_hsa_write_summary): Likewise.
1621 * symbol-summary.h (symtab_duplication): Use ::get function/
1622 (get): New function.
1624 2018-06-08 Martin Liska <mliska@suse.cz>
1626 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
1628 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
1629 (hsa_register_kernel): Likewise.
1630 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
1631 * hsa-gen.c (hsa_get_host_function): Likewise.
1632 (get_brig_function_name): Likewise.
1633 (generate_hsa): Likewise.
1634 (pass_gen_hsail::execute): Likewise.
1635 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
1636 (devirtualization_time_bonus): Likewise.
1637 (ipcp_propagate_stage): Likewise.
1638 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1639 (edge_set_predicate): Likewise.
1640 (evaluate_conditions_for_known_args): Likewise.
1641 (evaluate_properties_for_edge): Likewise.
1642 (ipa_fn_summary::reset): Likewise.
1643 (ipa_fn_summary_t::duplicate): Likewise.
1644 (dump_ipa_call_summary): Likewise.
1645 (ipa_dump_fn_summary): Likewise.
1646 (analyze_function_body): Likewise.
1647 (compute_fn_summary): Likewise.
1648 (estimate_edge_devirt_benefit): Likewise.
1649 (estimate_edge_size_and_time): Likewise.
1650 (estimate_calls_size_and_time): Likewise.
1651 (estimate_node_size_and_time): Likewise.
1652 (inline_update_callee_summaries): Likewise.
1653 (remap_edge_change_prob): Likewise.
1654 (remap_edge_summaries): Likewise.
1655 (ipa_merge_fn_summary_after_inlining): Likewise.
1656 (ipa_update_overall_fn_summary): Likewise.
1657 (read_ipa_call_summary): Likewise.
1658 (inline_read_section): Likewise.
1659 (write_ipa_call_summary): Likewise.
1660 (ipa_fn_summary_write): Likewise.
1661 (ipa_free_fn_summary): Likewise.
1662 * ipa-hsa.c (process_hsa_functions): Likewise.
1663 (ipa_hsa_write_summary): Likewise.
1664 (ipa_hsa_read_section): Likewise.
1665 * ipa-icf.c (sem_function::merge): Likewise.
1666 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1667 (do_estimate_edge_time): Likewise.
1668 (estimate_size_after_inlining): Likewise.
1669 (estimate_growth): Likewise.
1670 (growth_likely_positive): Likewise.
1671 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1672 (inline_call): Likewise.
1673 * ipa-inline.c (caller_growth_limits): Likewise.
1674 (can_inline_edge_p): Likewise.
1675 (can_inline_edge_by_limits_p): Likewise.
1676 (compute_uninlined_call_time): Likewise.
1677 (compute_inlined_call_time): Likewise.
1678 (want_inline_small_function_p): Likewise.
1679 (edge_badness): Likewise.
1680 (update_caller_keys): Likewise.
1681 (update_callee_keys): Likewise.
1682 (recursive_inlining): Likewise.
1683 (inline_small_functions): Likewise.
1684 (inline_to_all_callers_1): Likewise.
1685 (dump_overall_stats): Likewise.
1686 (early_inline_small_functions): Likewise.
1687 (early_inliner): Likewise.
1688 * ipa-inline.h (estimate_edge_growth): Likewise.
1689 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1690 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1691 * ipa-prop.h (IPA_NODE_REF): Likewise.
1692 (IPA_EDGE_REF): Likewise.
1693 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1694 (propagate_malloc): Likewise.
1695 * ipa-split.c (execute_split_functions): Likewise.
1696 * symbol-summary.h: Rename get to get_create.
1698 (get_create): Likewise.
1699 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1701 2018-06-08 Martin Liska <mliska@suse.cz>
1703 * symbol-summary.h (release): Move definition out of class
1705 (symtab_removal): Likewise.
1706 (symtab_duplication): Likewise.
1708 2018-06-08 Martin Liska <mliska@suse.cz>
1710 * symbol-summary.h (function_summary): Move constructor
1711 implementation out of class declaration.
1712 (release): Likewise.
1713 (symtab_insertion): Likewise.
1714 (symtab_removal): Likewise.
1715 (symtab_duplication): Likewise.
1718 2018-06-08 Martin Liska <mliska@suse.cz>
1720 * Makefile.in: Remove support for MPX (macros, related functions,
1721 fields in cgraph_node, ...).
1722 * builtin-types.def (BT_BND): Likewise.
1723 (BT_FN_BND_CONST_PTR): Likewise.
1724 (BT_FN_CONST_PTR_BND): Likewise.
1725 (BT_FN_VOID_PTR_BND): Likewise.
1726 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
1727 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
1728 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
1729 (expand_builtin_mempcpy_with_bounds): Likewise.
1730 (expand_builtin_memset_with_bounds): Likewise.
1731 (expand_builtin_memset_args): Likewise.
1732 (std_expand_builtin_va_start): Likewise.
1733 (expand_builtin): Likewise.
1734 (expand_builtin_with_bounds): Likewise.
1735 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
1736 (DEF_LIB_BUILTIN_CHKP): Likewise.
1737 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
1738 (DEF_CHKP_BUILTIN): Likewise.
1739 (BUILT_IN_MEMCPY): Likewise.
1740 (BUILT_IN_MEMMOVE): Likewise.
1741 (BUILT_IN_MEMPCPY): Likewise.
1742 (BUILT_IN_MEMSET): Likewise.
1743 (BUILT_IN_STPCPY): Likewise.
1744 (BUILT_IN_STRCAT): Likewise.
1745 (BUILT_IN_STRCHR): Likewise.
1746 (BUILT_IN_STRCPY): Likewise.
1747 (BUILT_IN_STRLEN): Likewise.
1748 (BUILT_IN_MEMCPY_CHK): Likewise.
1749 (BUILT_IN_MEMMOVE_CHK): Likewise.
1750 (BUILT_IN_MEMPCPY_CHK): Likewise.
1751 (BUILT_IN_MEMSET_CHK): Likewise.
1752 (BUILT_IN_STPCPY_CHK): Likewise.
1753 (BUILT_IN_STRCAT_CHK): Likewise.
1754 (BUILT_IN_STRCPY_CHK): Likewise.
1755 * calls.c (store_bounds): Likewise.
1756 (emit_call_1): Likewise.
1757 (special_function_p): Likewise.
1758 (maybe_warn_nonstring_arg): Likewise.
1759 (initialize_argument_information): Likewise.
1760 (finalize_must_preallocate): Likewise.
1761 (compute_argument_addresses): Likewise.
1762 (expand_call): Likewise.
1763 * cfgexpand.c (expand_call_stmt): Likewise.
1764 (expand_return): Likewise.
1765 (expand_gimple_stmt_1): Likewise.
1766 (pass_expand::execute): Likewise.
1767 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
1768 (cgraph_node::remove): Likewise.
1769 (cgraph_node::dump): Likewise.
1770 (cgraph_node::verify_node): Likewise.
1771 * cgraph.h (chkp_function_instrumented_p): Likewise.
1772 (symtab_node::get_alias_target): Likewise.
1773 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
1774 (cgraph_local_p): Likewise.
1775 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
1776 (cgraph_edge::rebuild_references): Likewise.
1777 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
1778 (walk_polymorphic_call_targets): Likewise.
1779 (cgraph_node::expand_thunk): Likewise.
1780 (symbol_table::output_weakrefs): Likewise.
1781 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
1782 (ix86_handle_option): Likewise.
1783 * config/i386/constraints.md: Likewise.
1784 * config/i386/i386-builtin-types.def (BND): Likewise.
1788 * config/i386/i386-builtin.def (BDESC_END): Likewise.
1789 (BDESC_FIRST): Likewise.
1791 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
1792 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
1793 * config/i386/i386.c (enum reg_class): Likewise.
1794 (ix86_target_string): Likewise.
1795 (ix86_option_override_internal): Likewise.
1796 (ix86_conditional_register_usage): Likewise.
1797 (ix86_valid_target_attribute_inner_p): Likewise.
1798 (ix86_set_indirect_branch_type): Likewise.
1799 (ix86_set_current_function): Likewise.
1800 (ix86_function_arg_regno_p): Likewise.
1801 (init_cumulative_args): Likewise.
1802 (ix86_function_arg_advance): Likewise.
1803 (ix86_function_arg): Likewise.
1804 (ix86_pass_by_reference): Likewise.
1805 (ix86_function_value_regno_p): Likewise.
1806 (ix86_function_value_1): Likewise.
1807 (ix86_function_value_bounds): Likewise.
1808 (ix86_return_in_memory): Likewise.
1809 (ix86_setup_incoming_vararg_bounds): Likewise.
1810 (ix86_va_start): Likewise.
1811 (indirect_thunk_need_prefix): Likewise.
1812 (print_reg): Likewise.
1813 (ix86_print_operand): Likewise.
1814 (ix86_expand_call): Likewise.
1815 (ix86_output_function_return): Likewise.
1816 (reg_encoded_number): Likewise.
1817 (BDESC_VERIFYS): Likewise.
1818 (ix86_init_mpx_builtins): Likewise.
1819 (ix86_init_builtins): Likewise.
1820 (ix86_emit_cmove): Likewise.
1821 (ix86_emit_move_max): Likewise.
1822 (ix86_expand_builtin): Likewise.
1823 (ix86_builtin_mpx_function): Likewise.
1824 (ix86_get_arg_address_for_bt): Likewise.
1825 (ix86_load_bounds): Likewise.
1826 (ix86_store_bounds): Likewise.
1827 (ix86_load_returned_bounds): Likewise.
1828 (ix86_store_returned_bounds): Likewise.
1829 (ix86_class_likely_spilled_p): Likewise.
1830 (ix86_hard_regno_mode_ok): Likewise.
1831 (x86_order_regs_for_local_alloc): Likewise.
1832 (ix86_mitigate_rop): Likewise.
1833 (ix86_bnd_prefixed_insn_p): Likewise.
1834 (ix86_mpx_bound_mode): Likewise.
1835 (ix86_make_bounds_constant): Likewise.
1836 (ix86_initialize_bounds): Likewise.
1837 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
1838 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
1839 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
1840 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
1841 (TARGET_CHKP_BOUND_MODE): Likewise.
1842 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
1843 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
1844 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
1845 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
1846 * config/i386/i386.h (TARGET_MPX): Likewise.
1847 (TARGET_MPX_P): Likewise.
1848 (VALID_BND_REG_MODE): Likewise.
1849 (FIRST_BND_REG): Likewise.
1850 (LAST_BND_REG): Likewise.
1851 (enum reg_class): Likewise.
1852 (BND_REG_P): Likewise.
1853 (BND_REGNO_P): Likewise.
1854 (BNDmode): Likewise.
1855 (ADJUST_INSN_LENGTH): Likewise.
1856 * config/i386/i386.md: Likewise.
1857 * config/i386/i386.opt: Likewise.
1858 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
1859 (defined): Likewise.
1860 (LINK_MPX): Likewise.
1861 (MPX_SPEC): Likewise.
1862 (LIBMPX_SPEC): Likewise.
1863 (LIBMPXWRAPPERS_SPEC): Likewise.
1864 (CHKP_SPEC): Likewise.
1865 * config/i386/predicates.md: Likewise.
1866 * dbxout.c (dbxout_type): Likewise.
1867 * doc/extend.texi: Likewise.
1868 * doc/invoke.texi: Likewise.
1869 * doc/md.texi: Likewise.
1870 * doc/tm.texi: Likewise.
1871 * doc/tm.texi.in: Likewise.
1872 * dwarf2out.c (is_base_type): Likewise.
1873 (gen_formal_types_die): Likewise.
1874 (gen_subprogram_die): Likewise.
1875 (gen_type_die_with_usage): Likewise.
1876 (gen_decl_die): Likewise.
1877 (dwarf2out_late_global_decl): Likewise.
1878 * expr.c (expand_assignment): Likewise.
1879 (emit_storent_insn): Likewise.
1880 (store_expr_with_bounds): Likewise.
1881 (store_expr): Likewise.
1882 (expand_expr_real_1): Likewise.
1883 * expr.h (store_expr_with_bounds): Likewise.
1884 * function.c (use_register_for_decl): Likewise.
1885 (struct bounds_parm_data): Likewise.
1886 (assign_parms_augmented_arg_list): Likewise.
1887 (assign_parm_find_entry_rtl): Likewise.
1888 (assign_parm_is_stack_parm): Likewise.
1889 (assign_parm_load_bounds): Likewise.
1890 (assign_bounds): Likewise.
1891 (assign_parms): Likewise.
1892 (expand_function_start): Likewise.
1893 * gcc.c (CHKP_SPEC): Likewise.
1894 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1895 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
1896 (wrestrict_dom_walker::check_call): Likewise.
1897 * gimple.c (gimple_build_call_from_tree): Likewise.
1898 * gimple.h (enum gf_mask): Likewise.
1899 (gimple_call_with_bounds_p): Likewise.
1900 (gimple_call_set_with_bounds): Likewise.
1901 * gimplify.c (gimplify_init_constructor): Likewise.
1902 * ipa-cp.c (initialize_node_lattices): Likewise.
1903 (propagate_constants_across_call): Likewise.
1904 (find_more_scalar_values_for_callers_subset): Likewise.
1905 * ipa-hsa.c (process_hsa_functions): Likewise.
1906 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
1907 * ipa-icf.c (sem_function::merge): Likewise.
1908 * ipa-inline.c (early_inliner): Likewise.
1909 * ipa-pure-const.c (warn_function_noreturn): Likewise.
1910 (warn_function_cold): Likewise.
1911 (propagate_pure_const): Likewise.
1912 * ipa-ref.h (enum GTY): Likewise.
1913 * ipa-split.c (find_retbnd): Likewise.
1914 (consider_split): Likewise.
1915 (split_function): Likewise.
1916 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1917 * ipa.c (walk_polymorphic_call_targets): Likewise.
1918 (symbol_table::remove_unreachable_nodes): Likewise.
1919 (process_references): Likewise.
1920 (cgraph_build_static_cdtor_1): Likewise.
1921 * lto-cgraph.c (lto_output_node): Likewise.
1922 (output_refs): Likewise.
1923 (compute_ltrans_boundary): Likewise.
1924 (input_overwrite_node): Likewise.
1925 (input_node): Likewise.
1926 (input_cgraph_1): Likewise.
1927 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
1928 * passes.c (pass_manager::execute_early_local_passes): Likewise.
1929 (class pass_chkp_instrumentation_passes): Likewise.
1930 (make_pass_chkp_instrumentation_passes): Likewise.
1931 * passes.def: Likewise.
1932 * rtl.h (struct GTY): Likewise.
1933 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
1934 * stor-layout.c (layout_type): Likewise.
1935 * symtab.c: Likewise.
1936 * target.def: Likewise.
1937 * targhooks.c (default_chkp_bound_type): Likewise.
1938 (default_chkp_bound_mode): Likewise.
1939 (default_builtin_chkp_function): Likewise.
1940 (default_chkp_function_value_bounds): Likewise.
1941 (default_chkp_make_bounds_constant): Likewise.
1942 (default_chkp_initialize_bounds): Likewise.
1943 * targhooks.h (default_chkp_bound_type): Likewise.
1944 (default_chkp_bound_mode): Likewise.
1945 (default_builtin_chkp_function): Likewise.
1946 (default_chkp_function_value_bounds): Likewise.
1947 (default_chkp_make_bounds_constant): Likewise.
1948 (default_chkp_initialize_bounds): Likewise.
1949 * toplev.c (compile_file): Likewise.
1950 (process_options): Likewise.
1951 * tree-core.h (DEF_BUILTIN): Likewise.
1952 (DEF_BUILTIN_CHKP): Likewise.
1953 * tree-inline.c (declare_return_variable): Likewise.
1954 (remap_gimple_stmt): Likewise.
1955 (copy_bb): Likewise.
1956 (initialize_inlined_parameters): Likewise.
1957 (expand_call_inline): Likewise.
1958 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
1959 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
1960 (make_pass_ipa_chkp_produce_thunks): Likewise.
1961 (make_pass_chkp): Likewise.
1962 (make_pass_chkp_opt): Likewise.
1963 (make_pass_chkp_instrumentation_passes): Likewise.
1964 * tree-pretty-print.c (dump_generic_node): Likewise.
1965 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
1966 * tree-ssa-dce.c (propagate_necessity): Likewise.
1967 (eliminate_unnecessary_stmts): Likewise.
1968 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
1969 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1970 * tree-ssa-sccvn.h: Likewise.
1971 * tree-ssa-strlen.c (get_string_length): Likewise.
1972 (valid_builtin_call): Likewise.
1973 (adjust_last_stmt): Likewise.
1974 (handle_builtin_strchr): Likewise.
1975 (handle_builtin_strcpy): Likewise.
1976 (handle_builtin_stxncpy): Likewise.
1977 (handle_builtin_memcpy): Likewise.
1978 (handle_builtin_strcat): Likewise.
1979 (strlen_check_and_optimize_stmt): Likewise.
1980 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
1981 * tree-streamer-in.c: Likewise.
1982 * tree-streamer.c (record_common_node): Likewise.
1983 * tree.c (tree_code_size): Likewise.
1984 (wide_int_to_tree_1): Likewise.
1985 (type_contains_placeholder_1): Likewise.
1986 (build_common_tree_nodes): Likewise.
1987 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
1988 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
1989 (POINTER_BOUNDS_P): Likewise.
1990 (BOUNDED_TYPE_P): Likewise.
1991 (BOUNDED_P): Likewise.
1992 (CALL_WITH_BOUNDS_P): Likewise.
1993 (pointer_bounds_type_node): Likewise.
1994 * value-prof.c (gimple_ic): Likewise.
1995 * var-tracking.c (vt_add_function_parameters): Likewise.
1996 * varasm.c (make_decl_rtl): Likewise.
1997 (assemble_start_function): Likewise.
1998 (output_constant): Likewise.
1999 (maybe_assemble_visibility): Likewise.
2000 * varpool.c (ctor_for_folding): Likewise.
2001 * chkp-builtins.def: Remove.
2002 * ipa-chkp.c: Remove.
2003 * ipa-chkp.h: Remove.
2004 * rtl-chkp.c: Remove.
2005 * rtl-chkp.h: Remove.
2006 * tree-chkp-opt.c: Remove.
2007 * tree-chkp.c: Remove.
2008 * tree-chkp.h: Remove.
2010 2018-06-07 Carl Love <cel@us.ibm.com>
2012 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
2013 vextract_fp_from_shortl): Add BE support.
2015 2018-06-07 Paul Koning <ni1d@arrl.net>
2017 * compare-elim.c (try_merge_compare): Don't merge compare if
2018 address contains a side effect.
2019 (try_eliminate_compare): Likewise.
2021 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
2023 * config.gcc: Support "tremont".
2024 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
2025 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2027 * config/i386/i386.c (m_TREMONT): Define.
2028 (processor_target_table): Add "tremont".
2029 (PTA_TREMONT): Define.
2030 (ix86_lea_outperforms): Add TARGET_TREMONT.
2031 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
2032 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
2033 and M_INTEL_GOLDMONT_PLUS.
2034 (fold_builtin_cpu): Add "tremont".
2035 (ix86_add_stmt_cost): Add TARGET_TREMONT.
2036 (ix86_option_override_internal): Add "tremont".
2037 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
2038 (processor_type): Add PROCESSOR_TREMONT.
2039 * config/i386/x86-tune.def: Add m_TREMONT.
2040 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
2042 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2044 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
2045 symbol defined for msp430i* devices to be lower case.
2047 gcc/testsuite/gcc.target/msp430/
2048 * msp430i-device-symbol.c: New test.
2049 * msp430f-device-symbol.c: New test.
2050 * msp430.h: New test header file.
2052 2018-06-07 Richard Biener <rguenther@suse.de>
2054 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
2055 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
2056 Properly wrap signed arithmetic if overflow wraps.
2058 2018-06-07 Jakub Jelinek <jakub@redhat.com>
2060 PR tree-optimization/69615
2061 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
2062 of a cast from a same precision integral SSA_NAME in a bb dominated
2063 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
2064 cast to utype if rhs2 has already a compatible type.
2066 2018-06-07 Richard Biener <rguenther@suse.de>
2068 PR tree-optimization/85935
2069 * graphite-scop-detection.c (find_params_in_bb): Analyze
2070 condition operands with respect to the correct loop. Assert
2071 the analysis doesn't fail.
2073 2018-06-04 Carl Love <cel@us.ibm.com>
2075 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
2076 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
2077 as it is slightly cheaper.
2078 (first_match_or_eos_index_<mode>):
2079 Calculate index using natural element order.
2080 (first_match_index_<mode>):
2081 Calculate index using natural element order.
2082 (first_match_or_eos_index_<mode>):
2083 Calculate index using natural order.
2084 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
2085 for BE and LE modes.
2086 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
2087 P9V_BUILTIN_VCLZLSBB_V16QI.
2088 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
2091 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2093 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
2094 indentation and line wrap for many prototypes. Add missing
2095 @smallexample directives around block of prototypes for vec_xl and
2098 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
2100 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
2101 track if we pass or return IEEE 128-bit floating point.
2102 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
2103 C++ mangling that is compatible with GCC 8.1.
2104 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
2105 (init_cumulative_args): Note if we pass or return IEEE 128-bit
2106 floating point types.
2107 (rs6000_function_arg_advance_1): Likewise.
2108 (rs6000_mangle_type): Optionally generate mangled names that match
2109 what GCC 8.1 generated for IEEE 128-bit floating point types.
2110 (rs6000_globalize_decl_name): If we have an external function that
2111 passes or returns IEEE 128-bit types, generate a weak reference
2112 from the mangled name used in GCC 8.1 to the current mangled
2114 (rs6000_init_builtins): Make __ibm128 use the long double type if
2115 long double is IBM extended double. Make __float128 use the long
2116 double type if long double is IEEE 128-bit.
2119 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
2120 macro for __ibm128 built-in functions.
2121 (PACK_IF): Add __ibm128 pack/unpack functions.
2122 (UNPACK_IF): Likewise.
2123 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
2124 enable long double built-in functions if long double is IEEE
2125 128-bit floating point.
2126 (rs6000_invalid_builtin): Update long double built-in function
2128 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
2129 functions, adjust the built-in function to use the long double
2130 built-in function if __ibm128 and long double are the same type.
2131 * doc/extend.texi (PowerPC builtins): Update documention for
2132 __builtin_{,un}pack_longdouble. Add documentation for
2133 __builtin_{,un}pack_ibm128.
2135 2018-06-06 Jim Wilson <jimw@sifive.com>
2137 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
2138 (struct machine_function): New field interrupt_mode.
2139 (riscv_handle_type_attribute): New function. Add forward declaration.
2140 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
2141 (riscv_expand_epilogue): Check interrupt_mode field.
2142 (riscv_set_current_function): Check interrupt attribute args and
2143 set interrupt_mode field.
2144 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
2145 (riscv_sret, riscv_uret): New.
2146 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
2147 new arguments to interrupt attribute.
2149 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
2152 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
2153 Don't handle -mcpu=power8 if -mpower9-vector is also used.
2155 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2157 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
2158 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
2159 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2160 several redundant entries.
2162 2018-06-06 David Malcolm <dmalcolm@redhat.com>
2164 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
2165 type from "rtx" to "rtx_insn *".
2166 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
2167 for local "call_insn", removing cast.
2168 (ix86_expand_call): Likewise, introducing a "call_insn" local.
2170 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
2172 PR tree-optimization/86066
2173 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
2174 for BIT_INSERT_EXPR stores.
2176 2018-06-06 Richard Biener <rguenther@suse.de>
2178 PR tree-optimization/86062
2179 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2180 component refs ontop
2181 of to be offsetted base.
2183 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2185 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
2186 to be static and remove check on interrupt attribute name.
2188 gcc/testsuite/gcc.target/msp430/
2189 * function-attributes-4.c: New test.
2190 * static-interrupts.c: New test.
2192 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
2194 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2195 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
2197 2018-06-05 Steve Ellcey <sellcey@cavium.com>
2200 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
2202 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
2203 Remove second argument, change how error is called.
2204 (aarch64_layout_arg): Remove second argument from
2205 aarch64_err_no_fpadvsimd call.
2206 (aarch64_init_cumulative_args): Ditto.
2207 (aarch64_gimplify_va_arg_expr): Ditto.
2208 * config/aarch64/aarch64.md (mov<mode>): Ditto.
2210 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
2212 * config/i386/i386.md (simple_return_indirect_internal): New expander.
2213 (*simple_return_indirect_internal<mode>): Rename from
2214 simple_return_indirect_internal. Use W mode iterator.
2215 (rstorssp): New expander.
2216 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
2217 (clrssbsy): New expander.
2218 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
2220 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
2222 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
2224 (cmse_check_pointed_object): Likewise.
2226 2018-06-05 Martin Liska <mliska@suse.cz>
2228 PR gcov-profile/47618
2229 * doc/invoke.texi: Document how -fprofile-dir format
2232 2018-06-05 Richard Biener <rguenther@suse.de>
2234 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
2235 removal pretend DOM info isn't available so we do not update
2236 it and only remove edges, not dominated blocks. Actually free
2237 DOM info in case we removed something. Remove unreachable blocks.
2238 (mfb_keep_latches): Work with either DOM info or marked backedges.
2239 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
2240 first. Mark backedges if DOM info isn't available.
2241 (Re-)compute DOM info after cleanup_control_flow_pre.
2243 2018-06-05 Richard Biener <rguenther@suse.de>
2245 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
2246 (locus_discrim_hasher::hash): Adjust.
2247 (locus_discrim_hasher::equal): Likewise.
2248 (next_discriminator_for_locus): Work on line directly.
2249 (same_line_p): Pass in expanded locus1 as well.
2250 (assign_discriminators): Avoid redundant location expansions.
2252 2018-06-05 Richard Biener <rguenther@suse.de>
2254 PR tree-optimization/86046
2255 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
2256 if required after clearing TREE_ADDRESSABLE.
2258 2018-06-05 Richard Biener <rguenther@suse.de>
2260 PR tree-optimization/86047
2261 * tree-ssa-loop.c (for_each_index): Glob handling of all
2262 decls and constants and really handle all of them.
2264 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2267 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
2268 qualifier_void_pointer and qualifier_const_void_pointer.
2269 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
2270 (arm_init_builtins): Handle the above.
2271 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
2272 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
2273 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
2276 2018-06-05 Martin Liska <mliska@suse.cz>
2278 * auto-profile.c (read_autofdo_file): Do not use
2279 gcov_ctr_summary struct.
2280 (afdo_callsite_hot_enough_for_early_inline): Likewise.
2281 * coverage.c (struct counts_entry): Likewise.
2282 (read_counts_file): Read just single summary entry.
2283 (get_coverage_counts): Use gcov_summary struct.
2284 * coverage.h (get_coverage_counts): Likewise.
2285 * gcov-dump.c (dump_working_sets): Likewise.
2286 (tag_summary): Dump just single summary.
2287 * gcov-io.c (gcov_write_summary): Write just histogram
2289 (gcov_read_summary): Read just single summary.
2290 (compute_working_sets): Use gcov_summary struct.
2291 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
2292 of GCOV_COUNTERS_SUMMABLE.
2293 (GCOV_COUNTERS_SUMMABLE): Remove.
2294 (GCOV_FIRST_VALUE_COUNTER): Replace with
2295 GCOV_COUNTER_V_INTERVAL.
2296 (struct gcov_ctr_summary): Remove.
2297 (struct gcov_summary): Directly use fields of former
2299 (compute_working_sets): Use gcov_summary struct.
2300 * gcov.c (read_count_file): Do not use ctrs fields.
2301 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
2303 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
2305 * profile.c: Likewise.
2306 * profile.h: Likewise.
2308 2018-06-05 Martin Liska <mliska@suse.cz>
2310 PR gcov-profile/84846
2311 * gcov.c (output_lines): Print working directory only
2312 in intermediate format.
2314 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
2316 * config/s390/s390-builtin-types.def: Add void function type.
2317 * config/s390/s390-builtins.def: Use the function type for the
2320 2018-06-04 Jim Wilson <jimw@sifive.com>
2322 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
2324 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
2325 and maybe_eh_return. Change regno to unsigned int. Use new args to
2326 handle EH_RETURN_DATA_REGNO registers properly.
2327 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
2328 (riscv_expand_epilogue): Update comment. Change argument name and
2329 type. Update code to use new name and type. Pass new args to
2330 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
2332 * config/riscv/riscv.md (NORMAL_RETURN): New.
2333 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
2334 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
2335 (eh_return): Call gen_eh_return_internal and emit barrier.
2336 (eh_return_internal): Call riscv_expand_epilogue.
2338 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
2340 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
2341 bit_insertion field and declare can_be_merged_into method.
2342 (merged_store_group::can_be_merged_into): New method.
2343 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
2344 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
2345 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
2347 2018-06-04 Richard Biener <rguenther@suse.de>
2349 PR tree-optimization/85955
2350 * builtins.c (fold_builtin_sincos): Convert pointers to
2351 destination to appropriate type before dereferencing.
2353 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
2355 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
2357 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
2359 * expr.c (expand_expr_real_1): Force the operand into memory if
2360 its TYPE_MODE is BLKmode and if there is no integer mode for
2361 the number of bits being extracted.
2363 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2367 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2368 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
2370 2018-06-04 Richard Biener <rguenther@suse.de>
2372 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
2373 (cleanup_tree_cfg_noloop): ... single caller. Do
2374 start_recording_case_labels later.
2376 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
2378 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
2379 to _IMMINTRIN_H_INCLUDED.
2380 * config/i386/pconfigintrin.h: Ditto.
2381 * config/i386/waitpkgintrin.h: Ditto.
2382 * config/i386/immintrin.h: Add includes for sgxintrin.h,
2383 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
2384 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
2385 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
2386 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
2387 waitpkgintrin.h and cldemoteintrin.h.
2389 2018-06-04 Richard Biener <rguenther@suse.de>
2391 PR tree-optimization/86038
2392 * tracer.c (find_best_successor): Check probability for
2393 being initialized, bail out if not.
2395 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
2398 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
2399 of bits to ignore when comparing architectures.
2401 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2403 PR tree-optimization/69615
2404 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
2405 maximum or minimum of the type, try to merge it also as if
2406 range1 is + [-, x - 1] or + [x + 1, -].
2409 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
2411 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
2413 PR tree-optimization/86034
2414 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
2415 the unsigned bitfield type in a bit insertion sequence if it does not
2416 have a larger precision than the bitfield size.
2417 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
2419 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
2421 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
2423 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2425 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
2426 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
2427 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
2428 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
2430 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2432 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
2433 Disable -fdelete-null-pointer-checks for ELF toolchain.
2435 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2436 Kito Cheng <kito.cheng@gmail.com>
2438 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
2439 (nds32le-*-*, nds32be-*-*): Integrate checking process.
2440 (nds32*-*-*): Add glibc and uclibc conditions.
2441 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
2442 (TARGET_EXCEPT_UNWIND_INFO): Define.
2443 * config/nds32/elf.h: New file.
2444 * config/nds32/linux.h: New file.
2445 * config/nds32/nds32-elf.opt: New file.
2446 * config/nds32/nds32-linux.opt: New file.
2447 * config/nds32/nds32-fp-as-gp.c
2448 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
2449 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
2451 (nds32_asm_file_end): Ditto.
2452 (nds32_print_operand): Ditto.
2453 (nds32_insert_attributes): Ditto.
2454 (nds32_init_libfuncs): New function.
2455 (TARGET_HAVE_TLS): Define.
2456 (TARGET_INIT_LIBFUNCS): Define.
2457 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
2459 (TARGET_ELF): Apply different mcmodel setting.
2460 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
2461 been migrated into elf.h and linux.h files.
2462 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
2463 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
2464 (mcmodel): The content has been migrated into nds32-elf.opt and
2465 nds32-linux.opt files.
2466 * config/nds32/t-elf: New file.
2467 * config/nds32/t-linux: New file.
2469 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2470 Shiva Chen <shiva0217@gmail.com>
2472 * config/nds32/constants.md (unspec_volatile_element): Add
2473 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
2474 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
2476 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
2477 (make_pass_nds32_fp_as_gp): Declare.
2478 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
2480 (nds32_asm_function_end_prologue): Remove unused asm output.
2481 (nds32_asm_function_begin_epilogue): Remove unused asm output.
2482 (nds32_asm_file_start): Output necessary fp_as_gp information.
2483 (nds32_option_override): Adjust register usage.
2484 (nds32_expand_prologue): Consider fp_as_gp situation.
2485 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
2486 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
2489 (simple_return): Ditto.
2490 (omit_fp_begin): Output special directive for fp_as_gp.
2491 (omit_fp_end): Output special directive for fp_as_gp.
2492 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
2493 mforbid-fp-as-gp): New options.
2495 2018-06-01 Mark Wielaard <mark@klomp.org>
2497 * dwarf2out.c (dwarf2out_finish): Remove generation of
2498 DW_AT_loclists_base.
2500 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
2502 * gimple-ssa-store-merging.c: Include gimple-fold.h.
2503 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
2504 (struct merged_store_group): Add bit_insertion field.
2505 (dump_char_array): Use standard hexadecimal format.
2506 (merged_store_group::merged_store_group): Set bit_insertion to false.
2507 (merged_store_group::apply_stores): Use optimal buffer size. Deal
2508 with BIT_INSERT_EXPR stores. Move up code updating the mask and
2509 also print the mask in the dump file.
2510 (pass_store_merging::gate): Minor tweak.
2511 (imm_store_chain_info::coalesce_immediate): Fix wrong association
2512 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
2513 stores with INTEGER_CST stores.
2514 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
2515 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
2516 and use it throughout. Generate bit insertion sequences if need be.
2517 (pass_store_merging::process_store): Remove redundant condition.
2518 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
2520 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
2522 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
2523 the 128-bit floating point types. Fix function comment.
2525 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2527 * config/aarch64/aarch64-simd.md
2528 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
2530 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
2533 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2535 PR tree-optimization/85989
2536 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
2538 (backprop::intersect_uses): Check it when deciding whether this
2539 is a backedge reference.
2540 (backprop::process_block): Add each phi to m_visited_phis
2541 after visiting it, then clear it at the end.
2543 2018-06-01 Richard Biener <rguenther@suse.de>
2545 * tree-vectorizer.h (vect_dr_stmt): New function.
2546 (vect_get_load_cost): Adjust.
2547 (vect_get_store_cost): Likewise.
2548 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
2549 Use vect_dr_stmt instead of DR_SMTT.
2550 (vect_record_base_alignments): Likewise.
2551 (vect_calculate_target_alignment): Likewise.
2552 (vect_compute_data_ref_alignment): Likewise and make static.
2553 (vect_update_misalignment_for_peel): Likewise.
2554 (vect_verify_datarefs_alignment): Likewise.
2555 (vector_alignment_reachable_p): Likewise.
2556 (vect_get_data_access_cost): Likewise. Pass down
2557 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
2558 (vect_get_peeling_costs_all_drs): Likewise.
2559 (vect_peeling_hash_get_lowest_cost): Likewise.
2560 (vect_enhance_data_refs_alignment): Likewise.
2561 (vect_find_same_alignment_drs): Likewise.
2562 (vect_analyze_data_refs_alignment): Likewise.
2563 (vect_analyze_group_access_1): Likewise.
2564 (vect_analyze_group_access): Likewise.
2565 (vect_analyze_data_ref_access): Likewise.
2566 (vect_analyze_data_ref_accesses): Likewise.
2567 (vect_vfa_segment_size): Likewise.
2568 (vect_small_gap_p): Likewise.
2569 (vectorizable_with_step_bound_p): Likewise.
2570 (vect_prune_runtime_alias_test_list): Likewise.
2571 (vect_analyze_data_refs): Likewise.
2572 (vect_supportable_dr_alignment): Likewise.
2573 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2574 (vect_gen_prolog_loop_niters): Likewise.
2575 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2576 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
2578 (vect_recog_mask_conversion_pattern): Likewise.
2579 (vect_try_gather_scatter_pattern): Likewise.
2580 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
2581 to vect_get_store_cost.
2582 (vect_get_store_cost): Get stmt_info instead of DR.
2583 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
2584 (vect_get_load_cost): Get stmt_info instead of DR.
2586 2018-06-01 Richard Biener <rguenther@suse.de>
2589 * gimple-fold.c (var_decl_component_p): Also allow offsetted
2590 vars wrapped in MEM_REFs.
2592 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2594 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2595 Fix subreg tests so that we only return a choice between
2596 GENERAL_REGS and FP_REGS if the original classes included both.
2598 2018-06-01 Richard Biener <rguenther@suse.de>
2601 * tree-ssa-structalias.c (get_function_part_constraint):
2602 Handle NULL fi->decl.
2603 (find_func_aliases_for_call): Properly handle indirect
2604 fi from direct call.
2605 (find_func_clobbers): Likewise.
2606 (ipa_pta_execute): Likewise.
2607 (create_variable_info_for): For functions that are ifunc_resolver
2608 resolve to a varinfo that contains the result of the resolver call.
2609 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
2612 2018-05-31 Michael Collison <michael.collison@arm.com>
2614 * config/aarch64/aarch64.md:
2615 (*fix_to_zero_extenddfdi2): New pattern.
2616 * gcc.target/aarch64/fix_extend1.c: New testcase.
2618 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
2622 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
2623 and BUILT_IN_STRNCMP_EQ.
2624 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
2625 BUILT_IN_STRNCMP_EQ.
2626 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
2627 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2628 (gimple_fold_builtin): Likewise.
2629 * tree-ssa-strlen.c (compute_string_length): New function.
2630 (determine_min_obsize): New function.
2631 (handle_builtin_string_cmp): New function to handle calls to
2632 string compare functions.
2633 (strlen_optimize_stmt): Add handling to builtin string compare
2635 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2636 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2637 * tree.c (build_common_builtin_nodes): Add new defines of
2638 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
2640 2018-05-31 Jakub Jelinek <jakub@redhat.com>
2643 * bb-reorder.c (pass_partition_blocks::gate): Return false for
2644 functions with naked attribute.
2646 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
2648 * config/i386/sse.md (avx_vec_concat<mode>):
2649 Substitute concat_tg_mode mode attribute with xtg_mode.
2650 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
2651 (concat_tg_mode): Remove mode attribute.
2653 2018-05-31 Martin Sebor <msebor@redhat.com>
2656 * calls.c (alloc_max_size): Correct a logic error/typo.
2657 Treat excessive arguments as infinite. Warn for invalid arguments.
2658 * doc/invoke.texi (-Walloc-size-larger-than): Update.
2660 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
2663 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
2664 and movx for Haswell.
2666 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
2667 Cesar Philippidis <cesar@codesourcery.com>
2670 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
2671 when emitting error on private/firstprivate reductions.
2672 * omp-low.c (lower_omp_target): Avoid reference-type processing
2673 on pointers for firstprivate clause.
2675 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2677 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
2680 * config/aarch64/aarch64-simd.md
2681 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
2682 (aarch64_ld1_x3_<mode>): Likewise
2683 (aarch64_st1x2<VALLDIF:mode>): Likewise
2684 (aarch64_st1_x2_<mode>): Likewise
2685 (aarch64_st1x3<VALLDIF:mode>): Likewise
2686 (aarch64_st1_x3_<mode>): Likewise
2687 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
2688 (vld1_s8_x3): Likewise.
2689 (vld1_u16_x3): Likewise.
2690 (vld1_s16_x3): Likewise.
2691 (vld1_u32_x3): Likewise.
2692 (vld1_s32_x3): Likewise.
2693 (vld1_u64_x3): Likewise.
2694 (vld1_s64_x3): Likewise.
2695 (vld1_f16_x3): Likewise.
2696 (vld1_f32_x3): Likewise.
2697 (vld1_f64_x3): Likewise.
2698 (vld1_p8_x3): Likewise.
2699 (vld1_p16_x3): Likewise.
2700 (vld1_p64_x3): Likewise.
2701 (vld1q_u8_x3): Likewise.
2702 (vld1q_s8_x3): Likewise.
2703 (vld1q_u16_x3): Likewise.
2704 (vld1q_s16_x3): Likewise.
2705 (vld1q_u32_x3): Likewise.
2706 (vld1q_s32_x3): Likewise.
2707 (vld1q_u64_x3): Likewise.
2708 (vld1q_s64_x3): Likewise.
2709 (vld1q_f16_x3): Likewise.
2710 (vld1q_f32_x3): Likewise.
2711 (vld1q_f64_x3): Likewise.
2712 (vld1q_p8_x3): Likewise.
2713 (vld1q_p16_x3): Likewise.
2714 (vld1q_p64_x3): Likewise.
2715 (vst1_s64_x2): Likewise.
2716 (vst1_u64_x2): Likewise.
2717 (vst1_f64_x2): Likewise.
2718 (vst1_s8_x2): Likewise.
2719 (vst1_p8_x2): Likewise.
2720 (vst1_s16_x2): Likewise.
2721 (vst1_p16_x2): Likewise.
2722 (vst1_s32_x2): Likewise.
2723 (vst1_u8_x2): Likewise.
2724 (vst1_u16_x2): Likewise.
2725 (vst1_u32_x2): Likewise.
2726 (vst1_f16_x2): Likewise.
2727 (vst1_f32_x2): Likewise.
2728 (vst1_p64_x2): Likewise.
2729 (vst1q_s8_x2): Likewise.
2730 (vst1q_p8_x2): Likewise.
2731 (vst1q_s16_x2): Likewise.
2732 (vst1q_p16_x2): Likewise.
2733 (vst1q_s32_x2): Likewise.
2734 (vst1q_s64_x2): Likewise.
2735 (vst1q_u8_x2): Likewise.
2736 (vst1q_u16_x2): Likewise.
2737 (vst1q_u32_x2): Likewise.
2738 (vst1q_u64_x2): Likewise.
2739 (vst1q_f16_x2): Likewise.
2740 (vst1q_f32_x2): Likewise.
2741 (vst1q_f64_x2): Likewise.
2742 (vst1q_p64_x2): Likewise.
2743 (vst1_s64_x3): Likewise.
2744 (vst1_u64_x3): Likewise.
2745 (vst1_f64_x3): Likewise.
2746 (vst1_s8_x3): Likewise.
2747 (vst1_p8_x3): Likewise.
2748 (vst1_s16_x3): Likewise.
2749 (vst1_p16_x3): Likewise.
2750 (vst1_s32_x3): Likewise.
2751 (vst1_u8_x3): Likewise.
2752 (vst1_u16_x3): Likewise.
2753 (vst1_u32_x3): Likewise.
2754 (vst1_f16_x3): Likewise.
2755 (vst1_f32_x3): Likewise.
2756 (vst1_p64_x3): Likewise.
2757 (vst1q_s8_x3): Likewise.
2758 (vst1q_p8_x3): Likewise.
2759 (vst1q_s16_x3): Likewise.
2760 (vst1q_p16_x3): Likewise.
2761 (vst1q_s32_x3): Likewise.
2762 (vst1q_s64_x3): Likewise.
2763 (vst1q_u8_x3): Likewise.
2764 (vst1q_u16_x3): Likewise.
2765 (vst1q_u32_x3): Likewise.
2766 (vst1q_u64_x3): Likewise.
2767 (vst1q_f16_x3): Likewise.
2768 (vst1q_f32_x3): Likewise.
2769 (vst1q_f64_x3): Likewise.
2770 (vst1q_p64_x3): Likewise.
2772 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2774 * config/msp430/msp430.c (msp430_output_labelref): Prepend
2775 user_label_prefix to name.
2777 * tree-core.h: Update comment about the format of NAME string
2778 passed to handler in attribute_spec.
2780 * config/msp430/msp430.md: Remove erroneous subreg expression from
2781 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
2782 zero_extend{q,h}isi2.
2784 2018-05-30 Borislav Petkov <bp@suse.de>
2786 * doc/extend.texi: Document some architecture specific
2787 constraints and sort entries.
2789 2018-05-30 Martin Sebor <msebor@redhat.com>
2792 * builtins.c (expand_builtin_stpcpy_1): New function.
2793 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
2794 only if the former succeeds.
2796 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2798 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
2801 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2803 * doc/invoke.texi (-flinker-output): Document
2805 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2807 * passes.c (ipa_write_summaries): Only modify statements if body
2809 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
2810 incrementally linking.
2811 (ipa_passes): Likewise.
2812 * lto-cgraph.c (lto_output_node): When incrementally linking do not
2813 pass down resolution info.
2814 * common.opt (flag_incremental_link): Update info.
2815 * gcc.c (plugin specs): Turn flinker-output=* to
2816 -plugin-opt=-linker-output-known
2817 * toplev.c (compile_file): Also cut compilation when doing incremental
2819 * flag-types. (enum lto_partition_model): Add
2820 LTO_LINKER_OUTPUT_NOLTOREL.
2821 (invoke.texi): Add -flinker-output docs.
2822 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
2823 link same way as WPA; do not stream in dead initializers.
2825 * dwarf2out.c (dwarf2out_die_ref_for_decl,
2826 darf2out_register_external_decl): Support incremental link.
2828 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2830 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
2832 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2834 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
2835 it down to simple_object_copy_lto_debug_sections.
2836 (run_gcc): Determine incremental LTO link time and configure
2837 lto1 into non-wpa mode, disable renaming of debug sections.
2839 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
2841 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2842 descriptions of various incorrectly documented functions.
2844 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
2847 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2848 address check not strict.
2850 2018-05-30 Richard Biener <rguenther@suse.de>
2852 PR tree-optimization/85964
2853 * tracer.c (better_p): Drop initialized count check, we only
2854 call the function with initialized counts now.
2855 (find_best_successor): Do find a best edge if one
2856 has uninitialized count.
2857 (find_best_predecessor): Likewise. Do BB frequency check only
2858 if count is initialized.
2860 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
2862 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
2863 (aarch64_ldrstr_offset_compare): New.
2864 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
2865 load/store orderings.
2866 (aarch64_gen_adjusted_ldpstp): Likewise.
2868 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
2870 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2871 Check for subset of GENERAL_REGS and FP_REGS.
2872 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
2875 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
2877 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
2878 and wi::to_poly_offset. Add the current offset and then check
2879 whether the sum fits, rather than using an unchecked addition of
2880 a checked term. Check for a shwi rather than a uhwi.
2881 * expr.c (get_bit_range): Use tree_to_poly_uint64.
2882 (store_constructor): Use poly_int_tree_p.
2883 (expand_expr_real_1): Likewise.
2884 * function.c (assign_temp): Likewise.
2885 * fold-const.c (const_binop): Use poly_int_tree_p and
2887 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
2889 * ipa-icf-gimple.c (func_checker::compare_operand): Use
2890 to_poly_offset for MEM offsets.
2891 * ipa-icf.c (sem_variable::equals): Likewise.
2892 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
2893 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
2894 wi::to_poly_offset for BIT_FIELD_REF offsets.
2895 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
2897 * var-tracking.c (emit_note_insn_var_location): Use
2898 tree_to_poly_uint64.
2900 2018-05-29 Jim Wilson <jimw@sifive.com>
2902 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
2904 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
2907 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
2908 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
2910 (sse4_1_round<mode>2): Use nonimmediate_operand
2911 for operand 1 predicate.
2913 2018-05-29 Martin Sebor <msebor@redhat.com>
2914 Richard Biener <rguenther@suse.de>
2917 * calls.c (get_size_range): Call determine_value_range instead
2918 of get_value_range..
2919 * tree-vrp.h (determine_value_range): Declared new function.
2920 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
2922 2018-05-29 Richard Biener <rguenther@suse.de>
2924 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
2925 sure to use non-pattern stmts for get_earlier_stmt arguments.
2926 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
2927 called on pattern stmts.
2928 (get_later_stmt): Likewise.
2930 2018-05-29 Martin Liska <mliska@suse.cz>
2932 PR gcov-profile/85759
2933 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
2936 2018-05-29 Jakub Jelinek <jakub@redhat.com>
2938 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
2940 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
2944 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2945 VEC_PACK_FLOAT_EXPR): New tree codes.
2946 * tree-pretty-print.c (op_code_prio): Handle
2947 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2948 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2949 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2950 * tree-inline.c (estimate_operator_cost): Likewise.
2951 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
2952 * fold-const.c (const_binop): Likewise.
2953 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
2954 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2955 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2956 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
2957 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2958 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2959 * expr.c (expand_expr_real_2): Likewise.
2960 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
2961 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
2962 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
2964 * optabs.c (expand_widen_pattern_expr): For
2965 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
2966 sign from result type rather than operand's type.
2967 (expand_binop_directly): For vec_packu_float_optab and
2968 vec_packs_float_optab allow result type to be different from operand's
2970 * optabs-tree.c (optab_for_tree_code): Handle
2971 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2972 VEC_PACK_FLOAT_EXPR. Formatting fixes.
2973 * tree-vect-generic.c (expand_vector_operations_1): Handle
2974 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2975 VEC_PACK_FLOAT_EXPR.
2976 * tree-vect-stmts.c (supportable_widening_operation): Handle
2978 (supportable_narrowing_operation): Handle FLOAT_EXPR.
2979 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
2980 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
2981 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
2982 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
2984 (vec_pack<floatprefix>_float_<mode>): New expander.
2985 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
2987 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
2988 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
2989 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
2990 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
2991 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
2993 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
2994 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
2995 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2996 VEC_PACK_FLOAT_EXPR): Document.
2998 2018-05-29 Richard Biener <rguenther@suse.de>
3000 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
3002 (stmt_vec_info_vec): Make pointer.
3003 (init_stmt_vec_info_vec): Remove.
3004 (free_stmt_vec_info_vec): Likewise.
3005 (set_stmt_vec_info_vec): New function.
3006 (free_stmt_vec_infos): Likewise.
3007 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
3008 (set_vinfo_for_stmt): Likewise.
3009 (get_earlier_stmt): Likewise.
3010 (get_later_stmt): Likewise.
3011 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
3012 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
3013 (vec_info::~vec_info): Free stmt_vec_infos.
3014 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
3015 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
3016 (pass_slp_vectorize::execute): Likewise.
3017 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
3018 (free_stmt_vec_info_vec): Likewise.
3019 (set_stmt_vec_info_vec): New function.
3020 (free_stmt_vec_infos): Likewise.
3021 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
3022 the global stmt_vec_info_vec.
3023 * tree-parloops.c (gather_scalar_reductions): Use
3024 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
3027 2018-05-29 Richard Biener <rguenther@suse.de>
3029 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
3031 2018-05-29 Martin Liska <mliska@suse.cz>
3032 David Malcolm <dmalcolm@redhat.com>
3034 * vec.c (test_reverse): New.
3035 (vec_c_tests): Add new test.
3036 * vec.h (vl_ptr>::reverse): New function.
3038 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
3040 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
3042 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
3045 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
3047 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
3049 2018-05-28 Richard Biener <rguenther@suse.de>
3051 PR tree-optimization/85933
3052 * tree-vect-data-refs.c (vect_record_base_alignments): Only
3053 look at stmts marked as vectorizable.
3055 2018-05-28 Richard Biener <rguenther@suse.de>
3057 PR tree-optimization/85934
3058 * tree-vect-generic.c (expand_vector_operations_1): Hoist
3059 vector boolean check before scalar optimization.
3061 2018-05-28 Jakub Jelinek <jakub@redhat.com>
3063 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
3066 2018-05-28 Mark Wielaard <mark@klomp.org>
3068 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
3069 if it is an expression containing a minus sign.
3071 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
3073 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
3075 2018-05-27 Paul Koning <ni1d@arrl.net>
3077 * config/pdp11/pdp11.md (truncsihi2): Remove.
3079 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
3080 Chung-Ju Wu <jasonwucj@gmail.com>
3082 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
3084 (unaligned_store_dw): Ditto.
3085 * config/nds32/nds32-memory-manipulation.c
3086 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
3087 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
3088 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
3089 (emit_setmem_word_loop): Rename to ...
3090 (emit_setmem_doubleword_loop): ... this.
3091 (nds32_gen_dup_4_byte_to_word_value): New function.
3092 (nds32_gen_dup_8_byte_to_double_word_value): New function.
3093 (nds32_expand_setmem_loop): Refine implementation.
3094 (nds32_expand_setmem_loop_v3m): Ditto.
3095 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
3098 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3100 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
3102 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3104 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
3105 (nds32_init_machine_status): Initialize machine->attr_naked_p and
3106 machine->attr_no_prologue_p.
3107 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
3108 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
3109 (nds32_expand_epilogue): Consider attr_naked_p.
3110 (nds32_expand_epilogue_v3pop): Likewise.
3111 (nds32_can_use_return_insn): Likewise.
3112 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
3113 attr_no_prologue_p fields.
3114 * config/nds32/nds32.opt (mret-in-naked-func): New option.
3116 2018-05-27 Jakub Jelinek <jakub@redhat.com>
3119 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
3121 * config/i386/sse.md
3122 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
3124 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
3126 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
3128 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
3130 (*<floatsuffix>floatv2div2sf2): Rename to ...
3131 (*float<floatunssuffix>v2div2sf2): ... this.
3132 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
3133 (float<floatunssuffix>v2div2sf2_mask): ... this.
3134 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
3135 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
3136 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
3138 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
3140 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3142 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3144 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3146 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3148 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
3149 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
3150 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
3151 gen_ufix_truncv8dfv8si2.
3152 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
3153 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
3154 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
3155 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
3156 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
3157 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
3158 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
3159 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
3161 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3165 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
3167 2018-05-25 Jim Wilson <jimw@sifive.com>
3169 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
3170 * config/riscv/riscv.c (struct machine_function): Add
3171 interrupt_handler_p and attribute_checked_p fields.
3172 (riscv_attribute_table): Add interrupt.
3173 (riscv_interrupt_type_p): New.
3174 (riscv_save_reg_p): Save extra regs for interrupt handler.
3175 (riscv_use_save_libcall): Return false for interrupt handler.
3176 (riscv_first_stack_step): Add forward declaration.
3177 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
3178 for interrupt handler with large frame. Use it for saved reg list.
3179 (riscv_expand_prologue): Move flag_stack_usage_info support to
3180 eliminate duplication.
3181 (riscv_expand_epilogue): Generate mret for interrupt handler.
3182 (riscv_epilogue_uses): New.
3183 (riscv_can_use_return_insn): Return false for interrupt handler.
3184 (riscv_function_ok_for_sibcall): Likewise.
3185 (riscv_set_current_function): Add interrupt handler support.
3186 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
3187 * config/riscv/riscv.md (UNSPECV_MRET): New.
3189 (riscv_frflags, riscv_fsflags): Use tab after opcode.
3191 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
3193 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
3195 PR tree-optimization/85712
3196 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
3197 this candidate has already been replaced in-situ by a copy.
3199 2018-05-25 Jason Merrill <jason@redhat.com>
3201 PR c++/80485 - inline function non-zero address.
3202 * symtab.c (nonzero_address): Check DECL_COMDAT.
3204 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
3207 * config/alpha/alpha.md (ashlsi3): New insn pattern.
3208 (*ashlsi_se): Rename from *ashldi_se. Define as sign
3209 extension of SImode operation. Use const123_operand predicate.
3210 (*saddsi_1): Remove.
3211 (*saddl_se_1): Ditto.
3213 (*ssubl_se_1): Ditto.
3214 * config/alpha/predicates.md (const123_operand): New predicate.
3215 * config/alpha/constraints.md (P): Use IN_RANGE.
3217 2018-05-25 Richard Biener <rguenther@suse.de>
3219 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
3221 (ref_maybe_used_by_stmt_p): Likewise.
3222 (stmt_may_clobber_ref_p): Likewise.
3223 (stmt_may_clobber_ref_p_1): Likewise.
3224 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
3226 (ref_maybe_used_by_stmt_p): Likewise.
3227 (stmt_may_clobber_ref_p): Likewise.
3228 (stmt_may_clobber_ref_p_1): Likewise.
3229 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
3230 the alias oracle to disambiguate DRs with stmts DR analysis
3232 (vect_analyze_data_refs): Do not give up on not analyzable
3233 DRs for BB vectorization. Remove code truncating the dataref
3236 2018-05-25 Jakub Jelinek <jakub@redhat.com>
3239 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3240 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
3241 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
3243 2018-05-25 Richard Biener <rguenther@suse.de>
3245 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
3246 function, combining stmt data ref gathering and fatal analysis
3248 (vect_analyze_data_refs): Remove now redudnant code and simplify.
3249 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
3250 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
3251 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
3252 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
3254 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3256 PR tree-optimization/85720
3257 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
3258 SCC if all partitions are builtins.
3259 (version_loop_by_alias_check): New parameter. Generate cancelable
3260 runtime alias check if all partitions are builtins.
3261 (distribute_loop): Update call to above function.
3263 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3265 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
3266 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
3267 (parm_default_def_partition_arg): Ditto.
3268 (set_parm_default_def_partition): Ditto.
3269 (get_parm_default_def_partitions): Ditto and make it static.
3270 (get_undefined_value_partitions): Ditto and make it static.
3271 (remove_ssa_form): Refactor call to init_var_map here.
3272 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
3273 computation for loop region.
3274 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
3275 (register_default_def): Delete.
3276 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
3277 (parm_default_def_partition_arg): Ditto.
3278 (set_parm_default_def_partition): Ditto.
3279 (get_parm_default_def_partitions): Ditto and make it static.
3280 (get_undefined_value_partitions): Ditto and make it static.
3281 (coalesce_with_default, coalesce_with_default): Update comment.
3282 (create_coalesce_list_for_region): New func factored out from
3283 create_outofssa_var_map.
3284 (populate_coalesce_list_for_outofssa): New func factored out from
3285 create_outofssa_var_map and coalesce_ssa_name.
3286 (create_outofssa_var_map): Delete.
3287 (coalesce_ssa_name): Refactor to support live range computation.
3288 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
3289 (get_parm_default_def_partitions): Delete.
3290 (get_undefined_value_partitions): Ditto.
3291 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
3292 computation for loop region.
3293 (new_tree_live_info, loe_visit_block): Ditto.
3294 (live_worklist, set_var_live_on_entry): Ditto.
3295 (calculate_live_on_exit, verify_live_on_entry): Ditto.
3296 * tree-ssa-live.h (struct _var_map): New fields.
3297 (init_var_map): Change decl.
3298 (region_contains_p): New.
3300 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3302 * tree-ssa-live.h (live_merge_and_clear): Delete.
3304 2018-05-25 Richard Biener <rguenther@suse.de>
3307 * tree-dump.c (dequeue_and_dump): Remove access to removed
3308 operand 2 of a SWITCH_EXPR.
3310 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3312 * doc/sourcebuild.texi (vect_double_cond_arith): Include
3313 multiplication and division.
3314 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
3315 (cond_udiv@var{m}, cond_umod@var{m}): Document.
3316 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
3317 (cond_udiv_optab, cond_umod_optab): New optabs.
3318 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
3319 (IFN_COND_RDIV): New internal functions.
3320 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
3321 TRUNC_MOD_EXPR and RDIV_EXPR.
3322 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
3323 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
3325 (SVE_INT_BINARY): Include mult.
3326 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
3327 (optab, sve_int_op): Handle mult.
3328 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
3330 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
3331 for SVE_INT_BINARY_SD.
3333 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3335 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
3336 (optab, sve_int_op): Handle div and udiv.
3337 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
3338 for SVE_INT_BINARY_SD.
3339 (*<optab><mode>3): New insn for the same.
3341 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3343 * tree-vect-patterns.c: Include predict.h.
3344 (vect_recog_divmod_pattern): Restrict check for division support
3345 to when optimizing for size.
3347 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3349 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
3350 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
3351 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
3352 (gimple_match_op::set_op): Likewise.
3353 (gimple_resimplify4): Declare.
3354 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
3355 (expr::gen_transform): Likewise.
3356 (decision_tree::gen): Generate a simplification routine for 4 operands.
3357 * gimple-match-head.c (gimple_simplify): Add an overload for
3358 4 operands. In the top-level function, handle up to 4 call
3359 arguments and call gimple_resimplify4.
3360 (gimple_resimplify4): New function.
3361 (build_call_internal): Pass a fourth operand.
3362 (maybe_push_to_seq): Likewise.
3363 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
3364 Fold VEC_COND_EXPRs of an operation and a default value into
3365 an IFN_COND_* function if possible.
3366 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
3368 (SVE_COND_FP_BINARY): Include them.
3369 (optab, sve_fp_op): Handle them.
3370 (SVE_INT_BINARY_REV): New code iterator.
3371 (SVE_COND_FP_BINARY_REV): New int iterator.
3372 (commutative): New int attribute.
3373 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
3375 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
3377 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
3378 (*cond_<optab><mode>): New patterns for reversed operands.
3380 2018-05-25 Richard Biener <rguenther@suse.de>
3382 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
3383 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
3384 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
3385 (STMT_VINFO_GROUPED_ACCESS): Adjust.
3386 * tree-vect-data-refs.c (everywhere): Adjust users.
3387 * tree-vect-loop.c (everywhere): Likewise.
3388 * tree-vect-slp.c (everywhere): Likewise.
3389 * tree-vect-stmts.c (everywhere): Likewise.
3390 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
3392 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3394 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
3396 (gcc_cv_as_section_exclude): ... this.
3397 Try Solaris as #exclude syntax.
3398 * configure: Regenerate.
3399 * config.in: Regenerate.
3400 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
3402 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
3403 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
3405 * varasm.c (default_elf_asm_named_section): Don't check if
3406 HAVE_GAS_SECTION_EXCLUDE is defined.
3408 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3410 * doc/md.texi: Update the documentation of the cond_* optabs
3411 to mention the new final operand. Fix GET_MODE_NUNITS call.
3412 Describe the scalar case too.
3413 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
3414 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
3416 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
3417 (get_conditional_internal_fn): Update comment.
3418 * tree-vect-loop.c (vectorizable_reduction): Pass the original
3419 accumulator value as a final argument to conditional functions.
3420 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
3421 a define_expand and add an "else" operand. Assert for now that
3422 the else operand is equal to operand 2. Use SVE_INT_BINARY and
3423 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
3424 (*cond_<optab><mode>): New patterns.
3425 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
3426 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
3427 (UNSPEC_COND_EOR): Delete.
3428 (optab): Remove associated mappings.
3429 (SVE_INT_BINARY): New code iterator.
3430 (sve_int_op): Remove int attribute and add "minus" to the code
3432 (SVE_COND_INT_OP): Delete.
3433 (SVE_COND_FP_OP): Rename to...
3434 (SVE_COND_FP_BINARY): ...this.
3436 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3438 * optabs.c (can_reuse_operands_p): New function.
3439 (maybe_legitimize_operands): Try to reuse the results for
3442 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3444 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
3445 Add {q} suffix to insn mnemonic.
3447 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3449 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
3450 (msp430_warn_func_return): New.
3452 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
3454 * fold-const.c (tree_nonzero_bits): New function.
3455 * fold-const.h (tree_nonzero_bits): Likewise.
3456 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
3457 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
3459 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3463 * varasm.c (assemble_alias): Check ifunc_resolver only on
3466 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3469 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
3470 when memory input operand is handled.
3472 2018-05-24 Luis Machado <luis.machado@linaro.org>
3474 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
3476 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
3478 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3480 * match.pd: Delay FMA folds until after vectorization.
3482 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
3485 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3486 address check not strict.
3488 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3490 * gimple-match.h (gimple_match_op): New class.
3491 (mprts_hook): Replace parameters with a gimple_match_op *.
3492 (maybe_build_generic_op): Likewise.
3493 (gimple_simplified_result_is_gimple_val): Replace parameters with
3494 a const gimple_match_op *.
3495 (gimple_simplify): Replace code_helper * and tree * parameters with
3496 a gimple_match_op * parameter.
3497 (gimple_resimplify1): Replace code_helper *, tree and tree *
3498 parameters with a gimple_match_op * parameter.
3499 (gimple_resimplify2): Likewise.
3500 (gimple_resimplify3): Likewise.
3501 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
3502 parameters with a gimple_match_op * parameter.
3503 * gimple-match-head.c (gimple_simplify): Change prototypes of
3504 auto-generated functions to take a gimple_match_op * instead of
3505 separate code_helper * and tree * parameters. Make the same
3506 change in the top-level overload and update calls to the
3507 gimple_resimplify routines. Update calls to the auto-generated
3508 functions and to maybe_push_res_to_seq in the publicly-facing
3509 operation-specific gimple_simplify overloads.
3510 (gimple_match_op::MAX_NUM_OPS): Define.
3511 (gimple_resimplify1): Replace rcode and ops with a single res_op
3512 parameter. Update call to gimple_simplify.
3513 (gimple_resimplify2): Likewise.
3514 (gimple_resimplify3): Likewise.
3515 (mprts_hook): Replace parameters with a gimple_match_op *.
3516 (maybe_build_generic_op): Likewise.
3517 (build_call_internal): Replace type, nargs and ops with
3518 a gimple_match_op *.
3519 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
3520 with a single gimple_match_op *. Update calls to mprts_hook,
3521 build_call_internal and gimple_simplified_result_is_gimple_val.
3522 Factor out code that is common to the tree_code and combined_fn cases.
3523 * genmatch.c (expr::gen_transform): Replace tem_code and
3524 tem_ops with a gimple_match_op called tem_op. Update calls
3525 to the gimple_resimplify functions and maybe_push_res_to_seq.
3526 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
3527 res_ops. Update call to the gimple_resimplify functions.
3528 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
3529 (decision_tree::gen): Make the functions take a gimple_match_op *
3530 called res_op instead of separate res_code and res_ops parameters.
3531 Update call accordingly.
3532 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
3533 and ops with a single res_op parameter. Update calls to
3534 maybe_build_generic_op and maybe_push_res_to_seq.
3535 (fold_stmt_1): Update calls to gimple_simplify and
3536 replace_stmt_with_simplification.
3537 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
3538 and gimple_simplified_result_is_gimple_val.
3539 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
3541 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
3542 with a gimple_match_op *.
3543 (vn_nary_build_or_lookup): Likewise. Update call to
3544 vn_nary_build_or_lookup_1.
3545 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
3546 gimple_match_op *. Update calls to the gimple_resimplify routines
3547 and to gimple_simplified_result_is_gimple_val.
3548 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
3549 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
3550 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
3551 (visit_nary_op): Likewise.
3552 (visit_reference_op_load): Likewise.
3554 2018-05-23 Luis Machado <luis.machado@linaro.org>
3556 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
3557 modifier for printing the step amount.
3559 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
3562 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
3565 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
3567 * doc/sourcebuild.texi (Endianness): New subsubsection.
3569 2018-05-23 Luis Machado <luis.machado@linaro.org>
3571 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3572 <prefetch_dynamic_strides>: New const bool field.
3573 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3574 prefetch_dynamic_strides.
3575 (exynosm1_prefetch_tune): Likewise.
3576 (thunderxt88_prefetch_tune): Likewise.
3577 (thunderx_prefetch_tune): Likewise.
3578 (thunderx2t99_prefetch_tune): Likewise.
3579 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
3581 (aarch64_override_options_internal): Update to set
3582 PARAM_PREFETCH_DYNAMIC_STRIDES.
3583 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3584 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3585 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3586 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3587 prefetch-dynamic-strides setting.
3589 2018-05-23 Luis Machado <luis.machado@linaro.org>
3591 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3592 <minimum_stride>: New const int field.
3593 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3594 minimum_stride field defaulting to -1.
3595 (exynosm1_prefetch_tune): Likewise.
3596 (thunderxt88_prefetch_tune): Likewise.
3597 (thunderx_prefetch_tune): Likewise.
3598 (thunderx2t99_prefetch_tune): Likewise.
3599 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
3600 <default_opt_level>: Set to 3.
3601 (aarch64_override_options_internal): Update to set
3602 PARAM_PREFETCH_MINIMUM_STRIDE.
3603 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3604 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3605 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3606 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3607 stride is constant and is below the minimum stride threshold.
3609 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3611 * config/arm/arm-cpus.in (mode26): Delete.
3612 (armv4): Delete mode26 reference.
3613 * config/arm/arm.c (arm_configure_build_target): Delete use of
3616 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3618 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
3620 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
3621 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
3623 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
3624 (floatunsdidf2): Ditto.
3626 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3628 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
3629 (fixuns_trunc<mode>si2_avx512f): Ditto.
3630 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
3631 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
3632 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
3634 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3636 PR rtl-optimization/79985
3637 * df-scan.c (df_insn_refs_collect): Remove special case for
3638 global registers and asm statements.
3640 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3642 * extend.texi (Global Register Variables): Rewrite the bullet list.
3643 Note that the register is available for allocation. Note that access
3644 via inline asm must use constraints. Add note about async-signal
3645 handlers. Remove paragraph about automagic register selection.
3647 2018-05-23 Richard Biener <rguenther@suse.de>
3649 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
3650 of fixed offset from memset VN.
3652 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
3654 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
3656 (alloc_cand_and_find_basis): Initialize first_interp field.
3657 (slsr_process_mul): Modify first_interp field.
3658 (slsr_process_add): Likewise.
3659 (slsr_process_cast): Modify first_interp field for each new
3661 (slsr_process_copy): Likewise.
3662 (dump_candidate): Dump first_interp field.
3663 (replace_mult_candidate): Process all interpretations, not just
3665 (replace_rhs_if_not_dup): Likewise.
3666 (replace_one_candidate): Likewise.
3668 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
3670 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
3672 (aarch64_needs_frame_chain): New function.
3673 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
3675 2018-05-23 Sudakshina Das <sudi.das@arm.com>
3678 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
3679 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
3680 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
3681 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
3682 as true for strict-align.
3683 (aarch64_can_inline_p): Perform checks even when callee has no
3684 attributes to check for strict alignment.
3685 * doc/extend.texi (AArch64 Function Attributes): Document
3687 * doc/invoke.texi: (AArch64 Options): Likewise.
3689 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
3691 PR tree-optimization/85853
3692 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
3693 the handling of the root of the node to...
3694 (vect_slp_analyze_node_operations_1): ...this new function,
3695 and run the whole thing with the child nodes' def types
3696 set according to their SLP node's def type.
3698 2018-05-23 Richard Biener <rguenther@suse.de>
3701 * tree-data-ref.c (create_runtime_alias_checks): Defer
3702 and ignore overflow warnings.
3704 2018-05-23 Yury Gribov <tetra2005@gmail.com>
3706 PR tree-optimization/85822
3707 * tree-vrp.c (is_masked_range_test): Fix handling of negative
3710 2018-05-23 Richard Biener <rguenther@suse.de>
3712 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3713 memset constants via native_interpret_expr.
3715 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
3718 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
3720 (cgraph_node::create_alias): Likewise.
3721 (cgraph_node::get_availability): Check ifunc_resolver instead
3722 of looking up ifunc attribute.
3723 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3724 * varasm.c (do_assemble_alias): Likewise.
3725 (assemble_alias): Likewise.
3726 (default_binds_local_p_3): Likewise.
3727 * cgraph.h (cgraph_node): Add ifunc_resolver.
3728 (cgraph_node::only_called_directly_or_aliased_p): Return false
3730 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
3732 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
3733 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
3734 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
3735 instead of looking up ifunc attribute.
3737 2018-05-22 Luis Machado <luis.machado@linaro.org>
3739 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
3741 2018-05-22 Martin Sebor <msebor@redhat.com>
3744 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
3745 only when expasion succeeds.
3746 (expand_builtin_strcmp): Same.
3747 (expand_builtin_strncmp): Same.
3749 2018-05-22 Martin Sebor <msebor@redhat.com>
3751 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
3753 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3754 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3756 * config/aarch64/aarch64-ldpstp.md: Replace uses of
3757 aarch64_mem_pair_operand with memory_operand and delete operand swapping
3759 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3760 Add check for legitimate_address.
3761 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
3762 (aarch64_swap_ldrstr_operands): New.
3763 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
3766 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3767 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3769 * config/aarch64/aarch64.md: New patterns to generate stp
3771 (store_pair_sw, store_pair_dw): New patterns to generate stp for
3772 single words and double words.
3773 (load_pair_sw, load_pair_dw): Likewise.
3774 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
3776 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
3778 * config/aarch64/aarch64-ldpstp.md: Modify peephole
3779 for different mode ldpstp and add peephole for merged zero stores.
3781 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3783 (aarch64_gen_store_pair): Rename calls to match new patterns.
3784 (aarch64_gen_load_pair): Rename calls to match new patterns.
3785 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
3786 (load_pair<DREG:mode><DREG2:mode>): ... This.
3787 (store_pair<mode>): Rename to...
3788 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
3789 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
3791 (V_INT_EQUIV): Handle SImode.
3792 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
3795 2018-05-22 Martin Sebor <msebor@redhat.com>
3798 * calls.c (maybe_warn_nonstring_arg): Use string length to set
3799 or ajust the presumed bound on an operation to avoid unnecessary
3802 2018-05-22 Martin Sebor <msebor@redhat.com>
3804 PR tree-optimization/85826
3805 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
3806 assuming that a DECL necesarily has a constant size.
3808 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
3811 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
3813 2018-05-22 Richard Biener <rguenther@suse.de>
3815 PR tree-optimization/85834
3816 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
3817 non-constant and non-zero memset arguments.
3819 2018-05-22 Martin Liska <mliska@suse.cz>
3822 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
3824 2018-05-22 Richard Biener <rguenther@suse.de>
3826 PR tree-optimization/85863
3827 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
3828 comparisons when vectype is specified.
3829 (vectorizable_condition): Do not specify vectype for
3830 vect_is_simple_cond when SLP vectorizing.
3832 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
3835 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
3836 define __ibm128 as long double.
3837 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
3838 as a distinct type when IEEE 128-bit support is enabled.
3839 (init_float128_ieee): Fix up conversions between IFmode and IEEE
3840 128-bit types to use the correct functions.
3841 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
3842 convert between 128-bit floating point types that have different
3843 modes but the same representation, instead of using gen_lowpart to
3845 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
3847 (IFKF_reg): New attributes to give the register constraints for
3849 (extend<mode>tf2_internal): New insns to mark an explicit
3850 conversion between 128-bit floating point types that have a
3851 different mode but share the same representation.
3853 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
3855 PR tree-optimization/85814
3856 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
3857 a null return from get_strinfo when unsharing the next
3858 strinfo in the chain.
3860 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3863 * varasm.c (weak_finish): Clean up weak_decls.
3865 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3867 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
3868 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
3869 UNSPEC_UADALP values.
3870 * config/aarch64/iterators.md (ABAL): New int iterator.
3873 (sur): Add mappings for the above.
3874 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
3876 (aarch64_<sur>abal<mode>_4): Likewise.
3877 (aarch64_<sur>adalp<mode>_3): Likewise.
3878 (<sur>sadv16qi): New define_expand.
3880 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
3882 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
3883 (*movdf_internal): Ditto.
3884 (*rcpsf2_sse): Ditto.
3885 (*rsqrtsf2_sse): Ditto.
3886 (*sqrt<mode>2_sse): Ditto.
3888 2018-05-21 Tamar Christina <tamar.christina@arm.com>
3890 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
3892 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
3893 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
3894 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
3895 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
3897 * config/aarch64/arm_neon.h: Likewise.
3898 * config/aarch64/iterators.md (VQ_I): New.
3900 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
3902 * config.gcc: Add arc/t-multilib-linux to tmake_file for
3904 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
3907 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3909 * config/nds32/constraints.md (S): New constraint.
3910 * config/nds32/nds32.md (call_internal): Use constraint S.
3911 (call_value_internal): Likewise.
3912 (sibcall_internal): Likewise.
3913 (sibcall_value_internal): Likewise.
3915 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3916 Chung-Ju Wu <jasonwucj@gmail.com>
3918 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
3921 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3922 Chung-Ju Wu <jasonwucj@gmail.com>
3924 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
3925 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
3926 (nds32_rtx_costs_impl): Simplify.
3927 (nds32_address_cost_impl): Simplify.
3928 (nds32_init_rtx_costs): New function.
3929 (nds32_rtx_costs_speed_prefer): Likewise.
3930 (nds32_rtx_costs_size_prefer): Likewise.
3931 (nds32_address_cost_speed_prefer): Likewise.
3932 (nds32_address_cost_speed_fwprop): Likewise.
3933 (nds32_address_cost_size_prefer): Likewise.
3934 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
3935 * config/nds32/nds32.c (nds32_option_override): Use
3936 nds32_init_rtx_costs function.
3938 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3940 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
3941 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
3942 (TARGET_PIPELINE_N8): Likewise.
3943 (TARGET_PIPELINE_N10): Likewise.
3944 (TARGET_PIPELINE_N13): Likewise.
3945 (TARGET_PIPELINE_GRAYWOLF): Likewise.
3947 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
3949 * config/nds32/nds32-fpu.md: Update copyright year.
3951 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3953 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
3955 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3957 * config/nds32/nds32.c
3958 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
3959 * config/nds32/nds32.opt (minline-asm-r15): New option.
3961 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3963 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
3965 * config/nds32/nds32.md (abssi2): New pattern.
3967 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
3969 * config/i386/i386.md (rex64namesuffix): New mode attribute.
3970 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
3971 Merge insn pattern from sse_cvtsi2ss<round_name> and
3972 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
3973 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
3974 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
3975 using SWI48 mode iterator.
3976 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
3977 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
3978 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
3979 pattern from sse_cvttss2si<round_saeonly_name>
3980 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
3981 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
3982 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
3983 using SWI48 mode iterator.
3984 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
3985 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
3986 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
3987 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
3988 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
3989 using SWI48 mode iterator.
3990 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
3991 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
3992 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
3993 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
3994 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
3995 SWI48 mode iterator.
3996 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
3997 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
3998 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
3999 pattern from sse_cvttsd2si<round_saeonly_name>
4000 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
4002 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4004 * config/nds32/nds32-md-auxiliary.c
4005 (nds32_valid_smw_lwm_base_p): Refine.
4006 (nds32_output_smw_single_word): Refine.
4007 (nds32_output_smw_double_word): New.
4008 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
4010 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4012 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
4013 (nds32_output_stack_pop): Refine.
4014 (nds32_expand_unaligned_load): Refine.
4015 (nds32_expand_unaligned_store): Refine.
4017 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4018 Chung-Ju Wu <jasonwucj@gmail.com>
4020 * config/nds32/constants.md: Add TP_REGNUM constant.
4021 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
4022 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
4024 * config/nds32/nds32-doubleword.md: Consider flag_pic.
4025 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
4026 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
4027 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
4028 and PIC code generation.
4029 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
4031 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
4033 * config/nds32/nds32.md: Support TLS and PIC.
4034 * config/nds32/nds32.c: Support TLS and PIC.
4035 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
4036 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
4039 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4041 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
4042 mode with E_ prefix.
4044 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4045 Chung-Ju Wu <jasonwucj@gmail.com>
4047 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
4048 * config/nds32/nds32-md-auxiliary.c
4049 (symbolic_reference_mentioned_p): New.
4050 (nds32_legitimize_ict_address): New.
4051 (nds32_expand_ict_move): New.
4052 (nds32_indirect_call_referenced_p): New.
4053 (nds32_symbol_binds_local_p): Delete.
4054 (nds32_long_call_p): Modify.
4055 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
4056 * config/nds32/nds32-protos.h
4057 (symbolic_reference_mentioned_p): Declare.
4058 (nds32_legitimize_ict_address): Declare.
4059 (nds32_expand_ict_move): Declare.
4060 (nds32_indirect_call_referenced_p): Declare.
4061 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
4062 (nds32_relax_group): Use nds32_ict_const_p as condition.
4063 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
4064 (nds32_asm_file_start): Output ict_model directive in asm code.
4065 (nds32_legitimate_address_p): Consider indirect call.
4066 (nds32_print_operand): Consider indirect call.
4067 (nds32_print_operand_address): Consider indirect call.
4068 (nds32_insert_attributes): Handle "indirect_call" attribute.
4069 (TARGET_LEGITIMATE_ADDRESS_P): Define.
4070 (TARGET_LEGITIMATE_CONSTANT_P): Define.
4071 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
4072 (TARGET_DELEGITIMIZE_ADDRESS): Define.
4073 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
4074 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
4075 (TARGET_ICT_MODEL_SMALL): Define.
4076 (TARGET_ICT_MODEL_LARGE): Define.
4077 * config/nds32/nds32.md (movsi): Consider ict model.
4078 (call, call_value): Consider ict model.
4079 (sibcall, sibcall_value): Consider ict model.
4080 * config/nds32/nds32.opt (mict-model): New option.
4081 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
4084 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
4085 Monk Chiang <sh.chiang04@gmail.com>
4086 Jim Wilson <jimw@sifive.com>
4088 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
4089 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
4090 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
4091 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
4092 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
4093 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
4094 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
4095 compute save_libcall_adjustment properly.
4096 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
4097 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
4098 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
4099 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
4100 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
4101 (ABI_SPEC): Handle mabi=ilp32e.
4102 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
4103 (RVE): Add RVE mask.
4104 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
4105 <-march>: Add rv32e as an example.
4107 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
4110 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
4111 (intra_create_variable_infos): Handle C++ constructors.
4113 2018-05-18 Martin Liska <mliska@suse.cz>
4115 * passes.def: Remove a redundant pass.
4117 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
4120 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
4122 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4124 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
4126 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
4128 (armv2, armv2a, armv3, armv3m): Delete architectures.
4129 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
4130 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
4131 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
4133 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
4134 (*mulsidi3adddi): Likewise.
4135 (mulsidi3): Likewise.
4136 (*mulsidi3_nov6): Likewise.
4137 (umulsidi3): Likewise.
4138 (umulsidi3_nov6): Likewise.
4139 (umaddsidi4): Likewise.
4140 (*umulsidi3adddi): Likewise.
4141 (smulsi3_highpart): Likewise.
4142 (*smulsi3_highpart_nov6): Likewise.
4143 (umulsi3_highpart): Likewise.
4144 (*umulsi3_highpart_nov6): Likewise.
4145 * config/arm/arm.h (arm_arch3m): Delete.
4146 * config/arm/arm.c (arm_arch3m): Delete.
4147 (arm_option_override_internal): Update armv3-related comment.
4148 (arm_configure_build_target): Delete use of isa_bit_mode32.
4149 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
4150 (arm_rtx_costs_internal): Delete check of arm_arch3m.
4151 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
4153 (mulusa3): Likewise.
4154 * config/arm/arm-protos.h (arm_arch3m): Delete.
4155 * config/arm/arm-tables.opt: Regenerate.
4156 * config/arm/arm-tune.md: Likewise.
4157 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
4158 deleted architectures.
4160 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4162 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
4163 (armv5t, armv5te): New features.
4164 (ARMv5, ARMv5e): Delete fgroups.
4165 (ARMv5t, ARMv5te): Adjust for above changes.
4167 (armv5, armv5e): Delete arches.
4168 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
4170 (*call_reg_arm): Likewise.
4171 (*call_value_reg_armv5): Likewise.
4172 (*call_value_reg_arm): Likewise.
4173 (*call_symbol): Likewise.
4174 (*call_value_symbol): Likewise.
4175 (*sibcall_insn): Likewise.
4176 (*sibcall_value_insn): Likewise.
4178 (prefetch): Likewise.
4179 (define_split and define_peephole2 dependent on arm_arch5):
4181 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
4183 (TARGET_ARM_QBIT): Likewise.
4184 (TARGET_DSP_MULTIPLY): Likewise.
4185 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
4186 (arm_arch5, arm_arch5e): Delete.
4187 (arm_arch5t, arm_arch5te): Declare.
4188 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
4189 (arm_arch5t): Declare.
4190 (arm_option_reconfigure_globals): Update for the above.
4191 (arm_options_perform_arch_sanity_checks): Update comment, replace
4192 use of arm_arch5 with arm_arch5t.
4193 (use_return_insn): Likewise.
4194 (arm_emit_call_insn): Likewise.
4195 (output_return_instruction): Likewise.
4196 (arm_final_prescan_insn): Likewise.
4197 (arm_coproc_builtin_available): Likewise.
4198 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
4199 arm_arch5e with arm_arch5t and arm_arch5te.
4200 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
4201 (arm_arch5t, arm_arch5te): Declare.
4202 * config/arm/arm-tables.opt: Regenerate.
4203 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
4204 * config/arm/t-multilib: Likewise.
4205 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
4206 instead of arm_arch5.
4207 (*call_reg_thumb1): Likewise.
4208 (*call_value_reg_thumb1_v5): Likewise.
4209 (*call_value_reg_thumb1): Likewise.
4210 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
4212 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
4214 2018-05-18 Martin Liska <mliska@suse.cz>
4216 PR gcov-profile/84846
4217 * doc/gcov.texi: Document -t option of gcov tool.
4219 2018-05-18 Martin Liska <mliska@suse.cz>
4221 PR gcov-profile/84846
4222 * gcov.c (print_usage): Add new -t option.
4223 (process_args): Handle the option.
4224 (generate_results): Use stdout as output when requested by
4227 2018-05-18 Martin Liska <mliska@suse.cz>
4229 PR gcov-profile/84846
4230 * coverage.c (coverage_init): Write PWD to .gcno file.
4231 * doc/gcov.texi: Document how working directory is printed.
4232 * gcov-dump.c (dump_gcov_file): Print PWD.
4233 * gcov.c (output_intermediate_file): Likewise.
4234 (read_graph_file): Read PWD string.
4235 (output_lines): Print PWD.
4237 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4240 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
4241 for retval and return false if all args to phi are zero.
4243 2018-05-18 Richard Biener <rguenther@suse.de>
4245 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
4247 (evrp_dom_walker::before_dom_children): Call it.
4249 2018-05-18 Richard Biener <rguenther@suse.de>
4251 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
4252 results when processing array refs with variable index.
4254 2018-05-18 Toon Moene <toon@moene.org>
4256 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
4257 directly after that of -floop-interchange. Indicate that both
4258 options are enabled by default when specifying -O3.
4260 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4262 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
4263 iterator. Delete separate integer-mode vec_set<mode> expander.
4264 (aarch64_simd_vec_setv2di): Delete.
4265 (vec_setv2di): Delete.
4266 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
4267 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
4268 the "w, r" alternative.
4270 2018-05-18 Martin Liska <mliska@suse.cz>
4272 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
4273 * tree-pass.h (make_pass_lower_switch_O0): New function.
4274 * tree-switch-conversion.c (node_has_low_bound): Remove.
4275 (node_has_high_bound): Likewise.
4276 (node_is_bounded): Likewise.
4277 (class pass_lower_switch): Make it a template type and create
4279 (pass_lower_switch::execute): Add template argument.
4280 (make_pass_lower_switch): New function.
4281 (make_pass_lower_switch_O0): New function.
4282 (do_jump_if_equal): Remove.
4283 (emit_case_nodes): Simplify to just handle all 3 cases and leave
4284 all the hard work to tree optimization passes.
4286 2018-05-18 Martin Liska <mliska@suse.cz>
4288 * dbgcnt.c (limit_low): Renamed from limit.
4289 (limit_high): New variable.
4290 (dbg_cnt_is_enabled): Check for upper limit.
4291 (dbg_cnt): Adjust dumping.
4292 (dbg_cnt_set_limit_by_index): Add new argument for high
4294 (dbg_cnt_set_limit_by_name): Likewise.
4295 (dbg_cnt_process_single_pair): Parse new format.
4296 (dbg_cnt_process_opt): Use strtok.
4297 (dbg_cnt_list_all_counters): Remove 'value' and add
4299 * doc/invoke.texi: Document changes.
4301 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
4303 * doc/sourcebuild.texi (scalar_all_fma): Document.
4304 * tree.def (FMA_EXPR): Delete.
4305 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
4306 * internal-fn.c (ternary_direct): New macro.
4307 (expand_ternary_optab_fn): Likewise.
4308 (direct_ternary_optab_supported_p): Likewise.
4309 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
4310 * builtins.c (fold_builtin_fma): Delete.
4311 (fold_builtin_3): Don't call it.
4312 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
4313 * expr.c (expand_expr_real_2): Likewise.
4314 * fold-const.c (operand_equal_p): Likewise.
4315 (fold_ternary_loc): Likewise.
4316 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
4317 * gimple.c (DEFTREECODE): Likewise.
4318 * gimplify.c (gimplify_expr): Likewise.
4319 * optabs-tree.c (optab_for_tree_code): Likewise.
4320 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
4321 * tree-eh.c (operation_could_trap_p): Likewise.
4322 (stmt_could_throw_1_p): Likewise.
4323 * tree-inline.c (estimate_operator_cost): Likewise.
4324 * tree-pretty-print.c (dump_generic_node): Likewise.
4325 (op_code_prio): Likewise.
4326 * tree-ssa-loop-im.c (stmt_cost): Likewise.
4327 * tree-ssa-operands.c (get_expr_operands): Likewise.
4328 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
4329 * fold-const-call.h (fold_fma): Delete.
4330 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
4331 CFN_FNMA and CFN_FNMS.
4333 * genmatch.c (combined_fn): New enum.
4334 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
4335 (commutative_op): New function.
4336 (commutate): Use it. Handle more than 2 operands.
4337 (dt_operand::gen_gimple_expr): Use commutative_op.
4338 (parser::parse_expr): Allow :c to be used with non-binary
4339 operators if the commutative operand is known.
4340 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
4341 CFN_FMS, CFN_FNMA and CFN_FNMS.
4342 (backprop::process_assign_use): Remove FMA_EXPR handling.
4343 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
4344 (gen_hsa_fma): New function.
4345 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
4346 IFN_FNMA and IFN_FNMS.
4347 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
4348 * gimple-fold.h (follow_all_ssa_edges): Declare.
4349 * gimple-fold.c (follow_all_ssa_edges): New function.
4350 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
4351 gimple_build interface and use follow_all_ssa_edges to fold the result.
4352 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
4353 instead of checking for optabs directly.
4354 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
4355 rather than FMA_EXPRs.
4356 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
4357 call to IFN_FMA instead of an FMA_EXPR.
4359 2018-05-17 Jim Wilson <jimw@sifive.com>
4361 * expr.c (do_tablejump): When converting index to Pmode, if we have a
4362 sign extended promoted subreg, and the range does not have the sign bit
4363 set, then do a sign extend.
4365 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
4366 test, check for sign extended subreg and/or constant operands, and
4367 do a sign extend in that case.
4369 2018-05-17 Steve Ellcey <sellcey@cavium.com>
4371 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
4372 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
4374 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
4375 Change logics_shift_reg to logics_shift_imm.
4376 (thunderx2t99_fp_loadpair_basic): Delete.
4377 (thunderx2t99_fp_storepair_basic): Delete.
4378 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
4379 (thunderx2t99_asimd_polynomial): Delete.
4380 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
4381 and neon_fp_mul_d_scalar_q.
4382 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
4383 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
4384 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
4385 (thunderx2t99_asimd_lut): Add missing tbl types.
4386 (thunderx2t99_asimd_ext): Delete.
4387 (thunderx2t99_asimd_load1_1_mult): Delete.
4388 (thunderx2t99_asimd_load1_2_mult): Delete.
4389 (thunderx2t99_asimd_load1_ldp): New.
4390 (thunderx2t99_asimd_load1): New.
4391 (thunderx2t99_asimd_load2): Add missing *load2* types.
4392 (thunderx2t99_asimd_load3): New.
4393 (thunderx2t99_asimd_load4): New.
4394 (thunderx2t99_asimd_store1_1_mult): Delete.
4395 (thunderx2t99_asimd_store1_2_mult): Delete.
4396 (thunderx2t99_asimd_store2_mult): Delete.
4397 (thunderx2t99_asimd_store2_onelane): Delete.
4398 (thunderx2t99_asimd_store_stp): New.
4399 (thunderx2t99_asimd_store1): New.
4400 (thunderx2t99_asimd_store2): New.
4401 (thunderx2t99_asimd_store3): New.
4402 (thunderx2t99_asimd_store4): New.
4404 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
4406 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
4407 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
4409 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
4410 Segher Boessenkool <segher@kernel.crashing.org>
4413 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
4416 2018-05-17 Richard Biener <rguenther@suse.de>
4418 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
4419 for pruning loop and prune defs feeding only already visited PHIs.
4421 2018-05-17 Richard Biener <rguenther@suse.de>
4423 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
4425 2018-05-17 Bin Cheng <bin.cheng@arm.com>
4426 Richard Biener <rguenther@suse.de>
4428 PR tree-optimization/85793
4429 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
4430 for VMAT_ELEMENTWISE.
4432 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4434 * internal-fn.h (lookup_internal_fn): Declare
4435 * internal-fn.c (lookup_internal_fn): New function.
4436 * gimple.c (gimple_build_call_from_tree): Handle calls to
4438 * gimple-pretty-print.c (dump_gimple_call): Print "." before
4439 internal function names.
4440 * tree-pretty-print.c (dump_generic_node): Likewise.
4441 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
4443 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4445 * gimple-fold.h (gimple_build): Make the function forms take
4446 combined_fn rather than built_in_function.
4447 (gimple_simplify): Likewise.
4448 * gimple-match-head.c (gimple_simplify): Likewise.
4449 * gimple-fold.c (gimple_build): Likewise.
4450 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
4451 rather than gimple_build_call_internal.
4452 (get_initial_defs_for_reduction): Likewise.
4453 (vect_create_epilog_for_reduction): Likewise.
4454 (vectorizable_live_operation): Likewise.
4456 2018-05-17 Martin Liska <mliska@suse.cz>
4458 * gimple-ssa-sprintf.c (format_directive): Do not use
4459 space in between 'G_' and '('.
4461 2018-05-17 Jakub Jelinek <jakub@redhat.com>
4464 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
4465 even if the mask is not all ones.
4468 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
4470 (ix86_gimple_fold_builtin): Likewise.
4473 * config/i386/i386.c: Include tree-vector-builder.h.
4474 (ix86_vector_shift_count): New function.
4475 (ix86_fold_builtin): Fold shift builtins by scalar count.
4476 (ix86_gimple_fold_builtin): Likewise.
4478 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
4479 _mm512_setzero): New intrinsics.
4481 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
4482 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4484 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
4485 code generation for cases where splatting a value is not useful.
4486 * simplify-rtx.c (simplify_ternary_operation): Simplify
4487 vec_merge across a vec_duplicate and a paradoxical subreg forming
4488 a vector mode to a vec_concat.
4490 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
4492 * config.gcc: Support "goldmont-plus".
4493 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4495 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4496 PROCESSOR_GOLDMONT_PLUS.
4497 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
4498 (processor_target_table): Add "goldmont-plus".
4499 (PTA_GOLDMONT_PLUS): Define.
4500 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
4501 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
4502 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
4503 (fold_builtin_cpu): Add "goldmont-plus".
4504 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
4505 (ix86_option_override_internal): Add "goldmont-plus".
4506 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
4507 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
4508 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
4509 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
4511 2018-05-17 Richard Biener <rguenther@suse.de>
4513 PR tree-optimization/85757
4514 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
4515 remove defs that only feed that PHI from further processing.
4517 2018-05-16 Jim Wilson <jimw@sifive.com>
4519 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
4521 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
4523 2018-05-16 Mark Wielaard <mark@klomp.org>
4525 * dwarf2out.c (count_index_strings): New function.
4526 (output_indirect_strings): Call count_index_strings and generate
4527 header for dwarf_version >= 5.
4529 2018-05-16 Mark Wielaard <mark@klomp.org>
4531 * dwarf2out.c (dwarf_FORM): New function.
4532 (set_indirect_string): Use dwarf_FORM.
4533 (reset_indirect_string): Likewise.
4534 (size_of_die): Likewise.
4535 (value_format): Likewise.
4536 (output_die): Likewise.
4537 (add_skeleton_AT_string): Likewise.
4538 (output_macinfo_op): Likewise.
4539 (index_string): Likewise.
4540 (output_index_string_offset): Likewise.
4541 (output_index_string): Likewise.
4542 (count_index_strings): Likewise.
4544 2018-05-16 Carl Love <cel@us.ibm.com>
4546 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4547 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
4549 2018-05-16 Martin Jambor <mjambor@suse.cz>
4551 * ipa-prop.c (ipa_free_all_edge_args): Remove.
4552 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
4554 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
4556 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
4557 (fnma<mode>4): Likewise.
4558 (fms<mode>4): Likewise.
4559 (fnms<mode>4): Likewise.
4560 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
4561 (aarch64_fnma<mode>4): Likewise.
4562 (aarch64_fms<mode>4): Likewise.
4563 (aarch64_fnms<mode>4): Likewise.
4564 (aarch64_fnmadd<mode>4): Likewise.
4566 2018-05-16 Jason Merrill <jason@redhat.com>
4568 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
4570 2018-05-16 Richard Biener <rguenther@suse.de>
4572 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
4573 (dump_stmt_cost): Declare.
4574 (add_stmt_cost): Dump cost we add.
4575 (add_stmt_costs): New function.
4576 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
4578 (vect_analyze_stmt): Adjust prototype.
4579 (vectorizable_condition): Likewise.
4580 (vectorizable_live_operation): Likewise.
4581 (vectorizable_reduction): Likewise.
4582 (vectorizable_induction): Likewise.
4583 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
4584 cost vector to pass to vectorizable_ and record afterwards.
4585 (vect_model_reduction_cost): Take cost vector argument and adjust.
4586 (vect_model_induction_cost): Likewise.
4587 (vectorizable_reduction): Likewise.
4588 (vectorizable_induction): Likewise.
4589 (vectorizable_live_operation): Likewise.
4590 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
4591 SLP_TREE_NUMBER_OF_VEC_STMTS.
4592 (vect_analyze_slp_cost_1): Remove.
4593 (vect_analyze_slp_cost): Likewise.
4594 (vect_slp_analyze_node_operations): Take visited args and
4595 a target cost vector. Avoid processing already visited stmt sets.
4596 (vect_slp_analyze_operations): Use a local cost vector to gather
4597 costs and register those of non-discarded instances.
4598 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
4599 (vect_schedule_slp_instance): Remove copying of
4600 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
4602 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
4603 adding cost. Record cost entry location.
4604 (vect_prologue_cost_for_slp_op): Function to compute cost of
4605 a constant or invariant generated for SLP vect in the prologue,
4606 split out from vect_analyze_slp_cost_1.
4607 (vect_model_simple_cost): Make static. Adjust for SLP costing.
4608 (vect_model_promotion_demotion_cost): Likewise.
4609 (vect_model_store_cost): Likewise, make static.
4610 (vect_model_load_cost): Likewise.
4611 (vectorizable_bswap): Add cost vector arg and adjust.
4612 (vectorizable_call): Likewise.
4613 (vectorizable_simd_clone_call): Likewise.
4614 (vectorizable_conversion): Likewise.
4615 (vectorizable_assignment): Likewise.
4616 (vectorizable_shift): Likewise.
4617 (vectorizable_operation): Likewise.
4618 (vectorizable_store): Likewise.
4619 (vectorizable_load): Likewise.
4620 (vectorizable_condition): Likewise.
4621 (vectorizable_comparison): Likewise.
4622 (can_vectorize_live_stmts): Likewise.
4623 (vect_analyze_stmt): Likewise.
4624 (vect_transform_stmt): Adjust calls to vectorizable_*.
4625 * tree-vectorizer.c: Include gimple-pretty-print.h.
4626 (dump_stmt_cost): New function.
4628 2018-05-16 Richard Biener <rguenther@suse.de>
4630 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
4631 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
4632 * tree-ssa-dse.c: Include tree-ssa-loop.h.
4633 (check_name): New callback.
4634 (dse_classify_store): Track cycles via a visited bitmap of PHI
4635 defs and simplify handling of in-loop and across loop dead stores
4636 and properly fail for loop-variant refs. Handle byte-tracking with
4637 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
4640 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
4642 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
4643 (vect_get_mask_type_for_stmt): Likewise.
4644 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
4646 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
4647 to determine the statement's vector type and the vector type that
4648 should be used for calculating nunits. Deal with cases in which
4649 the type has to be deferred.
4650 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
4651 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
4652 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
4653 (vect_determine_vf_for_stmt): New functions, split out from...
4654 (vect_determine_vectorization_factor): ...here.
4655 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
4656 (vect_get_mask_type_for_stmt): New functions, split out from
4657 vect_determine_vectorization_factor.
4659 2018-05-16 Richard Biener <rguenther@suse.de>
4661 * tree-cfg.c (verify_gimple_assign_ternary): Properly
4662 verify the [VEC_]COND_EXPR embedded comparison.
4664 2018-05-15 Martin Sebor <msebor@redhat.com>
4666 PR tree-optimization/85753
4667 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
4668 RECORD_TYPE in addition to ARRAY_TYPE.
4670 2018-05-15 Martin Sebor <msebor@redhat.com>
4673 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
4675 2018-05-15 Richard Biener <rguenther@suse.de>
4677 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
4678 add by_clobber_p one. Change algorithm to collect all defs
4679 representing uses we need to walk and try reducing them to
4680 a single one before failing.
4681 (dse_dom_walker::dse_optimize_stmt): Adjust.
4683 2018-05-13 Mark Wielaard <mark@klomp.org>
4685 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
4686 (size_of_loc_descr): Likewise.
4687 (output_loc_operands): Likewise.
4688 (output_loc_operands_raw): Likewise.
4689 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
4690 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
4691 (hash_loc_operands): Likewise.
4692 (compare_loc_operands): Likewise.
4694 2018-05-14 Mark Wielaard <mark@klomp.org>
4696 * dwarf2out.c (count_index_addrs): New function.
4697 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
4699 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4701 PR tree-optimization/83648
4702 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
4703 return value as malloc candidate.
4705 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4708 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
4709 param as true in call to suggest_attribute.
4711 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
4713 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
4714 -mreadonly-in-sdata.
4716 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4718 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
4720 (aarch64_crypto_aesd_fused): Likewise.
4722 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
4724 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
4725 (movsi_aarch64): Likewise.
4726 (load_pairsi): Likewise.
4727 (load_pairdi): Likewise.
4728 (store_pairsi): Likewise.
4729 (store_pairdi): Likewise.
4730 (load_pairsf): Likewise.
4731 (load_pairdf): Likewise.
4732 (store_pairsf): Likewise.
4733 (store_pairdf): Likewise.
4734 (zero_extend): Likewise.
4735 (trunc): Swap alternatives.
4736 (fcvt_target): Add '?' to prefer w over r.
4738 2018-05-14 Jakub Jelinek <jakub@redhat.com>
4741 * config/i386/i386.md: Disallow non-commutative arithmetics in
4742 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
4743 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
4744 in the peephole2 before it.
4746 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
4748 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
4749 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
4750 (ix86_handle_option): Handle -mcldemote.
4751 * config.gcc: New header.
4752 * config/i386/cldemoteintrin.h: New file.
4753 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
4754 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4756 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4757 OPTION_MASK_ISA_CLDEMOTE.
4758 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
4759 (ix86_valid_target_attribute_inner_p): Ditto.
4760 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
4761 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
4762 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
4763 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
4764 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
4766 * config/i386/i386.opt: Add -mcldemote.
4767 * config/i386/x86intrin.h: New header.
4768 * doc/invoke.texi: Add -mcldemote.
4770 2018-05-14 Richard Biener <rguenther@suse.de>
4772 * doc/match-and-simplify.texi: Adjust :s documentation.
4774 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
4776 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
4777 intended memcpy size.
4778 (REORDER_45): Likewise.
4780 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
4782 * sort.cc: New file.
4783 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
4784 * vec.c (qsort_chk): Use gcc_qsort.
4785 * Makefile.in (OBJS-libcommon): Add sort.o.
4786 (build/sort.o): New target. Use it...
4787 (BUILD_RTL): ... here, and...
4788 (build/gencfn-macros): ... here, and...
4789 (build/genmatch): ... here.
4791 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4792 Chung-Ju Wu <jasonwucj@gmail.com>
4794 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
4795 * config/nds32/nds32-graywolf.md: New file.
4796 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
4797 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
4799 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
4800 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
4801 * config/nds32/nds32.md (pipeline_model): Add graywolf.
4802 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
4803 * config/nds32/pipelines.md: Include n15 settings.
4805 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4806 Chung-Ju Wu <jasonwucj@gmail.com>
4808 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
4809 * config/nds32/nds32-n13.md: New file.
4810 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
4811 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
4813 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
4814 * config/nds32/nds32.md (pipeline_model): Add n13.
4815 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
4816 * config/nds32/pipelines.md: Include n13 settings.
4818 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4819 Chung-Ju Wu <jasonwucj@gmail.com>
4821 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
4822 * config/nds32/nds32-n10.md: New file.
4823 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
4824 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
4826 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
4827 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
4828 * config/nds32/nds32.md (pipeline_model): Add n10.
4829 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
4830 * config/nds32/pipelines.md: Include n10 settings.
4832 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
4833 Kito Cheng <kito.cheng@gmail.com>
4834 Chung-Ju Wu <jasonwucj@gmail.com>
4836 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
4837 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4838 Add enum values for DSP extension instructions.
4839 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
4841 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
4842 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
4844 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
4845 * config/nds32/nds32-dspext.md: New file for DSP implementation.
4846 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
4847 * config/nds32/nds32-intrinsic.md: Likewise.
4848 * config/nds32/nds32_intrinsic.h: Likewise.
4849 * config/nds32/nds32-md-auxiliary.c: Likewise.
4850 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
4851 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
4852 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
4853 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
4854 * config/nds32/nds32-protos.h: New declarations for DSP extension.
4855 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
4856 TYPE_DMAC in switch statement.
4857 * config/nds32/nds32.c: New checking and implementation for DSP
4858 extension instructions.
4859 * config/nds32/nds32.h: Likewise.
4860 * config/nds32/nds32.md: Likewise.
4861 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
4862 * config/nds32/predicates.md: Implement new predicates for DSP
4865 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4867 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
4868 Reformat alternatives and attributes so it is easier to identify
4869 which constraints/attributes go with which instruction.
4870 (mov<mode>_hardfloat32, FMOVE64): Likewise.
4871 (mov<mode>_softfloat32, FMOVE64): Likewise.
4872 (mov<mode>_hardfloat64, FMOVE64): Likewise.
4873 (mov<mode>_softfloat64, FMOVE64): Likewise.
4875 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
4877 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4879 (Basic PowerPC Built-in Functions): The new name of the
4880 subsection previously known as "PowerPC Built-in Functions".
4881 (Basic PowerPC Built-in Functions Available on all Configurations):
4883 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
4884 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
4885 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
4886 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
4888 2018-05-11 Martin Jambor <mjambor@suse.cz>
4891 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
4894 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4897 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
4899 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
4901 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
4902 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
4903 (ix86_handle_option): Handle -mwaitpkg.
4904 * config.gcc: New header.
4905 * config/i386/cpuid.h (bit_WAITPKG): New bit.
4906 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
4907 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
4909 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4910 OPTION_MASK_ISA_WAITPKG.
4911 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
4912 (ix86_option_override_internal): Add PTA_WAITPKG.
4913 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
4914 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
4915 IX86_BUILTIN_TPAUSE.
4916 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
4917 __builtin_ia32_umwait and __builtin_ia32_tpause.
4918 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
4919 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
4920 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
4921 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
4922 UNSPECV_TPAUSE): New.
4923 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
4924 * config/i386/i386.opt: Add -mwaitpkg.
4925 * config/i386/waitpkgintrin.h: New file.
4926 * config/i386/x86intrin.h: New header.
4927 * doc/invoke.texi: Add -mwaitpkg.
4929 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4932 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
4934 (cortex-m0): Use armv6s-m isa.
4935 (cortex-m0plus): Likewise.
4936 (cortex-m1): Likewise.
4937 (cortex-m0.small-multiply): Likewise.
4938 (cortex-m0plus.small-multiply): Likewise.
4939 (cortex-m1.small-multiply): Likewise.
4941 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
4942 Jakub Jelinek <jakub@redhat.com>
4944 PR tree-optimization/85692
4945 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
4946 source permute as well.
4948 2018-05-11 Martin Liska <mliska@suse.cz>
4951 * doc/extend.texi: Document LLVM style format for no_sanitize
4954 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
4956 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
4957 mode_supports_vsx_dform_quad to mode_supports_dq_form.
4958 (mode_supports_vsx_dform_quad): Likewise.
4959 (mode_supports_vmx_dform): Move these functions to be next to the
4960 other mode_supports functions.
4961 (mode_supports_dq_form): Likewise.
4962 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
4963 mode_supports_dq_form.
4964 (reg_offset_addressing_ok_p): Likewise.
4965 (offsettable_ok_by_alignment): Likewise.
4966 (rs6000_legitimate_offset_address_p): Likewise.
4967 (legitimate_lo_sum_address_p): Likewise.
4968 (rs6000_legitimize_address): Likewise.
4969 (rs6000_legitimize_reload_address): Likewise.
4970 (rs6000_secondary_reload_inner): Likewise.
4971 (rs6000_preferred_reload_class): Likewise.
4972 (rs6000_output_move_128bit): Likewise.
4974 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4976 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
4977 Generate SImode target register for null target.
4978 <case IX86_BUILTIN_XGETBV>: Ditto.
4979 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
4980 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
4982 2018-05-10 Carl Love <cel@us.ibm.com>
4984 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4985 dcbtt and dcbtstt if operands[2] is 0.
4987 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4990 * config/i386/sse.md (usadv64qi): New expander.
4992 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
4994 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
4995 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
4996 -maltivec=be support.
4997 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
4998 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
4999 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
5000 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
5001 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
5002 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
5003 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
5004 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
5005 altivec_vsumsws): Adjust.
5006 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
5007 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
5008 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
5010 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
5011 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
5012 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
5013 (altivec_lve<VI_char>x): Delete expand.
5014 (*altivec_lve<VI_char>x_internal): Rename to...
5015 (altivec_lve<VI_char>x): ... this.
5016 (altivec_lvxl_<mode>): Delete expand.
5017 (*altivec_lvxl_<mode>_internal): Rename to ...
5018 (altivec_lvxl_<mode>): ... this.
5019 (altivec_stvxl_<mode>): Delete expand.
5020 (*altivec_stvxl_<mode>_internal): Rename to ...
5021 (altivec_stvxl_<mode>): ... this.
5022 (altivec_stve<VI_char>x): Delete expand.
5023 (*altivec_stve<VI_char>x_internal): Rename to ...
5024 (altivec_stve<VI_char>x): ... this.
5025 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
5026 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
5027 reduc_plus_scal_<mode>): Adjust.
5028 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
5030 (rs6000_cpu_cpp_builtins): Adjust.
5031 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
5032 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
5033 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
5034 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
5035 -maltivec=be support.
5036 (rs6000_split_vec_extract_var): Adjust.
5037 (rs6000_split_v4si_init): Adjust.
5038 (swap_selector_for_mode): Delete.
5039 (altivec_expand_lvx_be, altivec_expand_stvx_be,
5040 altivec_expand_stvex_be): Delete.
5041 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
5042 -maltivec=be support.
5043 (rs6000_gimple_fold_builtin): Ditto.
5044 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
5046 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
5047 (TARGET_DIRECT_MOVE_64BIT): Adjust.
5048 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
5049 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
5050 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
5051 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
5052 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
5053 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
5054 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
5055 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
5056 anonymous split): Adjust.
5057 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
5058 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
5060 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
5062 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
5063 when --with-gxx-include-dir is also specified.
5064 * configure: Regenerate.
5066 2018-05-09 Jim Wilson <jimw@sifive.com>
5069 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
5070 * config/riscv/t-withmultilib: New.
5071 * config/riscv/withmultilib.h: New.
5072 * doc/install.texi: Document RISC-V --with-multilib-list support.
5074 2018-05-09 Richard Biener <rguenther@suse.de>
5076 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
5078 (vect_bb_vectorization_profitable_p): Adjust. Compute
5079 actual scalar cost using the cost vector and the add_stmt_cost
5082 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5084 PR rtl-optimization/85645
5085 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
5086 in the REG_CFA_REGISTER note for LR, don't leave it empty.
5088 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5090 PR rtl-optimization/85645
5091 * shrink-wrap.c (spread_components): Return a boolean saying if
5092 anything was changed.
5093 (try_shrink_wrapping_separate): Iterate spread_components until
5094 nothing changes anymore.
5096 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5098 PR rtl-optimization/85645
5099 * regrename.c (build_def_use): Also kill the chains that include the
5100 destination of a REG_CFA_REGISTER note.
5102 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5104 PR rtl-optimization/85645
5105 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
5106 insn that has a REG_CFA_REGISTER note.
5108 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
5110 * cfgexpand.c (expand_clobber): New function.
5111 (expand_gimple_stmt_1): Use it.
5112 * tree-vect-stmts.c (vect_clobber_variable): New function,
5114 (vectorizable_simd_clone_call): ...here.
5115 (vectorizable_store): Emit a clobber either side of an
5116 IFN_STORE_LANES sequence.
5117 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
5119 2018-05-09 Tom de Vries <tom@codesourcery.com>
5122 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
5123 (define_insn "trap_if_false"): Add exit after trap.
5125 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
5127 PR rtl-optimization/85638
5128 * bb-reorder.c: Include common/common-target.h.
5129 (create_forwarder_block): New function extracted from...
5130 (fix_up_crossing_landing_pad): ...here. Rename into...
5131 (dw2_fix_up_crossing_landing_pad): ...this.
5132 (sjlj_fix_up_crossing_landing_pad): New function.
5133 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
5134 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
5135 from both partitions and exit the loop after one iteration.
5137 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5140 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5142 (Basic PowerPC Built-in Functions): The new name of the
5143 subsection previously known as "PowerPC Built-in Functions".
5144 (Basic PowerPC Built-in Functions Available on all Configurations):
5146 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5148 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5150 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5152 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5155 2018-05-08 Jim Wilson <jimw@sifive.com>
5157 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
5158 (LD_EMUL_SUFFIX): New.
5159 (LINK_SPEC): Use it.
5161 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5163 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5165 (Basic PowerPC Built-in Functions): The new name of the
5166 subsection previously known as "PowerPC Built-in Functions".
5167 (Basic PowerPC Built-in Functions Available on all Configurations):
5169 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5171 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5173 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5175 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5178 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5181 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
5182 after cmpelim optimization.
5184 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
5186 * config.gcc: Support "goldmont".
5187 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
5188 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5190 * config/i386/i386.c (m_GOLDMONT): Define.
5191 (processor_target_table): Add "goldmont".
5192 (PTA_GOLDMONT): Define.
5193 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
5194 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
5195 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
5196 (fold_builtin_cpu): Add "goldmont".
5197 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
5198 (ix86_option_override_internal): Add "goldmont".
5199 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
5200 (processor_type): Add PROCESSOR_GOLDMONT.
5201 * config/i386/i386.md: Add CPU "glm".
5202 * config/i386/glm.md: New file.
5203 * config/i386/x86-tune.def: Add m_GOLDMONT.
5204 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
5206 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5209 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
5211 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
5212 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
5213 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
5216 * config/i386/i386.c (ix86_fold_builtin): Handle
5217 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
5220 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
5221 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
5223 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
5226 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
5227 (check_arch): Likewise.
5228 (check_fpu): Return the result rather than printing it.
5229 (end arch): Fix operator precedence.
5230 (end cpu): Likewise.
5231 (END): Print the result from check_fpu.
5233 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5234 Alan Hayward <alan.hayward@arm.com>
5235 David Sherwood <david.sherwood@arm.com>
5237 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
5238 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
5239 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
5240 (*fcmuo<mode>_and): New patterns.
5242 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5244 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
5245 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
5246 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
5247 (cmp_op, sve_imm_con): New code attributes.
5248 (SVE_COND_INT_CMP, imm_con): Delete.
5249 (cmp_op): Remove above unspecs from int attribute.
5250 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
5252 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
5253 comparison-specific unspecs.
5254 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
5255 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
5256 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
5257 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
5258 (*vec_fcm<cmp_op><mode>): Rename to...
5259 (*fcm<cmp_op><mode>): ...this and adjust likewise.
5260 (*vec_fcmuo<mode>): Rename to...
5261 (*fcmuo<mode>): ...this and adjust likewise.
5262 (*pred_fcm<cmp_op><mode>): New pattern.
5263 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
5264 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
5266 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
5268 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
5269 (aarch64_emit_sve_predicated_cond): New function.
5270 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
5271 (aarch64_emit_unspec_cond_or): Replace with...
5272 (aarch64_emit_sve_or_conds): ...this new function. Use
5273 aarch64_emit_sve_ptrue_op for the individual comparisons and
5274 aarch64_emit_binop to OR them together.
5275 (aarch64_emit_inverted_unspec_cond): Replace with...
5276 (aarch64_emit_sve_inverted_cond): ...this new function. Use
5277 aarch64_emit_sve_ptrue_op for the comparison and
5278 aarch64_emit_unop to invert the result.
5279 (aarch64_expand_sve_vec_cmp_float): Update after the above
5280 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
5282 2018-05-07 Nathan Sidwell <nathan@acm.org>
5284 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
5285 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
5286 (Backwards Compatibility): Likewise.
5288 2018-05-07 Luis Machado <luis.machado@linaro.org>
5292 2018-05-07 Luis Machado <luis.machado@linaro.org>
5294 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5295 <prefetch_dynamic_strides>: New const bool field.
5296 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5297 prefetch_dynamic_strides.
5298 (exynosm1_prefetch_tune): Likewise.
5299 (thunderxt88_prefetch_tune): Likewise.
5300 (thunderx_prefetch_tune): Likewise.
5301 (thunderx2t99_prefetch_tune): Likewise.
5302 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5304 (aarch64_override_options_internal): Update to set
5305 PARAM_PREFETCH_DYNAMIC_STRIDES.
5306 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5307 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5308 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5309 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5310 prefetch-dynamic-strides setting.
5312 2018-05-07 Luis Machado <luis.machado@linaro.org>
5314 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5315 <minimum_stride>: New const int field.
5316 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5317 minimum_stride field.
5318 (exynosm1_prefetch_tune): Likewise.
5319 (thunderxt88_prefetch_tune): Likewise.
5320 (thunderx_prefetch_tune): Likewise.
5321 (thunderx2t99_prefetch_tune): Likewise.
5322 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5323 (aarch64_override_options_internal): Update to set
5324 PARAM_PREFETCH_MINIMUM_STRIDE.
5325 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5326 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5327 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5328 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5329 stride is constant and is below the minimum stride threshold.
5331 2018-05-07 Luis Machado <luis.machado@linaro.org>
5333 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
5336 2018-05-07 Luis Machado <luis.machado@linaro.org>
5338 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5339 <prefetch_dynamic_strides>: New const bool field.
5340 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5341 prefetch_dynamic_strides.
5342 (exynosm1_prefetch_tune): Likewise.
5343 (thunderxt88_prefetch_tune): Likewise.
5344 (thunderx_prefetch_tune): Likewise.
5345 (thunderx2t99_prefetch_tune): Likewise.
5346 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5348 (aarch64_override_options_internal): Update to set
5349 PARAM_PREFETCH_DYNAMIC_STRIDES.
5350 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5351 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5352 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5353 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5354 prefetch-dynamic-strides setting.
5356 2018-05-07 Luis Machado <luis.machado@linaro.org>
5358 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5359 <minimum_stride>: New const int field.
5360 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5361 minimum_stride field.
5362 (exynosm1_prefetch_tune): Likewise.
5363 (thunderxt88_prefetch_tune): Likewise.
5364 (thunderx_prefetch_tune): Likewise.
5365 (thunderx2t99_prefetch_tune): Likewise.
5366 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5367 (aarch64_override_options_internal): Update to set
5368 PARAM_PREFETCH_MINIMUM_STRIDE.
5369 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5370 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5371 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5372 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5373 stride is constant and is below the minimum stride threshold.
5375 2018-05-06 Jakub Jelinek <jakub@redhat.com>
5378 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
5379 the type is addressable. Don't force op into register if it has
5382 2018-05-05 Roland McGrath <mcgrathr@google.com>
5385 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
5386 any section for which we don't know a specific type it should have,
5387 regardless of name. Previously this was done only for the exact
5388 names ".init_array", ".fini_array", and ".preinit_array".
5389 (default_elf_asm_named_section): Add comment about
5390 relationship with default_section_type_flags and SECTION_NOTYPE.
5391 (get_section): Don't consider it a type conflict if one side has
5392 SECTION_NOTYPE and the other doesn't, as long as neither has the
5393 SECTION_BSS et al used in the default_section_type_flags logic.
5395 2018-05-05 Tom de Vries <tom@codesourcery.com>
5398 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
5399 (workaround_barsyncs): New function.
5400 (nvptx_reorg): Use workaround_barsyncs.
5401 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5402 (define_expand "nvptx_membar_cta"): New define_expand.
5403 (define_insn "*nvptx_membar_cta"): New insn.
5405 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
5407 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
5408 To improve optimization opportunities.
5409 * builtin-types.def: The new needed builtin types for the above.
5411 2018-05-04 Richard Biener <rguenther@suse.de>
5413 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
5414 * gimple-ssa-store-merging.c
5415 (imm_store_chain_info::output_merged_store): Remove redundant create,
5416 release split_store vector contents on failure.
5417 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
5418 scalar stmt vector on cache hit.
5420 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
5422 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
5424 * config.gcc (powerpc-xilinx-eabi*): Remove.
5425 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
5427 (fusion_addis_mem_combo_load): Ditto.
5428 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
5430 (rs6000_cpu_cpp_builtins): Ditto.
5431 * config/rs6000/rs6000-linux.c
5432 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
5433 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
5434 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
5436 (rs6000_setup_reg_addr_masks): Ditto.
5437 (rs6000_init_hard_regno_mode_ok): Ditto.
5438 (rs6000_option_override_internal): Ditto.
5439 (legitimate_lo_sum_address_p): Ditto.
5440 (rs6000_legitimize_address): Ditto.
5441 (rs6000_legitimize_reload_address): Ditto.
5442 (rs6000_legitimate_address_p): Ditto.
5443 (abi_v4_pass_in_fpr): Ditto.
5444 (setup_incoming_varargs): Ditto.
5445 (rs6000_gimplify_va_arg): Ditto.
5446 (rs6000_split_multireg_move): Ditto.
5447 (rs6000_savres_strategy): Ditto.
5448 (rs6000_emit_prologue_components): Ditto.
5449 (rs6000_emit_epilogue_components): Ditto.
5450 (rs6000_emit_prologue): Ditto.
5451 (rs6000_emit_epilogue): Ditto.
5452 (rs6000_elf_file_end): Ditto.
5453 (rs6000_function_value): Ditto.
5454 (rs6000_libcall_value): Ditto.
5455 * config/rs6000/rs6000.h: Ditto.
5456 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
5457 (TARGET_MINMAX): ... this. New.
5458 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
5459 * config/rs6000/rs6000.md: Remove Xilinx FP support.
5460 (*movsi_internal1_single): Delete.
5461 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
5462 mfpu=, mxilinx-fpu): Delete.
5463 * config/rs6000/singlefp.h: Delete.
5464 * config/rs6000/sysv4.h: Remove Xilinx FP support.
5465 * config/rs6000/t-rs6000: Ditto.
5466 * config/rs6000/t-xilinx: Delete.
5467 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
5468 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
5469 (VStype_simple): Delete.
5470 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
5471 * config/rs6000/xfpu.h: Delete.
5472 * config/rs6000/xfpu.md: Delete.
5473 * config/rs6000/xilinx.h: Delete.
5474 * config/rs6000/xilinx.opt: Delete.
5475 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
5476 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
5478 2018-05-04 Tom de Vries <tom@codesourcery.com>
5481 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
5484 2018-05-04 Richard Biener <rguenther@suse.de>
5487 * tree-complex.c (update_complex_assignment): We are always in SSA form.
5488 (expand_complex_div_wide): Likewise.
5489 (expand_complex_operations_1): Likewise.
5490 (expand_complex_libcall): Preserve EH info of the original stmt.
5491 (tree_lower_complex): Handle removed blocks.
5492 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
5493 on complex multiplication and division libcall builtins.
5495 2018-05-04 Richard Biener <rguenther@suse.de>
5498 * fold-const.c (negate_expr_p): Restrict negation of operand
5499 zero of a division to when we know that can happen without
5501 (fold_negate_expr_1): Likewise.
5503 2018-05-04 Jakub Jelinek <jakub@redhat.com>
5506 * real.h (real_nextafter): Declare.
5507 * real.c (real_nextafter): New function.
5508 * fold-const-call.c (fold_const_nextafter): New function.
5509 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
5510 CASE_CFN_NEXTTOWARD.
5511 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
5512 even when arg1_mode is different from arg0_mode.
5514 2018-05-03 Nathan Sidwell <nathan@acm.org>
5516 * doc/extend.texi (Deprecated Features): Remove
5518 (Backwards Compatibility): Likewise.
5519 * doc/invoke.texi (C++ Language Options): Likewise.
5520 (C++ Dialect Options): Likewise.
5522 2018-05-03 Jakub Jelinek <jakub@redhat.com>
5525 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
5526 _mm512_mask_mullox_epi64): New intrinsics.
5528 2018-05-03 Tom de Vries <tom@codesourcery.com>
5531 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5532 dump files): Add offload-tree.
5534 2018-05-03 Richard Biener <rguenther@suse.de>
5536 PR tree-optimization/85615
5537 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
5538 to loops not nested in BBs loop father to avoid creating multi-entry
5541 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5543 PR tree-optimization/70291
5544 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
5545 arguments. Change return type to tree. Emit libcall as a new
5546 statement rather than replacing existing one when inplace_p is true.
5547 (expand_complex_multiplication_components): New function.
5548 (expand_complex_multiplication): Expand floating-point complex
5549 multiplication using the above.
5550 (expand_complex_division): Rename inner_type parameter to type.
5551 Update expand_complex_libcall call-site.
5552 (expand_complex_operations_1): Update expand_complex_multiplication
5553 and expand_complex_division call-sites.
5555 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5558 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5559 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5560 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
5561 the highest significant bit of the shift count mask is clear. In
5562 check whether and[sq]i3 is needed verify that all significant bits
5563 of the shift count other than the highest are set.
5565 2018-05-02 Tom de Vries <tom@codesourcery.com>
5568 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
5569 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
5570 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
5571 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
5572 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
5573 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
5574 __builtin_goacc_parlevel_size.
5576 2018-05-02 Richard Biener <rguenther@suse.de>
5578 PR tree-optimization/85597
5579 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
5580 do not use split vect_get_vec_defs call but call vect_get_slp_defs
5583 2018-05-02 Tom de Vries <tom@codesourcery.com>
5586 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5587 dump files): Add ltrans-tree.
5589 2018-05-02 Tom de Vries <tom@codesourcery.com>
5592 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5593 dump files): Add wpa-ipa.
5595 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
5597 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
5598 powerpc*-*-linux*paired* target.
5599 * config/rs6000/750cl.h: Delete.
5600 * config/rs6000/paired.h: Delete.
5601 * config/rs6000/paired.md: Delete.
5602 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
5604 * config/rs6000/rs6000-builtin.def: Remove paired float support.
5605 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
5606 comment. Remove paired float support.
5607 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
5608 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
5610 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
5611 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
5613 * config/rs6000/rs6000.c: Remove paired float support.
5614 (paired_expand_vector_init, paired_expand_vector_move,
5615 paired_emit_vector_compare, paired_emit_vector_cond_expr,
5616 (paired_expand_lv_builtin, paired_expand_stv_builtin,
5617 paired_expand_builtin, paired_expand_predicate_builtin,
5618 paired_init_builtins): Delete.
5619 * config/rs6000/rs6000.h: Remove paired float support.
5620 * config/rs6000/rs6000.md: Remove paired float support.
5621 (move_from_CR_ov_bit): Delete.
5622 * config/rs6000/rs6000.opt (mpaired): Delete.
5623 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
5624 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
5626 2018-05-02 Richard Biener <rguenther@suse.de>
5629 * gimplify.c (gimplify_save_expr): When in SSA form allow
5630 SAVE_EXPRs to compute to SSA vars.
5632 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5635 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5636 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5637 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
5638 clobber operands[2], instead use a new pseudo. Formatting fixes.
5640 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
5642 PR tree-optimization/85586
5643 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
5644 exit early for statements in the same group if the accesses are
5647 2018-05-02 Tom de Vries <tom@codesourcery.com>
5650 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
5653 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
5655 PR tree-optimization/85143
5656 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
5658 2018-05-01 Tom de Vries <tom@codesourcery.com>
5661 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
5662 not found" error message.
5664 2018-05-01 Tom de Vries <tom@codesourcery.com>
5667 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
5668 * vec.c (test_ordered_remove_if): New function.
5669 (vec_c_tests): Call test_ordered_remove_if.
5670 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
5671 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
5672 * tree-vect-patterns.c (vect_pattern_recog_1): Use
5673 VEC_ORDERED_REMOVE_IF.
5675 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5677 PR tree-optimization/82665
5678 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
5679 pointer subtraction where arguments come from a memchr call.
5681 2018-05-01 Jakub Jelinek <jakub@redhat.com>
5683 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
5684 --push-state --as-needed and --pop-state instead of --as-needed and
5685 --no-as-needed if ld supports it.
5686 * configure: Regenerated.
5689 * doc/install.texi2html: Replace _002d with - and _002a with * in
5690 generated html files using sed.
5692 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5695 * gcc-rich-location.c (blank_line_before_p): New function.
5696 (use_new_line): New function.
5697 (gcc_rich_location::add_fixit_insert_formatted): New function.
5698 * gcc-rich-location.h
5699 (gcc_rich_location::add_fixit_insert_formatted): New function.
5701 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5703 * selftest.c (assert_streq): Rename "expected" and "actual" to
5704 "val1" and "val2". Extend NULL-handling to cover both inputs
5705 symmetrically, while still requiring both to be non-NULL for a pass.
5706 * selftest.h (assert_streq): Rename "expected" and "actual" to
5708 (ASSERT_EQ): Likewise.
5709 (ASSERT_EQ_AT): Likewise.
5710 (ASSERT_KNOWN_EQ): Likewise.
5711 (ASSERT_KNOWN_EQ_AT): Likewise.
5712 (ASSERT_NE): Likewise.
5713 (ASSERT_MAYBE_NE): Likewise.
5714 (ASSERT_MAYBE_NE_AT): Likewise.
5715 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
5716 the assertion to pass.
5717 (ASSERT_STREQ_AT): Likewise.
5719 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5721 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
5722 interaction with -pie.
5724 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5726 * selftest.h: Fix alphabetization of per-source-file selftest
5729 2018-04-30 Jason Merrill <jason@redhat.com>
5731 PR c++/61982 - dead stores to destroyed objects.
5732 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
5735 2018-04-30 Jason Merrill <jason@redhat.com>
5737 * tree.c (build_clobber): New.
5738 * tree.h: Declare it.
5739 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
5741 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5743 * diagnostic-show-locus.c (layout::layout): Update for
5744 location_get_source_line returning a char_span.
5745 (struct char_span): Move to input.h.
5746 (struct correction): Update for fields in char_span becoming
5748 (struct source_line): Update for location_get_source_line
5749 returning a char_span.
5750 (layout::print_line): Likewise.
5751 * edit-context.c (edited_file::print_content): Likewise.
5752 (edited_file::print_diff_hunk): Likewise.
5753 (edited_file::print_run_of_changed_lines): Likewise.
5754 (edited_file::get_num_lines): Likewise.
5755 (edited_line::edited_line): Likewise.
5756 * final.c (asm_show_source): Likewise.
5757 * input.c (location_get_source_line): Convert return type
5758 from const char * to char_span, losing the final "line_len"
5760 (dump_location_info): Update for the above.
5761 (get_substring_ranges_for_loc): Likewise. Use a char_span
5762 when handling the literal within the line.
5763 (test_reading_source_line): Update for location_get_source_line
5764 returning a char_span.
5765 * input.h (class char_span): Move here from
5766 diagnostic-show-locus.c, converting from a struct to a class.
5767 Make data members private.
5768 (char_span::operator bool): New.
5769 (char_span::length): New.
5770 (char_span::get_buffer): New.
5771 (char_span::operator[]): New.
5772 (char_span::subspan): Make const.
5773 (char_span::xstrdup): New.
5774 (location_get_source_line): Convert return type from const char *
5775 to char_span, losing the final "line_size" param.
5777 2018-04-30 Jan Hubicka <jh@suse.cz>
5779 * lto-wrapper.c (ltrans_priorities): New static var.
5780 (cmp_priority): New.
5781 (run_gcc): Read priorities and if doing parallel build order
5782 the Makefile by them.
5784 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5786 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
5788 2018-04-30 Richard Biener <rguenther@suse.de>
5790 * tree-cfg.c (verify_address): Remove base argument, add
5791 flag whether to check TREE_ADDRESSABLE and do that.
5792 (verify_expr): Remove.
5793 (verify_types_in_gimple_reference): Add pieces from verify_expr.
5794 (verify_gimple_assign_single): Likewise.
5795 (verify_gimple_switch): Likewise.
5796 (verify_expr_location_1): Dereference tp once. Add (disabled)
5797 piece from verify_expr.
5798 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
5800 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5802 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
5804 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5806 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
5807 (small_data_pattern): Likewise.
5808 (arc_rewrite_small_data): Likewise.
5809 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
5810 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
5811 (get_symbol_alignment): New function.
5812 (legitimate_small_data_address_p): Likewise.
5813 (legitimate_scaled_address): Update, call
5814 legitimate_small_data_address_p.
5815 (output_sdata): New static variable.
5816 (arc_print_operand): Update how we handle small data operands.
5817 (arc_print_operand_address): Likewise.
5818 (arc_legitimate_address_p): Update, use
5819 legitimate_small_data_address_p.
5820 (arc_rewrite_small_data_p): Remove.
5821 (arc_rewrite_small_data_1): Likewise.
5822 (arc_rewrite_small_data): Likewise.
5823 (small_data_pattern): Likewise.
5824 (compact_sda_memory_operand): Update to use
5825 legitimate_small_data_address_p and get_symbol_alignment.
5826 (prepare_move_operands): Don't rewite sdata pattern.
5827 (prepare_extend_operands): Remove.
5828 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
5830 (zero_extendqisi2): Likewise.
5831 (zero_extendhisi2): Likewise.
5832 (extendqihi2): Likewise.
5833 (extendqisi2): Likewise.
5834 (extendhisi2): Likewise.
5838 * config/arc/constraints.md (Usd): Change it to memory constraint.
5840 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5842 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
5843 as source of std instructions.
5844 * config/arc/arc.md (movsi_insn): Update pattern predicate to
5845 allow 6-bit constants as source for store instructions.
5846 (movdi_insn): Update instruction pattern to allow 6-bit constants
5847 as source for store instructions.
5849 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5851 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
5853 2018-04-30 Nathan Sidwell <nathan@acm.org>
5854 Sandra Loosemore <sandra@codesourcery.com>
5856 * dumpfile.c (dump_open): Allow '-' for stdout.
5857 * doc/invoke.texi (Developer Options): Document dump filename
5858 determination early. Document stdin/stdout selection.
5860 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
5862 Microblaze Target: PIC data text relative
5864 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
5865 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
5867 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
5868 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
5869 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
5870 New addressing mode for data-text relative position indepenedent code.
5871 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
5872 'ADDRESS_SYMBOLIC_TXT_REL'.
5873 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
5874 (microblaze_legitimate_pic_operand): Exclude function calls from
5875 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
5876 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
5878 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5879 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5880 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
5881 for 'address + offset'.
5882 (microblaze_expand_prologue): Add new function prologue call for
5884 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
5885 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
5886 table in case of TARGET_PIC_DATA_TEXT_REL.
5887 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
5888 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
5889 Add new macros 'UNSPEC_TEXT',
5890 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
5891 + exclude function calls from 'UNSPEC_PLT' in case of data text
5893 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5894 new target hook for generating address diff vector tables in case of
5896 * doc/tm.texi : Regenerate.
5897 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
5898 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
5899 of addr diff vector generation.
5900 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5901 target hook definition.
5902 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
5903 Add default function for generate_pic_addr_diff_vec -> flag_pic.
5904 * doc/invoke.texi (Add new pic option): Add new microblaze pic
5905 option for data text relative.
5907 2018-04-30 Richard Biener <rguenther@suse.de>
5909 * tree-chrec.h (evolution_function_is_constant_p): Remove
5911 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
5913 2018-04-30 Richard Biener <rguenther@suse.de>
5916 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
5918 2018-04-30 Richard Biener <rguenther@suse.de>
5920 PR tree-optimization/28364
5921 PR tree-optimization/85275
5922 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
5923 copying first exit test.
5925 2018-04-28 Mark Wielaard <mark@klomp.org>
5927 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
5929 (dwarf_AT): Handle DW_AT_addr_base.
5930 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
5932 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
5935 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
5936 (*ashl<dwi>3_doubleword_mask_1): Ditto.
5937 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
5938 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
5940 2018-04-28 Richard Biener <rguenther@suse.de>
5942 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
5943 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
5944 to reflect use. Only add interesting stmts.
5946 2018-04-27 Martin Jambor <mjambor@suse.cz>
5949 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
5950 the jump function allows for passing through aggregate values.
5952 2018-04-27 David Malcolm <dmalcolm@redhat.com>
5954 * input.h (in_system_header_at): Convert from macro to inline
5956 (from_macro_expansion_at): Likewise.
5957 (from_macro_definition_at): Likewise.
5959 2018-04-27 Jeff Law <law@redhat.com>
5961 * config.gcc: Mark tile* targets as deprecated/obsolete.
5963 2018-04-27 Richard Biener <rguenther@suse.de>
5965 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
5968 2018-04-27 Richard Biener <rguenther@suse.de>
5970 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
5972 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
5974 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
5975 with Yd constraint. Set "preferred_for_speed" attribute from
5976 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
5978 (*movdi_internal): Ditto.
5979 (movti_interunit splitters): Remove
5980 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
5981 (movdi_interunit splitters): Ditto.
5982 * config/i386/constraints.md (Ye): Remove.
5983 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
5985 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5988 * config/aarch64/constraints.md (Usg): Limit to 31.
5991 2018-04-27 Jakub Jelinek <jakub@redhat.com>
5993 PR tree-optimization/85529
5994 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
5995 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
5996 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
5997 zero extension or masking of the MSB bit.
5998 (optimize_range_tests): Add FIRST_BB argument, pass it through
5999 to optimize_range_tests_var_bound.
6000 (maybe_optimize_range_tests, reassociate_bb): Adjust
6001 optimize_range_tests callers.
6003 2018-04-26 Richard Biener <rguenther@suse.de>
6004 Jakub Jelinek <jakub@redhat.com>
6006 * cgraph.h (symbol_table): Just declare debug method here.
6007 * symtab.c (symbol_table::debug): Define.
6009 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
6011 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
6013 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
6015 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
6016 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
6017 (*movdi_internal): Substitute Yi and Yj constraint with x
6018 and Ym and Yn constraint with y constraint. Update "isa"
6019 attribute and set "preferred_for_speed" attribute from
6020 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
6021 (*movsi_internal): Ditto.
6022 (*movdf_internal): Ditto.
6023 (*movsf_internal): Ditto.
6024 (*zero_extendsidi2): Ditto.
6025 * config/i386/sse.md (vec_set<mode>_0): Ditto.
6026 (sse2_loadld): Ditto.
6027 (*vec_extract<ssevecmodelower>_0): Ditto.
6028 (*vec_extractv4si_0_zext_sse4): Ditto.
6029 (vec_concatv2di): Ditto.
6030 (*vec_dup<mode>): Ditto.
6031 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
6032 * config/i386/constraints.md (Yi): Remove.
6037 2018-04-26 Nathan Sidwell <nathan@acm.org>
6039 * dumpfile.c (dump_open): New.
6040 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
6041 (dump_finish): Detect stdio/stderr by value not name.
6043 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
6045 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
6047 2018-04-26 Tom de Vries <tom@codesourcery.com>
6050 * config/nvptx/nvptx.c (verify_neutering_jumps)
6051 (verify_neutering_labels): New function
6052 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
6054 2018-04-26 Tom de Vries <tom@codesourcery.com>
6057 * config/nvptx/nvptx.c (needs_neutering_p): New function.
6058 (nvptx_single): Use needs_neutering_p to skip over insns that do not
6061 2018-04-26 Richard Biener <rguenther@suse.de>
6062 Tom de Vries <tom@codesourcery.com>
6065 * lto-streamer-out.c (output_function): Fixup loops if required to match
6066 discovery done in the reader.
6068 2018-04-26 Richard Biener <rguenther@suse.de>
6070 PR tree-optimization/85116
6071 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
6072 have a loop exit from the single latch predecessor. Remove
6073 case of header with just condition.
6074 (ch_base::copy_headers): Exclude infinite loops from any
6076 (pass_ch::execute): Record exits.
6078 2018-04-26 Richard Biener <rguenther@suse.de>
6080 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
6081 prologue cost vector and pass it to vect_get_load_cost.
6082 (vect_get_peeling_costs_all_drs): Likewise.
6083 (vect_peeling_hash_get_lowest_cost): Likewise.
6084 (vect_enhance_data_refs_alignment): Likewise.
6086 2018-04-26 Richard Biener <rguenther@suse.de>
6089 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
6090 checking of integer<->pointer conversions.
6091 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
6092 sign-/zero-extending pointer types.
6093 (expand_omp_for_static_chunk): Likewise.
6095 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
6096 Jean Lee <xiaoyur347@gmail.com>
6098 * config/mips/mips.c (mips_asan_shadow_offset): New function.
6099 (TARGET_ASAN_SHADOW_OFFSET): Define.
6100 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
6101 true for -fsanitize=address.
6103 2018-04-25 Mark Wielaard <mark@klomp.org>
6105 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
6108 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6110 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
6111 than "alu", remove explicit "memory" and "imm_disp" attributes.
6112 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
6115 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
6116 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
6117 gen_lowpart_no_emit.
6119 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
6122 * config/i386/i386.c (ix86_expand_builtin): Change memory
6123 operand to XI, extend p0 to Pmode.
6124 * config/i386/i386.md: Change unspec volatile and operand
6125 1 mode to XI, change operand 0 mode to P.
6127 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6129 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
6130 GET_MODE_MASK before any checking.
6131 (nds32_can_use_bset_p): Likewise.
6132 (nds32_can_use_btgl_p): Likewise.
6134 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6136 * config/nds32/nds32-doubleword.md: New define_split pattern for
6137 illegal register number.
6139 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6141 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
6143 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6145 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
6147 2018-04-25 Richard Biener <rguenther@suse.de>
6149 * lto-streamer.h (LTO_major_version): Bump to 8.
6151 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6153 * BASE-VER: Set to 9.0.0.
6155 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
6157 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
6158 in __abskf2 and __powikf2.
6160 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6163 * config/aarch64/constraints.md (Usg, Usj): New constraints.
6164 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
6165 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
6166 Use the above on operand 2. Reindent.
6167 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
6169 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6172 * common/config/i386/i386-common.c (ix86_handle_option): Don't
6174 * config/i386/i386.opt (mcet): Removed.
6175 * doc/install.texi: Remove -mcet documentation.
6176 * doc/invoke.texi: Likewise.
6178 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6181 * doc/install.texi: Remove -mcet from bootstrap-cet.
6183 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6186 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
6187 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
6191 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
6192 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
6193 containing a CONST_VECTOR.
6195 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
6197 * doc/install.texi: Update newlib dependency for nvptx.
6199 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6202 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
6203 instead of INTVAL when shifting x left.
6205 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
6207 PR tree-optimization/85478
6208 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
6209 vect_grouped_store_supported for single element vectors.
6211 2018-04-24 Richard Biener <rguenther@suse.de>
6214 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
6215 load cost increase to the case of non-constant step.
6217 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6220 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
6221 destination if any_malformed_asm.
6223 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
6226 * expr.c (store_field): In the bitfield case, if the value comes from
6227 a function call and is returned in registers by means of a PARALLEL,
6228 do not change the mode of the temporary unless BLKmode and VOIDmode.
6230 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
6232 PR rtl-optimization/85423
6233 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
6234 dependencies to debug insns when the previous insn is non-debug.
6236 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6238 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
6239 enums into a single definition.
6240 (fls): Fix predicates and printing.
6243 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6245 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
6246 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
6247 and short u6 immediate.
6248 (check_if_valid_sleep_operand): Remove.
6249 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
6251 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6253 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
6254 flag_always_save_lp condition.
6255 * config/nds32/nds32.opt (malways-save-lp): New option.
6257 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6259 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
6260 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
6261 * config/nds32/nds32.h
6262 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
6263 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
6265 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6267 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
6268 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
6270 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6271 Chung-Ju Wu <jasonwucj@gmail.com>
6273 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
6275 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
6276 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
6278 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6280 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
6282 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6284 * config/nds32/nds32-protos.h (nds32_data_alignment,
6285 nds32_local_alignment): Declare.
6286 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
6287 nds32_local_alignment): New functions.
6288 (TARGET_CONSTANT_ALIGNMENT): Define.
6289 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
6291 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6293 * config/nds32/nds32.c
6294 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
6295 (TARGET_MODES_TIEABLE_P): Likewise.
6297 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6299 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
6302 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
6303 Chung-Ju Wu <jasonwucj@gmail.com>
6305 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6306 for unaligned access.
6307 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6309 * config/nds32/nds32-intrinsic.md: Likewise.
6310 * config/nds32/nds32_intrinsic.h: Likewise.
6311 * config/nds32/nds32.h (nds32_builtins): Likewise.
6312 * config/nds32/nds32.opt (munaligned-access): New option.
6313 * config/nds32/nds32.c (nds32_asm_file_start): Display
6314 flag_unaligned_access status.
6316 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6318 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
6319 -mno-relax is present.
6320 * config/riscv/linux.h (LINK_SPEC): Ditto.
6322 2018-04-20 Martin Sebor <msebor@redhat.com>
6325 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
6327 (gimple_fold_builtin_stxcpy_chk): Same.
6328 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
6330 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
6333 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
6334 __powikf2 when long double is IEEE 128-bit.
6336 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6338 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
6339 step to make sure stack always aligned.
6341 2018-04-20 Carl Love <cel@us.ibm.com>
6344 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
6345 size check for arg0.
6347 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
6348 Tom de Vries <tom@codesourcery.com>
6351 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
6352 Emit insns for calls too.
6353 (nvptx_find_par): Always look for worker-level predecessor insn.
6354 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
6356 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
6357 (nvptx_process_pars): Propagate frames for calls.
6359 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
6362 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
6364 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
6365 (ix86_handle_option): Don't handle OPT_mibt.
6366 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
6368 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
6370 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
6371 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
6372 (ix86_target_macros): Define __CET__ with flag_cf_protection
6373 for -fcf-protection.
6374 * config/i386/i386.c (isa2_opts): Remove -mibt.
6375 * config/i386/i386.h (TARGET_IBT): Removed.
6376 (TARGET_IBT_P): Likewise.
6377 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
6378 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
6379 * config/i386/i386.opt (mcet): Update help message.
6382 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
6383 -mcet as an alias for -mshstk.
6385 2018-04-20 Richard Biener <rguenther@suse.de>
6388 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
6389 complexity by forcing a single use of the multiply operand.
6391 2018-04-20 Martin Jambor <mjambor@suse.cz>
6394 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
6395 recursion dependency to only apply to non-clones.
6397 2018-04-20 Martin Jambor <mjambor@suse.cz>
6400 * ipa-cp.c (create_specialized_node): Check that clones of
6401 self-recursive edges exist during IPA-CP.
6403 2018-04-19 Toon Moene <toon@moene.org>
6405 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
6408 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6410 PR tree-optimization/85467
6411 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
6412 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
6413 VECTOR_CST element to type.
6415 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6418 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
6419 * config/i386/i386.md (builtin_setjmp_setup): Removed.
6420 (builtin_longjmp): Likewise.
6421 (save_stack_nonlocal): New pattern.
6422 (restore_stack_nonlocal): Likewise.
6424 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6427 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6428 Replace ASM_OUTPUT_LABEL with fprintf.
6430 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6433 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6434 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
6435 * config/i386/i386-c.c (ix86_target_macros_internal): Also
6436 define __IBT__ and __SHSTK__ for -fcf-protection.
6437 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
6439 (ix86_trampoline_init): Likewise.
6440 (x86_output_mi_thunk): Likewise.
6441 (ix86_notrack_prefixed_insn_p): Likewise.
6442 (ix86_option_override_internal): Don't disallow -fcf-protection.
6443 * config/i386/i386.md (rdssp<mode>): Also enable for
6445 (incssp<mode>): Likewise.
6446 (nop_endbr): Likewise.
6447 * config/i386/i386.opt (mcet): Change help message to built-in
6451 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
6452 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
6453 enable CET built-in functions.
6455 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
6457 * common/config/i386/i386-common.c
6458 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
6459 OPTION_MASK_ISA_MOVDIRI_UNSET,
6460 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
6461 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
6462 * config.gcc (movdirintrin.h): New header.
6463 * config/i386/cpuid.h (bit_MOVDIRI,
6464 bit_MOVDIR64B): New bits.
6465 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
6467 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
6468 (VOID, PVOID, PCVOID)): New function types.
6469 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
6470 __builtin_ia32_directstoreu_u64,
6471 __builtin_ia32_movdir64b): New builtins.
6472 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
6473 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
6475 (ix86_valid_target_attribute_inner_p): Ditto.
6476 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
6477 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
6478 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
6479 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
6480 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
6481 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
6482 (movdiri<mode>, movdir64b_<mode>): New.
6483 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
6484 * config/i386/immintrin.h: Include movdirintrin.h.
6485 * config/i386/movdirintrin.h: New file.
6486 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
6488 2018-04-19 Richard Biener <rguenther@suse.de>
6491 * cfg.c (clear_bb_flags): When loop state says we have
6492 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
6494 2018-04-19 Richard Biener <rguenther@suse.de>
6496 PR tree-optimization/84737
6497 * tree-vect-data-refs.c (vect_copy_ref_info): New function
6498 copying restrict info.
6499 (vect_setup_realignment): Use it.
6500 * tree-vectorizer.h (vect_copy_ref_info): Declare.
6501 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
6502 the first DR to all generated stores.
6503 (vectorizable_load): Likewise for loads.
6505 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6507 PR tree-optimization/85446
6508 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
6509 the integral and pointer types to have the same precision.
6511 * doc/install.texi: Document --disable-cet being the default and
6514 2018-04-18 Martin Liska <mliska@suse.cz>
6516 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
6519 2018-04-18 Martin Liska <mliska@suse.cz>
6522 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6525 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6526 arguments if they are comparable.
6528 2018-04-18 Martin Liska <mliska@suse.cz>
6531 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6534 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6537 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
6540 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
6541 ENDBR after calling __morestack.
6543 2018-04-18 David Malcolm <dmalcolm@redhat.com>
6546 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
6547 by using gcc_base_ver to generate a gcc_driver_version, and use
6548 it when generating GCC_DRIVER_NAME.
6549 * configure: Regenerate.
6551 2018-04-18 Jakub Jelinek <jakub@redhat.com>
6554 * config.gcc: Obsolete powerpc*-*-*spe*.
6556 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6559 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
6560 (stabstr_D): Change type of unum from unsigned int to
6561 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
6564 2018-04-17 Jim Wilson <jimw@sifive.com>
6567 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
6568 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
6569 Set arg_pointer_offset after using pretend_args_size.
6571 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6573 PR rtl-optimization/85431
6574 * dse.c (record_store): Ignore zero width stores.
6577 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
6578 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
6579 __builtin_stack_restore rather than after it.
6580 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
6581 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
6582 argument instead of virtual_dynamic_stack_rtx.
6584 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
6586 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
6588 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
6589 Add note to error message to explain internal mapping of overloaded
6590 built-in function name to non-overloaded built-in function name.
6591 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
6594 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6597 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
6598 where the inputs overlap with the output.
6600 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6602 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
6603 (=v, v) alternative and explicit "memory" attribute.
6604 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
6605 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6607 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
6608 "sselog1" type instead of "sselog".
6609 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
6610 "sselog". Remove explicit "memory" attribute.
6611 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
6612 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6614 (vec_extract_hi_v32hi): Merge all alternatives into one, use
6615 "sselog1" type instead of "sselog". Remove explicit "memory"
6617 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
6618 use "sselog1" type instead of "sselog". Remove explicit "memory"
6620 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
6621 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6623 (vec_extract_hi_v64qi): Merge all alternatives into one, use
6624 "sselog1" type instead of "sselog". Remove explicit "memory"
6626 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
6627 use "sselog1" type instead of "sselog". Remove explicit "memory"
6631 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
6634 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
6637 2018-04-17 Martin Jambor <mjambor@suse.cz>
6640 * ipa-cp.c (create_specialized_node): Call
6641 expand_all_artificial_thunks if necessary.
6643 2018-04-17 Martin Liska <mliska@suse.cz>
6646 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
6647 in message, remote space in between '_G' and '('.
6649 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6652 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
6653 avx512f_vmcmp<mode>3<round_saeonly_name>,
6654 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
6655 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
6656 avx512f_rndscale<mode><round_saeonly_name>,
6657 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
6658 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
6659 Use %<iptr>2 instead of %2 for -masm=intel.
6660 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
6661 avx512f_vcvttss2usi<round_saeonly_name>,
6662 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
6664 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
6665 avx512f_vcvttsd2usi<round_saeonly_name>,
6666 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
6667 Use %q1 instead of %1 for -masm=intel.
6668 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
6669 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
6670 of %3 for -masm=intel.
6671 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
6673 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
6675 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
6677 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
6678 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
6680 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
6682 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
6683 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
6684 %g1 and one with %0 and %1.
6685 (avx512er_vmrcp28<mode><round_saeonly_name>,
6686 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
6688 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
6689 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
6690 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
6691 of %0 and %{%4%} for -masm=intel.
6692 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
6693 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
6694 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
6695 order of %0 and %{%5%}%{z%} for -masm=intel.
6697 2018-04-17 Jan Hubicka <jh@suse.cz>
6700 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
6702 2018-04-17 Martin Liska <mliska@suse.cz>
6705 * multiple_target.c (create_dispatcher_calls): Set apostrophes
6706 for target_clone error message. Make default implementation
6707 clone to be a local declaration.
6708 (separate_attrs): Add new argument and check for an empty
6710 (expand_target_clones): Handle it.
6711 (ipa_target_clone): Make redirection just for target_clones
6714 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
6715 Tom de Vries <tom@codesourcery.com>
6718 * omp-expand.c (expand_oacc_for): Add dummy false branch for
6719 tiled basic blocks without omp continue statements.
6721 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
6724 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
6725 vec_extract expression as having side effects to make sure it gets
6728 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
6731 * config/i386/i386.c (get_builtin_code_for_version): Check
6734 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
6737 * gcc/config.gcc: Support "skylake".
6738 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
6740 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
6741 (processor_target_table): Add "skylake".
6742 (ix86_option_override_internal): Add "skylake".
6743 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
6744 PROCESSOR_CANNONLAKE.
6745 (get_builtin_code_for_version): Fix priority for
6746 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
6747 PROCESSOR_SKYLAKE-AVX512.
6748 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
6749 (processor_type): Add PROCESSOR_SKYLAKE.
6751 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
6752 Jason Merrill <jason@redhat.com>
6755 * convert.c (convert_to_integer_1): Use direct recursion for
6756 enumeral types and types with a precision less than the number
6757 of bits in their mode.
6759 2018-04-16 Julia Koval <julia.koval@intel.com>
6762 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
6763 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
6765 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
6768 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
6769 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
6770 and -mno-direct-move.
6772 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
6775 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
6776 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
6777 Ensure negative shifts result in {0}.
6779 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
6781 PR rtl-optimization/79916
6782 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
6783 regs (if any) to define how to gnerate SD moves when LRA is in
6786 2018-04-13 Jakub Jelinek <jakub@redhat.com>
6788 PR rtl-optimization/85393
6789 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
6790 * except.c (expand_dw2_landing_pad_for_region): Make static.
6791 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
6792 a label and unconditional jump to old_bb, rather than
6793 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
6796 PR rtl-optimization/85376
6797 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
6798 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
6799 instead of a specific value.
6801 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6802 Bin Cheng <bin.cheng@arm.com>
6804 PR tree-optimization/82965
6805 PR tree-optimization/83991
6806 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
6807 by_profile_only parameter.
6808 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
6809 information if the loop was predicted to iterate too many times.
6810 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
6812 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6815 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
6818 2018-04-13 Martin Liska <mliska@suse.cz>
6819 Jakub Jelinek <jakub@redhat.com>
6822 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
6823 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
6824 * builtins.c (expand_builtin_memory_copy_args): Use
6825 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
6826 handle dest_addr == pc_rtx.
6828 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
6831 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
6832 asked to not generate direct moves.
6833 (fix_trunc<mode>si2_stfiwx): Similar.
6834 (fix_trunc<mode>si2_internal): Similar.
6836 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6839 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
6840 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
6841 lookup if dest in some wider mode is known to be const0_rtx and
6842 if so, record permanent equivalence for it to be ZERO_EXTEND of
6843 the narrower mode destination.
6845 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6847 * lto-streamer-out.c (output_function): Revert 259346.
6848 * omp-expand.c (expand_oacc_for): Likewise.
6850 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
6852 PR rtl-optimization/85354
6853 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
6854 * sel-sched.c (sel_global_init): ... here.
6856 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
6859 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
6860 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6861 mode for PE-COFF targets.
6862 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
6863 (i386_pe_asm_lto_end): Likewise.
6864 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
6865 (TARGET_ASM_LTO_END): Likewise.
6866 * config/i386/winnt.c (saved_debug_info_level): New static variable.
6867 (i386_pe_asm_lto_start): New function.
6868 (i386_pe_asm_lto_end): Likewise.
6870 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6871 Richard Biener <rguenther@suse.de>
6874 * lto-streamer-out.c (output_function): Fix CFG loop state before
6876 * omp-expand.c (expand_oacc_for): Handle calls to internal
6877 functions like regular functions.
6879 2018-04-12 Richard Biener <rguenther@suse.de>
6882 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
6883 for the early LTO debug to properly generate references to it
6884 during DIE emission. Do not re-use that for the skeleton for
6886 (dwarf2out_early_finish): Likewise.
6888 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6891 * config/i386/sse.md
6892 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
6893 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
6894 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
6895 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
6896 and output is a reg, avoid creating invalid lowpart subreg, but
6897 instead split into a 512-bit move. Don't split if not AVX512VL,
6898 input is xmm16+ reg and output is a mem.
6899 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
6900 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
6901 xmm16+ reg and output is a mem.
6903 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6905 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
6906 also for flag_dwarf2_cfi_asm.
6908 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6910 PR rtl-optimization/85342
6911 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
6912 a bool scalar var inside of the loop instead. Don't try to update
6913 recog_data.operand after failed apply_change_group.
6915 2018-04-12 Tom de Vries <tom@codesourcery.com>
6918 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
6919 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
6920 array with flexible array member as array without given dimension.
6921 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
6922 argument for undefined param to true.
6924 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
6927 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6928 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
6929 from PowerPC section.
6930 * config/rs6000/sysv4.opt (mcall-): Improve help text.
6931 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
6932 help text that is too long.
6933 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
6934 help text that is too long.
6935 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
6936 help text that is too long.
6938 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
6940 * config/alpha/alpha.md (stack_probe_internal): Rename
6941 from "probe_stack". Update all callers.
6943 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6945 PR rtl-optimization/84566
6946 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
6947 sched_macro_fuse_insns.
6949 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6952 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
6953 (compute_block_dependences): ... from here.
6955 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6957 PR tree-optimization/85331
6958 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
6959 from int to HOST_WIDE_INT.
6961 2018-04-11 Martin Jambor <mjambor@suse.cz>
6964 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
6965 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
6966 not the same as the source val.
6967 (cgraph_edge_brings_value_p): New parameter.
6968 (gather_edges_for_value): Pass destination value to
6969 cgraph_edge_brings_value_p.
6970 (perhaps_add_new_callers): Likewise.
6971 (get_info_about_necessary_edges): Likewise and exclude values brought
6972 only by self-recursive edges.
6973 (create_specialized_node): Redirect only clones of self-calling edges.
6974 (+self_recursive_pass_through_p): New function.
6975 (find_more_scalar_values_for_callers_subset): Use it.
6976 (find_aggregate_values_for_callers_subset): Likewise.
6977 (known_aggs_to_agg_replacement_list): Removed.
6978 (decide_whether_version_node): Re-calculate known constants for all
6979 remaining context clones.
6981 2018-04-11 Richard Biener <rguenther@suse.de>
6984 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
6985 from early DWARF output.
6986 (dwarf2out_early_finish): Output line info unconditionally into
6987 early DWARF and add reference to it.
6989 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6992 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
6993 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
6994 other than V2DFmode using iptr mode attribute.
6995 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
6997 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6999 PR rtl-optimization/84659
7000 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
7002 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7005 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
7007 (output_loc_list): Pass address of a dummy size variable even in the
7008 locview handling loop.
7009 (index_location_lists): Add comment on why skip_loc_list_entry can't
7012 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
7015 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
7018 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
7021 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7022 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
7023 and -mstring-compare-inline-limit.
7025 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7028 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
7029 for stack clash protection in a register whenever we need it to be in
7032 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7034 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
7035 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
7037 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7040 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
7042 (mlong-double-): Ditto.
7043 * config/rs6000/sysv4.opt (msdata=): Ditto.
7044 (mtls-size=): Ditto.
7046 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7048 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
7049 erroneous entries for
7050 "vector int vec_ldl (int, long int *)", and
7051 "vector unsigned int vec_ldl (int, unsigned long int *)".
7052 Add comments and entries for
7053 "vector bool char vec_ldl (int, bool char *)",
7054 "vector bool short vec_ldl (int, bool short *)",
7055 "vector bool int vec_ldl (int, bool int *)",
7056 "vector bool long long vec_ldl (int, bool long long *)",
7057 "vector pixel vec_ldl (int, pixel *)",
7058 "vector long long vec_ldl (int, long long *)",
7059 "vector unsigned long long vec_ldl (int, unsigned long long *)".
7060 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
7061 type tree bool_long_long_type_node and correct definition of
7062 bool_V2DI_type_node to make reference to this new type tree.
7063 (rs6000_mangle_type): Replace erroneous reference to
7064 bool_long_type_node with bool_long_long_type_node.
7065 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
7066 comments to emphasize sign distinctions for char and int types and
7067 replace RS6000_BTI_bool_long constant with
7068 RS6000_BTI_bool_long_long constant. Also add comment to restrict
7069 use of RS6000_BTI_pixel.
7070 (bool_long_type_node): Remove this macro definition.
7071 (bool_long_long_type_node): New macro definition
7073 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7075 PR rtl-optimization/85300
7076 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
7077 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
7078 simplify_unary_operation fails.
7080 2018-04-10 Martin Liska <mliska@suse.cz>
7082 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
7083 cgraph_edge and ipa_ref.
7085 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7089 * config/i386/sse.md
7090 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
7091 computation of the VEC_MERGE selector from mask.
7092 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
7093 Fix decoding of the VEC_MERGE selector into mask.
7095 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7097 PR tree-optimization/85286
7098 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
7100 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7102 * final.c (final_1): Set insn_last_address as well as
7103 insn_current_address.
7105 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7108 * explow.c (emit_stack_probe): Call validize_mem on memory location
7109 before passing it to gen_probe_stack. Create address operand and
7110 legitimize it for the probe_stack_address case.
7112 2018-04-09 Jan Hubicka <jh@suse.cz>
7115 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
7116 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
7117 * tree.c (free_lang_data_in_type): Fix handling of binfos;
7119 (free_lang_data): Rebuild type inheritance graph.
7121 2018-04-09 Martin Sebor <msebor@redhat.com>
7123 * invoke.texi (-finline-small-functions): Mention other optimization
7125 (-findirect-inlining, -fpartial-inlining): Same.
7126 (-finline-functions-called-once): Same.
7127 (-freorder-blocks-and-partition): Same.
7129 2018-04-09 Jan Hubicka <jh@suse.cz>
7132 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
7133 jumps; choose last target that matches the criteria (i.e.
7134 no partition changes for non-crossing jumps).
7135 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
7136 support for redirecting crossing jumps to non-crossing.
7138 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
7140 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
7141 also for naked functions.
7143 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
7145 * config/arc/arc.md (add_shift): New pattern.
7146 (add_shift2): Likewise.
7147 (sub_shift): Likewise.
7148 (sub_shift_cmp0_noout): Likewise.
7149 (compare_si_ashiftsi): Likewise.
7150 (xbfu_cmp0_noout): New combine pattern.
7151 (xbfu_cmp0"): Likewise.
7152 (movsi_set_cc_insn): Place the predicable variant first.
7153 (commutative_binary_cmp0_noout): Remove clobber.
7154 (commutative_binary_cmp0): New pattern.
7155 (noncommutative_binary_cmp0): Likewise.
7156 (noncommutative_binary_cmp0_noout): Likewise.
7157 (noncommutative_binary_comparison_result_used): Removed.
7158 (rsub_cmp0): New pattern.
7159 (rsub_cmp0_noout): Likewise.
7160 (extzvsi): Changed, keep only meaningful variants.
7161 (SQH, SEZ): New iterators.
7162 (SQH_postfix): New mode attribute.
7163 (SEZ_prefix): New code attribute.
7164 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
7165 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
7166 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
7168 (noncommutative_operator): Check the availability of barrel
7171 2018-04-09 Richard Biener <rguenther@suse.de>
7173 PR tree-optimization/85284
7174 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
7175 Only use the niter constraining form of simple_iv when the exit
7178 2018-04-09 Tom de Vries <tom@codesourcery.com>
7181 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7182 (define_expand "*memory_barrier"): New define_expand.
7183 (define_insn "memory_barrier"): New insn.
7185 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7187 PR rtl-optimization/80463
7188 PR rtl-optimization/83972
7189 PR rtl-optimization/83480
7191 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
7192 correct producer for the insn.
7193 (tidy_control_flow): Fixup seqnos in case of debug insns.
7195 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7197 PR rtl-optimization/83913
7199 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
7200 different sched-times when merging exprs.
7202 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7204 PR rtl-optimization/83962
7206 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
7207 tidy_fallthru_edge and tidy_control_flow.
7209 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7211 PR rtl-optimization/83530
7213 * sel-sched.c (force_next_insn): New global variable.
7214 (remove_insn_for_debug): When force_next_insn is true, also leave only
7215 next insn in the ready list.
7216 (sel_sched_region): When the region wasn't scheduled, make another pass
7217 over it with force_next_insn set to 1.
7219 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7221 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
7223 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7224 for interrupt control.
7225 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7226 functions for interrupt control.
7227 * config/nds32/nds32-intrinsic.md: Likewise.
7228 * config/nds32/nds32_intrinsic.h: Likewise.
7229 * config/nds32/nds32.h (nds32_builtins): Likewise.
7231 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7233 * config/nds32/nds32.c (nds32_init_machine_status,
7234 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
7235 strict_aligned_p field.
7236 (nds32_expand_to_rtl_hook): New function.
7237 (TARGET_EXPAND_TO_RTL_HOOK): Define.
7238 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
7240 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7241 Chung-Ju Wu <jasonwucj@gmail.com>
7243 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
7244 * config/nds32/nds32-n7.md: New file.
7245 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
7246 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
7248 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
7249 * config/nds32/nds32.md (pipeline_model): Add n7.
7250 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
7251 * config/nds32/pipelines.md: Include n7 settings.
7253 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7254 Chung-Ju Wu <jasonwucj@gmail.com>
7256 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
7257 * config/nds32/nds32-e8.md: New file.
7258 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
7259 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
7261 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
7262 * config/nds32/nds32.md (pipeline_model): Add e8.
7263 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
7264 * config/nds32/pipelines.md: Include e8 settings.
7266 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7267 Chung-Ju Wu <jasonwucj@gmail.com>
7269 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
7270 * config/nds32/nds32-n8.md: New file.
7271 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
7272 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
7274 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
7275 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
7276 * config/nds32/nds32.md (pipeline_model): Add n8.
7277 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
7278 * config/nds32/pipelines.md: Include n8 settings.
7280 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7281 Chung-Ju Wu <jasonwucj@gmail.com>
7283 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
7284 * config/nds32/nds32-n9-2r1w.md: New file.
7285 * config/nds32/nds32-n9-3r2w.md: New file.
7286 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
7287 nds32_register_ports): New or modify for cpu n9.
7288 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
7290 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
7291 * config/nds32/nds32-utils.c: New file.
7292 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
7293 TARGET_MUL_SLOW): Define.
7294 * config/nds32/nds32.md (pipeline_model): New attribute.
7295 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
7296 New options that support cpu n9.
7297 * config/nds32/pipelines.md: Include n9 settings.
7298 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
7300 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7302 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
7303 information if necessary.
7304 (output_cond_branch_compare_zero): Likewise.
7305 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
7306 (nds32_target_alignment): Refine for alignment.
7307 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
7308 (FUNCTION_BOUNDARY): Modify.
7309 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
7311 * config/nds32/nds32.opt (malways-align, malign-functions): New.
7313 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7315 * config/nds32/constants.md (unspec_volatile_element): Add values for
7316 TLB operation and data prefetch.
7317 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7318 functions for TLB operation and data prefetch.
7319 * config/nds32/nds32-intrinsic.md: Likewise.
7320 * config/nds32/nds32_intrinsic.h: Likewise.
7321 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
7322 (nds32_print_operand): Likewise.
7323 * config/nds32/nds32.h (nds32_builtins): Likewise.
7325 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
7326 Andrew Pinski <pinsika@gcc.gnu.org>
7329 * match.pd: Use constant_boolean_node of correct type instead of
7330 boolean_true_node or boolean_false_node for simplifying
7331 pointer comparisons to zero.
7333 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7335 PR tree-optimization/80021
7336 * tree.c (verify_type_variant): Make error call in verify_variant_match
7337 translatable and remove final full stop.
7339 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7341 * config/nds32/constants.md (unspec_volatile_element): Add
7342 UNSPEC_VOLATILE_EH_RETURN.
7343 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
7344 nds32_output_stack_pop): Support dwarf exception handling process.
7345 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
7346 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
7347 exception handling process.
7348 (nds32_compute_stack_frame): Likewise.
7349 (nds32_return_addr_rtx): Likewise.
7350 (nds32_initial_elimination_offset): Likewise.
7351 (nds32_expand_prologue): Likewise.
7352 (nds32_expand_epilogue): Likewise.
7353 (nds32_dynamic_chain_address): New function.
7354 * config/nds32/nds32.h (machine_function): Add fields for dwarf
7356 (DYNAMIC_CHAIN_ADDRESS): Define.
7357 (EH_RETURN_DATA_REGNO): Define.
7358 (EH_RETURN_STACKADJ_RTX): Define.
7359 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
7360 patterns for dwarf exception handling.
7362 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7364 * config/nds32/nds32.h: Clean up obsolete macros.
7366 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7368 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7369 Add enum values for particular instructions.
7370 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7371 particular intrinsic functions.
7372 * config/nds32/nds32-intrinsic.md: Likewise.
7373 * config/nds32/nds32_intrinsic.h: Likewise.
7374 * config/nds32/nds32.h (nds32_builtins): Likewise.
7375 * config/nds32/nds32.md (type): Add pbsad and pbsada.
7376 (btst, ave): New patterns for particular instructions.
7378 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7380 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7381 Add enum values for atomic load/store and memory sync.
7382 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
7384 * config/nds32/nds32-intrinsic.md: Likewise.
7385 * config/nds32/nds32_intrinsic.h: Likewise.
7386 * config/nds32/nds32.h (nds32_builtins): Likewise.
7388 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7390 PR tree-optimization/85257
7391 * fold-const.c (native_encode_vector): If not all elts could fit
7392 and off is -1, return 0 rather than offset.
7393 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
7394 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
7395 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
7396 adjust buffer in native_interpret_expr call.
7398 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7400 * config/nds32/constants.md (unspec_volatile_element): Add cache
7401 control enum values.
7402 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
7403 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
7404 * config/nds32/nds32.c (nds32_cctl_names): New.
7405 (nds32_print_operand): Handle cache control register names.
7406 * config/nds32/nds32.h (nds32_builtins): New enum values.
7407 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
7409 * config/nds32/nds32.md (type): Add mmu.
7410 * config/nds32/pipelines.md (simple_insn): Add mmu.
7412 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7414 * config/nds32/nds32.md (type): Remove call.
7415 * config/nds32/pipelines.md (simple_insn): Likewise.
7417 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7419 * config/nds32/constants.md (unspec_volatile_element): Add
7420 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
7421 UNSPEC_VOLATILE_FMFCFG.
7422 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
7423 description for fmfcfg and fmfcsr.
7424 (bdesc_1arg): Add fmtcsr.
7425 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
7426 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
7427 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
7428 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
7429 unspec_fmfcfg): New patterns.
7430 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
7431 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
7432 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
7433 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
7434 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
7435 __nds32__fmfcfg): Define.
7437 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7439 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
7440 intrinsic register names.
7441 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
7442 intrinsic register enum values and macros.
7444 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7446 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
7447 for load/store addressing form.
7448 (nds32_print_operand_address): Likewise.
7450 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
7453 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
7454 based on LABEL_REF. Remove useless assertion.
7455 (pic_address_needs_scratch): Fix formatting.
7456 (sparc_legitimize_pic_address): Minor tweaks.
7457 (sparc_delegitimize_address): Adjust assertion accordingly.
7458 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
7459 into symbolic_operand.
7460 (movsi_high_pic_label_ref): Likewise.
7461 (movsi_lo_sum_pic_label_ref): Likewise.
7462 (movdi_pic_label_ref): Likewise.
7463 (movdi_high_pic_label_ref): Likewise.
7464 (movdi_lo_sum_pic_label_ref): Likewise.
7466 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
7468 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
7469 custom LIB_SPEC setup.
7471 2018-04-06 Ruslan Bukin <br@bsdpad.com>
7472 Kito Cheng <kito.cheng@gmail.com>
7474 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
7475 * config/riscv/freebsd.h: New.
7477 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7479 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
7480 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
7483 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7484 Kito Cheng <kito.cheng@gmail.com>
7486 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
7487 nds32_output_call, nds32_symbol_binds_local_p): New functions.
7488 * config/nds32/nds32-protos.h (nds32_output_call,
7489 nds32_output_return): Declare.
7490 * config/nds32/nds32.md: Refine all the call and return patterns.
7492 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7495 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
7496 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
7498 PR rtl-optimization/84872
7499 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
7500 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
7503 2018-04-06 Tamar Christina <tamar.christina@arm.com>
7505 * expr.c (copy_blkmode_to_reg): Revert 254862.
7506 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
7508 2018-04-06 Richard Biener <rguenther@suse.de>
7511 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
7512 after seeing a component reference with an adjacent field. Treat
7513 refs to arrays at struct end of external decls similar to
7514 refs to unconstrained commons.
7516 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7519 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
7520 look through SAVE_EXPRs with non-side-effects argument. Adjust
7522 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
7525 2018-04-06 Richard Biener <rguenther@suse.de>
7528 * alias.c (find_base_term): New wrapper around find_base_term
7529 unwinding CSELIB_VAL_PTR changes.
7530 (find_base_term): Do not restore CSELIB_VAL_PTR during the
7533 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7535 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
7537 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
7538 constant definitions.
7539 ("nop"): lr 0,0 -> nopr r0
7540 ("nop_lr0", "nop_lr1"): New insn definitions.
7542 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7544 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
7545 NDS32_V3PUSH_AVAILABLE_P macro.
7547 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
7548 Chung-Ju Wu <jasonwucj@gmail.com>
7550 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
7551 (nds32*-*-*): Add float and fpu_config into supported_defaults.
7552 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
7553 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
7554 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
7555 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
7556 * config/nds32/constraints.md: New constraints and checking for hard
7557 float configuration.
7558 * config/nds32/iterators.md: New mode iterator and attribute for hard
7559 float configuration.
7560 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
7562 * config/nds32/nds32-fpu.md: New file.
7563 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
7564 deal with hard float code generation.
7565 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
7567 (abi_type, float_reg_number): New enum type.
7568 * config/nds32/nds32-predicates.c: New predicates for hard float.
7569 * config/nds32/nds32-protos.h: Declare functions for hard float.
7570 * config/nds32/nds32.c: Implementation for hard float configuration.
7571 * config/nds32/nds32.h: Definitions for hard float configuration.
7572 * config/nds32/nds32.md: Include hard float machine description and
7573 modify patterns for hard float configuration.
7574 * config/nds32/nds32.opt: New options for hard float configuration.
7575 * config/nds32/predicates.md: New predicates for hard float
7578 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7580 * common/config/nds32/nds32-common.c
7581 (nds32_option_optimization_table): Enable -mreleax-hint by default.
7583 2018-04-05 Jakub Jelinek <jakub@redhat.com>
7586 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
7587 CONSTRUCTOR_ELT (ctor, ...)->value.
7589 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
7592 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
7594 2018-04-05 Tom de Vries <tom@codesourcery.com>
7597 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
7600 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7601 Kito Cheng <kito.cheng@gmail.com>
7603 * config/nds32/constraints.md (U33): Fine-tune checking condition.
7604 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
7605 * config/nds32/nds32.h (nds32_16bit_address_type): Add
7606 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
7608 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7609 Kito Cheng <kito.cheng@gmail.com>
7611 * config/nds32/constraints.md (Ufe): New memory constraint.
7612 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
7613 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
7614 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
7616 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
7617 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
7619 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7621 * config/nds32/nds32.md: Use optimize_size in the condition for
7622 alu-shift instructions.
7624 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7626 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
7628 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7630 * config/nds32/nds32.md (negsi2): Refine pattern.
7632 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7633 Chung-Ju Wu <jasonwucj@gmail.com>
7635 * config/nds32/iterators.md (shift_rotate): New code iterator.
7636 (shift): New code attribute.
7637 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
7638 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
7639 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
7640 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
7641 bit-wise operations.
7642 (andsi3, *andsi3): Ditto.
7643 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
7644 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
7645 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
7646 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
7647 nds32_ior_operand, nds32_xor_operand): New predicates.
7649 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7651 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
7652 (addsi3, subsi3): ... this.
7654 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7656 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
7658 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7660 * config/nds32/nds32.md: Adjust indention.
7662 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7664 * config/nds32/nds32.md (feature): New attribute.
7666 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7668 * config/nds32/nds32.md (subtype): New attribute.
7670 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
7673 * config/arm/arm-builtins.c (arm_expand_builtin): Change
7674 expansion to perform a bitwise AND of the argument followed by a
7675 boolean negation of the result.
7677 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
7679 PR rtl-optimization/84878
7680 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
7681 the basic block. Assert the use reference is not artificial and that
7682 it has an associated insn.
7684 2018-04-04 Michael Matz <matz@suse.de>
7686 * builtins.c (compute_objsize): Pass correct operand
7687 to array_at_struct_end_p.
7689 2018-04-04 Richard Biener <rguenther@suse.de>
7692 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
7693 from contexts for DINFO_LEVEL_TERSE and below.
7695 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7697 * config/nds32/nds32-doubleword.md (move_<mode>): Require
7698 resiter_operand condition.
7699 * config/nds32/nds32.md (*move<mode>): Ditto.
7701 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7702 Monk Chiang <sh.chiang04@gmail.com>
7704 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
7706 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7708 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
7710 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7711 Kito Cheng <kito.cheng@gmail.com>
7713 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
7714 nds32_cond_code_str, output_cond_branch,
7715 output_cond_branch_compare_zero, nds32_expand_cbranch,
7716 nds32_expand_cstore, nds32_expand_movcc,
7717 nds32_output_cbranchsi4_equality_zero,
7718 nds32_output_cbranchsi4_equality_reg,
7719 nds32_output_cbranchsi4_equality_reg_or_const_int,
7720 nds32_output_cbranchsi4_greater_less_zero: New functions.
7721 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
7722 nds32_expand_cstore, nds32_expand_movcc,
7723 nds32_output_cbranchsi4_equality_zero,
7724 nds32_output_cbranchsi4_equality_reg,
7725 nds32_output_cbranchsi4_equality_reg_or_const_int,
7726 nds32_output_cbranchsi4_greater_less_zero): Declare.
7727 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
7728 nds32_rimm11s_operand): New predicates.
7729 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
7730 * config/nds32/nds32.md: Rewrite all the branch and conditional move
7733 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7735 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
7736 * config/nds32/nds32.md: Ditto.
7737 * config/nds32/pipelines.md: Ditto.
7739 2018-04-04 Richard Biener <rguenther@suse.de>
7741 PR tree-optimization/85168
7742 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
7743 propagating abnormals.
7745 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7747 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
7749 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7750 Kito Cheng <kito.cheng@gmail.com>
7752 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
7753 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
7754 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
7755 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
7756 * config/nds32/nds32.md (sibcall_internal): New.
7757 (sibcall_register): Remove.
7758 (sibcall_immediate): Remove.
7759 (sibcall_value_internal): New.
7760 (sibcall_value_register): Remove.
7761 (sibcall_value_immediate): Remove.
7762 * config/nds32/predicates.md (nds32_general_register_operand): New.
7763 (nds32_call_address_operand): New.
7765 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7767 PR rtl-optimization/85167
7768 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
7769 bb_defs if *split_p, instead preinitialize it to NULL.
7771 PR tree-optimization/85156
7772 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
7773 evaluating the argument multiple times.
7775 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
7777 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
7779 (_mm_cvtpd_ps): Likewise.
7780 (_mm_cvttpd_epi32): Likewise.
7781 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
7782 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
7783 vector, pixel, and bool following altivec.h include.
7785 2018-04-03 Martin Sebor <msebor@redhat.com>
7787 * doc/extend.texi (Common Function Attributes): Clarify.
7788 (const attribute): Likewise.
7789 (pure attribute): Likewise.
7791 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7794 * config/i386/i386.c (ix86_expand_vector_set): Use
7795 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
7797 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
7799 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
7800 instructions when changing rounding bits to preserve precision bits
7801 in the x87 control word.
7803 2018-04-03 Martin Liska <mliska@suse.cz>
7805 PR tree-optimization/82491
7806 * rtl.h (strip_offset_and_add): Replace += suboffset with
7807 poly_uint64 () + suboffset.
7809 2018-03-29 Martin Liska <mliska@suse.cz>
7810 Martin Jambor <mjambor@suse.cz>
7813 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
7814 param_type is not an integral or pointer type.
7816 2018-04-03 Richard Biener <rguenther@suse.de>
7818 * sese.h (recompute_all_dominators): Remove.
7820 2018-04-02 Martin Sebor <msebor@redhat.com>
7822 * doc/invoke.texi (-Wrestrict): Fix typos.
7824 2018-04-02 Jim Wilson <jimw@sifive.com>
7826 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
7827 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
7828 (<optab>di3, <optab>si3_extend): Likewise.
7829 (<optab>si3_mask, <optab>si3_mask_1): New.
7830 (<optab>di3_mask, <optab>di3_mask_1): New.
7831 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
7832 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
7833 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
7835 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
7837 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
7840 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
7842 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
7843 (nds32_canonicalize_comparison): New function.
7845 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7846 Kito Cheng <kito.cheng@gmail.com>
7847 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7849 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
7850 * config/nds32/constants.md (unspec_volatile_element): Add
7851 UNSPEC_VOLATILE_RELAX_GROUP.
7852 * config/nds32/nds32-relax-opt.c: New file.
7853 * config/nds32/nds32-predicates.c
7854 (nds32_symbol_load_store_p): New function.
7855 * config/nds32/nds32-protos.h
7856 (nds32_symbol_load_store_p): Declare function.
7857 (make_pass_nds32_relax_opt): Declare new rtl pass function.
7858 * config/nds32/nds32.c
7859 (nds32_register_pass): New function to register pass.
7860 (nds32_register_passes): New function to register passes.
7861 * config/nds32/nds32.md (relax_group): New pattern.
7862 * config/nds32/nds32.opt (mrelax-hint): New option.
7863 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
7865 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
7867 * config/nds32/t-nds32: Modify files dependency.
7869 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7871 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
7872 (PROFILE_HOOK): Define its implementation.
7874 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7876 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
7877 type and 32-bit size.
7879 2018-04-01 Jakub Jelinek <jakub@redhat.com>
7882 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
7883 (V_128_256): New mode iterator.
7884 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
7885 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
7886 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
7888 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
7889 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
7891 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
7894 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
7895 NaN inputs correctly.
7897 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
7900 * config/rs6000/vsx.md (??r): New mode attribute.
7901 (*vsx_mov<mode>_64bit): Use it.
7902 (*vsx_mov<mode>_32bit): Likewise.
7904 2018-03-30 Martin Sebor <msebor@redhat.com>
7906 PR tree-optimization/84818
7907 * builtins.c (check_access): Use warning_n.
7909 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7912 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
7914 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
7917 2018-03-30 Julia Koval <julia.koval@intel.com>
7920 * x86-tune.def (movx, partial_reg_dependency): Enable for
7923 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
7926 * lra-constraints.c (process_alt_operands): Move setting
7927 this_alternative_matches below.
7929 2018-03-29 Martin Liska <mliska@suse.cz>
7932 * doc/invoke.texi: Document how LTO works with debug info.
7933 Describe auto-load support of binutils. Mention 'x86-64'
7934 as valid option value of -march option.
7936 2018-03-29 Jakub Jelinek <jakub@redhat.com>
7938 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
7941 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
7942 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
7943 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
7946 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7949 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
7950 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
7951 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
7952 for RS6000_BTM_POWERPC64.
7953 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
7954 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
7955 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
7960 2018-03-28 Carl Love <cel@us.ibm.com>
7963 2017-09-27 Carl Love <cel@us.ibm.com>
7965 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
7966 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
7967 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
7970 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7972 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
7973 instead of __vector bool.
7974 (_mm_max_pu8): Likewise.
7975 (_mm_min_pi16): Likewise.
7977 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7980 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
7981 (DIVWEUO): Likewise.
7983 (DIVDEUO): Likewise.
7984 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
7985 DIVWEUO and DIVDEUO.
7986 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
7987 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
7988 (div_extend): Likewise.
7989 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
7991 (__builtin_divweuo): Likewise.
7992 (__builtin_divdeo): Likewise.
7993 (__builtin_divdeuo): Likewise.
7995 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7998 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
7999 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
8001 PR tree-optimization/82004
8002 * gimple-match-head.c (optimize_pow_to_exp): New function.
8003 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
8004 Don't fold to exp if optimize_pow_to_exp is false.
8006 2018-03-28 Martin Liska <mliska@suse.cz>
8009 * calls.c (initialize_argument_information): Fix trailing space.
8010 * common.opt: Fix typo and provide better explanation for
8011 -fsanitize-coverage option.
8012 * config/i386/i386.opt: Fix typo.
8014 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8015 Martin Liska <mliska@suse.cz>
8018 * gimplify.c (asan_poison_variable): Don't do the check for
8019 gimplify_omp_ctxp here.
8020 (gimplify_decl_expr): Do it here.
8021 (gimplify_target_expr): Likewise.
8023 2018-03-28 Martin Liska <mliska@suse.cz>
8026 * config/i386/i386.c (ix86_function_arg_advance): Do not call
8027 chkp_type_bounds_count if MPX is not enabled.
8029 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
8031 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
8033 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
8036 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
8037 function to create the function decl for complex long double
8038 multiply and divide for -mabi=ieeelongdouble.
8039 (init_float128_ieee): Call it.
8041 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
8044 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
8045 -fcf-protection=branch -mibt.
8046 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
8048 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8051 * config/arm/arm.c (arm_valid_symbolic_address): Handle
8052 arm_word_relocations.
8054 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
8057 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
8058 extern array declarations.
8060 2018-03-27 Richard Biener <rguenther@suse.de>
8063 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
8064 explicit single_use checks.
8066 2018-03-27 Richard Biener <rguenther@suse.de>
8068 PR tree-optimization/85082
8069 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
8072 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8074 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
8075 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
8076 Turn on fasynchronous-unwind-tables and funwind-tables.
8078 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
8081 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
8082 (*bmi_blsr_<mode>_ccz): Ditto.
8084 2018-03-26 Tom de Vries <tom@codesourcery.com>
8086 PR tree-optimization/85063
8087 * omp-general.c (offloading_function_p): New function. Factor out
8089 * omp-offload.c (pass_omp_target_link::gate): ... here.
8090 * omp-general.h (offloading_function_p): Declare.
8091 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
8092 with attribute omp declare target for offloading functions.
8094 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
8096 PR tree-optimization/84005
8097 * tree-data-ref.h (get_base_for_alignment): Declare.
8098 * tree-data-ref.c (get_base_for_alignment_1): New function.
8099 (get_base_for_alignment): Likewise.
8100 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
8101 get_base_for_alignment to find a suitable base object, instead
8102 of always using drb->base_address.
8104 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8107 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
8108 known size by default.
8110 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
8113 * lra-constraints.c (process_alt_operands): Don't match BLKmode
8114 and non BLKmode operands.
8116 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8119 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
8120 Clean up attributes.
8122 2018-03-23 Richard Biener <rguenther@suse.de>
8125 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
8126 we are going to emit early debug for LTO.
8128 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8131 * function.c (match_asm_constraints_1): Don't optimize if input
8132 doesn't satisfy general_operand predicate for output's mode.
8135 * alias.c (write_dependence_p): Don't require for x_canonicalized
8136 non-VOIDmode if x has VOIDmode.
8139 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
8140 just don't try to optimize it rather than assert it never happens.
8142 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8144 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
8145 macro expansions for definition of ST_INTERNAL_<mode> and
8146 LD_INTERNAL_<mode> builtins.
8147 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
8149 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
8151 (altivec_expand_st_builtin): Likewise.
8152 (altivec_expand_builtin): Remove calls to deleted functions.
8153 (rs6000_address_for_altivec): Delete this function.
8154 * config/rs6000/vector.md: Remove expands for
8155 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
8157 2018-03-22 Sudakshina Das <sudi.das@arm.com>
8160 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
8161 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
8162 re-computing once computed.
8163 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
8164 (arm_init_machine_status): Initialize
8165 machine->static_chain_stack_bytes.
8167 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8170 * doc/extend.texi: Add four new prototypes for vec_ld.
8171 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
8172 definitions for more logical presentation.
8173 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
8174 entries for V1TI variants of __builtin_altivec_ld builtin.
8175 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
8176 handling of V1TI variant of LVX icode pattern.
8177 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
8178 (rs6000_gimple_fold_builtin): Likewise.
8179 (altivec_init_builtins): Add code to define
8180 __builtin_altivec_lvx_v1ti function.
8182 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8185 * function.c (match_asm_constraints_1): Don't do the optimization
8186 if input isn't a REG, SUBREG, MEM or constant.
8188 2018-03-22 Tom de Vries <tom@codesourcery.com>
8190 PR tree-optimization/84956
8191 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
8192 bb_has_abnormal_pred.
8194 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8197 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
8198 DECL_INITIAL (decl) to decl at the end.
8199 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
8202 2018-03-21 Joseph Myers <joseph@codesourcery.com>
8204 * doc/extend.texi (__builtin_tgmath): Document when complex
8205 integer types are treated as _Complex _Float64.
8207 2018-03-21 Tom de Vries <tom@codesourcery.com>
8209 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
8211 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8213 PR tree-optimization/84960
8214 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
8215 if it is ENTRY block, move them into single succ of ENTRY in that case.
8217 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
8219 PR tree-optimization/84811
8220 * poly-int.h (poly_span_traits): Remove the T3 parameter and
8221 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
8222 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
8223 (known_subrange_p): Update accordingly. Cast each value involved
8224 in the size comparison, rather than casting the result of the
8227 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8229 PR tree-optimization/84982
8230 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
8231 by flipping the least significant bit rather than all bits from
8232 bitpos to bitpos + bitsize - 1.
8234 2018-03-21 Nathan Sidwell <nathan@acm.org>
8236 * doc/extend.texi (Deprecated Features): Remove mention of
8237 long-deleted deprecations.
8239 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8242 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
8243 * configure: Regenerate.
8245 2018-03-21 Tom de Vries <tom@codesourcery.com>
8247 PR tree-optimization/83126
8248 * tree-parloops.c (num_phis): New function.
8249 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
8251 2018-03-21 Nathan Sidwell <nathan@acm.org>
8253 * doc/extend.texi (Deprecated Features): Update deprecated flags,
8254 mention anon-struct/union members and trailing attributes.
8256 2018-03-21 Bin Cheng <bin.cheng@arm.com>
8258 PR tree-optimization/84969
8259 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
8260 builtin memset partitions if they set different rhs values.
8262 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8264 PR rtl-optimization/84989
8265 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
8266 VEC_DUPLICATE with scalar result mode.
8268 2018-03-21 Martin Liska <mliska@suse.cz>
8271 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
8272 not intended return statement.
8274 2018-03-21 Martin Liska <mliska@suse.cz>
8277 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
8278 (chkp_find_bound_slots_1): Limit number of iterations.
8280 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
8283 * Minor grammar fixes for x86 options.
8285 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8288 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
8289 holding REG_CFA_RESTORE notes, instead turn them into a USE.
8291 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
8294 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
8295 (altivec_lvx_<mode>_1op): Likewise.
8296 (altivec_stvx_<mode>_2op): Likewise.
8297 (altivec_stvx_<mode>_1op): Likewise.
8298 (altivec_lvx_<VM2:mode>): New define_expand.
8299 (altivec_stvx_<VM2:mode>): Likewise.
8300 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
8301 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8302 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
8303 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8304 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
8305 (rs6000_gen_lvx): Likewise.
8306 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
8307 (altivec_expand_stv_builtin): Likewise.
8308 (altivec_expand_builtin): Likewise.
8309 * config/rs6000/vector.md: Likewise.
8311 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8314 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
8317 2018-03-20 Richard Biener <rguenther@suse.de>
8320 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
8321 sign-conversions as zero, fall back to standard scalar_stmt
8324 2018-03-20 Martin Liska <mliska@suse.cz>
8327 * predict.c (rebuild_frequencies): Handle case when we have
8328 PROFILE_ABSENT, but flag_guess_branch_prob is false.
8330 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8333 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
8334 flag_section_anchors.
8335 * varasm.c (use_blocks_for_decl_p): Remove hack for
8336 dw2_force_const_mem.
8339 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
8341 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
8342 be created, use lowpart_subreg of operands[0] rather than operands[0]
8344 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
8345 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
8346 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
8347 and n constraint instead of aarch64_shift_imm_di and Usd.
8348 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
8349 (*aarch64_<optab>_reg_minus<mode>3): ... this.
8351 2018-03-20 Sudakshina Das <sudi.das@arm.com>
8354 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
8355 to favor GPR over NEON registers.
8356 (<shift>di3_neon): Likewise.
8358 2018-03-20 Tom de Vries <tom@codesourcery.com>
8361 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
8362 (nvptx_process_pars): Emit bar.sync asap and alap.
8364 2018-03-20 Tom de Vries <tom@codesourcery.com>
8367 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
8368 seen_label if seen_label is already set.
8370 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8373 * config/i386/i386.c (fold_builtin_cpu): For features above 31
8374 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
8375 Use 1U instead of 1. Formatting fixes.
8378 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
8379 instead of TREE_TYPE (s1) for the return value.
8381 2018-03-19 Jakub Jelinek <jakub@redhat.com>
8383 PR tree-optimization/84946
8384 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
8385 bitsize + bitsize in poly_uint64 rather than poly_int64.
8388 * dwarf2asm.c: Include fold-const.c.
8389 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
8390 of decl rather than decl itself.
8392 PR rtl-optimization/84643
8393 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
8395 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
8398 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
8399 calling assemble_variable.
8401 2018-03-19 Sudakshina Das <sudi.das@arm.com>
8404 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
8405 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
8407 2018-03-19 Jim Wilson <jimw@sifive.com>
8410 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
8411 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
8412 (riscv_first_stack_step): Likewise.
8413 (riscv_option_override): Use STACK_BOUNDARY instead of
8415 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
8417 (BIGGEST_ALIGNMENT): Set to 128.
8418 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
8419 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
8422 2018-03-19 Richard Biener <rguenther@suse.de>
8424 PR tree-optimization/84933
8425 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
8426 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
8428 2018-03-19 Richard Biener <rguenther@suse.de>
8430 PR tree-optimization/84859
8431 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
8432 (cond_if_else_store_replacement): Perform sinking operation on
8433 single-store BBs regardless of MAX_STORES_TO_SINK setting.
8434 Generalize what a BB with a single eligible store is.
8436 2018-03-19 Richard Biener <rguenther@suse.de>
8438 PR tree-optimization/84929
8439 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
8440 chrec_is_positive against non-chrec arg.
8442 2018-03-19 Tamar Christina <tamar.christina@arm.com>
8445 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
8447 2018-03-18 Martin Liska <mliska@suse.cz>
8449 PR rtl-optimization/84635
8450 * regrename.c (build_def_use): Use matches_mode only when
8453 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
8455 PR tree-optimization/84913
8456 * tree-vect-loop.c (vectorizable_reduction): Don't try to
8457 vectorize chains of COND_EXPRs.
8459 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8461 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
8463 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8465 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
8467 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8469 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
8471 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8472 Kito Cheng <kito.cheng@gmail.com>
8474 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
8475 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
8476 (nds32_adjust_reg_alloc_order): New function.
8477 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
8479 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
8481 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
8482 nds32_print_operand, nds32_print_operand_address): Use
8483 HOST_WIDE_INT_PRINT_DEC instead.
8485 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8487 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
8489 2018-03-17 Jakub Jelinek <jakub@redhat.com>
8492 * config/i386/i386.c (initial_ix86_tune_features,
8493 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
8495 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
8496 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
8497 rather than 1u << ix86_tune. Formatting fix.
8498 (ix86_option_override_internal): Change ix86_arch_mask from
8499 unsigned int to unsigned HOST_WIDE_INT, initialize to
8500 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
8501 (ix86_function_specific_restore): Likewise.
8503 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8506 * postreload.c (reload_combine_recognize_pattern): Perform
8507 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
8508 truncate_int_for_mode the result for the destination's mode.
8511 * hsa-gen.c (mem_type_for_type): Fix comment typo.
8512 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
8514 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8517 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
8520 * lra-assigns.c (lra_split_hard_reg_for): Don't use
8521 regno_allocno_class_array and sorted_pseudos.
8522 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
8523 insns where regno is used.
8525 2018-03-16 Martin Liska <mliska@suse.cz>
8528 * multiple_target.c (create_dispatcher_calls): Redirect
8529 reference in the symbol table.
8531 2018-03-16 Martin Liska <mliska@suse.cz>
8534 * multiple_target.c (create_dispatcher_calls): Redirect also
8537 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8541 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
8542 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
8543 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
8545 2018-03-16 Julia Koval <julia.koval@intel.com>
8547 * doc/invoke.texi (Skylake Server): Add CLWB.
8548 Cannonlake): Remove CLWB.
8550 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8552 PR tree-optimization/84841
8553 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
8555 (FLOAT_ONE_CONST_TYPE): Define.
8556 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
8557 (sort_by_operand_rank): Put entries with higher constant_type last
8558 rather than first to match comments.
8560 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
8562 * config/nios2/nios2.md (movsi_internal): Fix thinko in
8565 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8568 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
8569 check and use address of target always.
8571 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
8574 * config/i386/i386.c (indirect_thunk_needed): Update comments.
8575 (indirect_thunk_bnd_needed): Likewise.
8576 (indirect_thunks_used): Likewise.
8577 (indirect_thunks_bnd_used): Likewise.
8578 (indirect_return_needed): New.
8579 (indirect_return_bnd_needed): Likewise.
8580 (output_indirect_thunk_function): Add a bool argument for
8582 (output_indirect_thunk_function): Don't generate alias for
8583 function return thunk.
8584 (ix86_code_end): Call output_indirect_thunk_function to generate
8585 function return thunks.
8586 (ix86_output_function_return): Set indirect_return_bnd_needed
8587 and indirect_return_needed instead of indirect_thunk_bnd_needed
8588 and indirect_thunk_needed.
8590 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
8592 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
8593 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
8594 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
8596 2018-03-15 David Malcolm <dmalcolm@redhat.com>
8597 Paul Hua <paul.hua.gm@gmail.com>
8600 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
8602 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
8604 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
8605 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
8608 2018-03-15 Tamar Christina <tamar.christina@arm.com>
8611 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8612 instead of GET_MODE_SIZE when comparing Units.
8614 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
8617 * varasm.c (hash_section): Return an unchangeble hash value
8618 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
8619 Return !aarch64_can_use_per_function_literal_pools_p ().
8621 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8624 * optabs.c (emit_conditional_move): Pass address of cmode's copy
8625 rather than address of cmode as last argument to prepare_cmp_insn.
8627 2018-03-15 Julia Koval <julia.koval@intel.com>
8629 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
8630 F_AVX512VNNI, F_AVX512BITALG): New.
8632 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
8635 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
8636 insn for floating-point loads and stores.
8638 2018-03-14 Carl Love <cel@us.ibm.com>
8640 * config/rs6000/rs6000-c.c: Add macro definitions for
8641 ALTIVEC_BUILTIN_VEC_PERMXOR.
8642 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
8643 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
8644 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
8645 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
8647 * config/doc/extend.texi: Add prototypes for vec_permxor.
8649 2018-03-14 David Malcolm <dmalcolm@redhat.com>
8652 * diagnostic-show-locus.c (class layout_point): Convert m_line
8653 from int to linenum_type.
8654 (line_span::comparator): Use linenum "compare" function when
8655 comparing line numbers.
8656 (test_line_span): New function.
8657 (layout_range::contains_point): Convert param "row" from int to
8659 (layout_range::intersects_line_p): Likewise.
8660 (layout::will_show_line_p): Likewise.
8661 (layout::print_source_line): Likewise.
8662 (layout::should_print_annotation_line_p): Likewise.
8663 (layout::print_annotation_line): Likewise.
8664 (layout::print_leading_fixits): Likewise.
8665 (layout::annotation_line_showed_range_p): Likewise.
8666 (struct line_corrections): Likewise for field m_row.
8667 (line_corrections::line_corrections): Likewise for param "row".
8668 (layout::print_trailing_fixits): Likewise.
8669 (layout::get_state_at_point): Likewise.
8670 (layout::get_x_bound_for_row): Likewise.
8671 (layout::print_line): Likewise.
8672 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
8673 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
8674 * input.c (selftest::test_linenum_comparisons): New function.
8675 (selftest::input_c_tests): Call it.
8676 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
8677 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
8678 * selftest.h (ASSERT_GT): New macro.
8679 (ASSERT_GT_AT): New macro.
8680 (ASSERT_LT): New macro.
8681 (ASSERT_LT_AT): New macro.
8683 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
8685 PR rtl-optimization/84780
8686 * combine.c (distribute_links): Don't make a link based on pc_rtx.
8688 2018-03-14 Martin Liska <mliska@suse.cz>
8690 * tree.c (record_node_allocation_statistics): Use
8691 get_stats_node_kind.
8692 (get_stats_node_kind): New function extracted from
8693 record_node_allocation_statistics.
8694 (free_node): Use get_stats_node_kind.
8696 2018-03-14 Richard Biener <rguenther@suse.de>
8698 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
8699 that the value-set of ANTIC_IN doesn't grow.
8702 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8704 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8705 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8707 2018-03-14 Julia Koval <julia.koval@intel.com>
8709 * config.gcc (icelake-client, icelake-server): New.
8711 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
8712 (initial_ix86_arch_features): Ditto.
8713 (PTA_SKYLAKE): Add SGX.
8714 (PTA_ICELAKE): Remove.
8715 (PTA_ICELAKE_CLIENT): New.
8716 (PTA_ICELAKE_SERVER): New.
8717 (ix86_option_override_internal): Split up icelake on icelake client and
8719 (get_builtin_code_for_version): Ditto.
8720 (fold_builtin_cpu): Ditto.
8721 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
8722 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
8723 * config/i386/i386.h (processor_type): Ditto.
8724 * doc/invoke.texi: Ditto.
8726 2018-03-14 Jakub Jelinek <jakub@redhat.com>
8729 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
8730 INTEGER_CST offset, add it together with bitpos / 8 and
8731 sign extend based on POINTER_SIZE.
8735 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
8738 * config/i386/constraints.md (Yc): New register constraint.
8739 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
8740 Use Yc constraint for alternative 2 of operand 0. Remove
8741 preferred_for_speed attribute.
8743 2018-03-14 Richard Biener <rguenther@suse.de>
8745 PR tree-optimization/84830
8746 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
8747 with the old one to avoid oscillations.
8749 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
8752 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
8754 (assign_by_spills): Return a flag of reload assignment failure.
8755 Do not process the reload assignment failures. Do not spill other
8756 reload pseudos if they has the same reg class. Update n if
8758 (lra_assign): Add a return arg. Set up from the result of
8759 assign_by_spills call.
8760 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8761 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8762 usage_insns if it is not NULL.
8763 (spill_hard_reg_in_range): New function.
8764 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8765 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8766 function prototypes.
8767 (lra_assign): Change prototype.
8768 * lra.c (lra): Add code to deal with fails by splitting hard reg
8771 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
8773 * config/riscv/riscv.opt (mrelax): New option.
8774 * config/riscv/riscv.c (riscv_file_start): Emit ".option
8775 "norelax" when riscv_mrelax is disabled.
8776 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
8778 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8781 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
8782 reassociation for int modes.
8784 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8786 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
8787 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
8789 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
8790 * config/aarch64/aarch64-sve.md
8791 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
8792 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
8793 (*extend<mode><Vwide>2): Rename to...
8794 (aarch64_sve_extend<mode><Vwide>2): ...this.
8795 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
8796 renaming the old pattern to...
8797 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
8799 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
8800 define_expand, renaming the old pattern to...
8801 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
8802 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
8803 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
8804 account when deciding which SVE instruction the optab should use.
8805 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
8807 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8809 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
8810 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
8811 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
8812 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
8813 (tlsdesc_small_<mode>): Turn a define_expand and use
8814 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
8815 (tlsdesc_small_advsimd_<mode>): ...this.
8816 (tlsdesc_small_sve_<mode>): New pattern.
8818 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8820 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
8821 (UNSPEC_UMUL_HIGHPART): New constants.
8822 (MUL_HIGHPART): New int iteraor.
8823 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
8824 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
8826 (*<su>mul<mode>3_highpart): New define_insn.
8828 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8831 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8834 2018-03-13 Martin Liska <mliska@suse.cz>
8837 * (sem_item_optimizer::sem_item_optimizer): Initialize new
8839 (sem_item_optimizer::~sem_item_optimizer): Release it.
8840 (sem_item_optimizer::merge_classes): Register variable aliases.
8841 (sem_item_optimizer::fixup_pt_set): New function.
8842 (sem_item_optimizer::fixup_points_to_sets): Likewise.
8843 * ipa-icf.h: Declare new variables and functions.
8845 2018-03-13 Jakub Jelinek <jakub@redhat.com>
8848 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
8849 integer_pow2p@2 and test integer_pow2p in condition.
8850 (A < 0 ? C : 0): Similarly for @1.
8853 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
8854 characters starting at p contain '\0' character, don't look beyond
8858 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
8859 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
8862 * reg-stack.c (change_stack): Change update_end var from int to
8863 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
8864 also call set_block_for_insn on the newly added insns and rescan.
8867 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
8868 on the last operand.
8871 * tree.c (stabilize_reference_1): Return save_expr (e) for
8872 STATEMENT_LIST even if it doesn't have side-effects.
8874 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
8876 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
8878 2018-03-12 Renlin Li <renlin.li@arm.com>
8880 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
8881 aarch64_output_scalar_simd_mov_immediate.
8883 2018-03-12 Martin Sebor <msebor@redhat.com>
8885 PR tree-optimization/83456
8886 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
8887 for perfectly overlapping calls to memcpy.
8888 (gimple_fold_builtin_memory_chk): Same.
8889 (gimple_fold_builtin_strcpy): Handle no-warning.
8890 (gimple_fold_builtin_stxcpy_chk): Same.
8891 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
8893 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8895 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
8896 parameter. Use it for SFmode.
8897 (rs6000_function_arg_advance_1): Adjust.
8898 (rs6000_function_arg): Adjust.
8899 (rs6000_gimplify_va_arg): Pass false for that new parameter.
8901 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8903 PR rtl-optimization/84169
8904 PR rtl-optimization/84780
8905 * combine.c (can_combine_p): Check for a 2-insn combination whether
8906 the destination register is used between the two insns, too.
8908 2018-03-12 Richard Biener <rguenther@suse.de>
8910 PR tree-optimization/84803
8911 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
8912 for refs DR analysis didn't process.
8914 2018-03-12 Richard Biener <rguenther@suse.de>
8916 PR tree-optimization/84777
8917 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
8918 force-vectorize loops ignore whether we are optimizing for size.
8920 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
8922 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
8923 (TARGET_MD_ASM_ADJUST): Define.
8925 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
8926 Kito Cheng <kito.cheng@gmail.com>
8927 Chung-Ju Wu <jasonwucj@gmail.com>
8929 * config/nds32/nds32.c (nds32_compute_stack_frame,
8930 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
8931 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
8932 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
8933 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
8934 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
8935 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
8936 * config/nds32/nds32.md (prologue, epilogue): Use macro
8937 NDS32_V3PUSH_AVAILABLE_P to do checking.
8939 2018-03-11 Jakub Jelinek <jakub@redhat.com>
8942 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
8943 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
8944 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
8945 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
8946 addition of most attributes on !orig_type_die or the attribute not
8947 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
8949 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8950 Chung-Ju Wu <jasonwucj@gmail.com>
8952 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
8954 * config/nds32/nds32.opt (mvh): New option.
8956 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8957 Chung-Ju Wu <jasonwucj@gmail.com>
8959 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
8961 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
8962 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
8965 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8966 Chung-Ju Wu <jasonwucj@gmail.com>
8968 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
8970 * config/nds32/nds32-multiple.md (strlensi): New pattern.
8971 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
8973 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
8974 Kito Cheng <kito.cheng@gmail.com>
8975 Chung-Ju Wu <jasonwucj@gmail.com>
8977 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
8978 UNSPEC_FFMISM and UNSPEC_FLMISM.
8979 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
8980 for ffb, ffmism and flmism.
8981 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
8982 (unspec_ffmism): Ditto.
8983 (unspec_flmism): Ditto.
8984 (nds32_expand_builtin_impl): Check if string extension is available.
8985 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
8986 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
8988 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
8991 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8994 * lra-assigns.c (assign_by_spills): Return a flag of reload
8995 assignment failure. Do not process the reload assignment
8996 failures. Do not spill other reload pseudos if they has the same
8998 (lra_assign): Add a return arg. Set up from the result of
8999 assign_by_spills call.
9000 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9001 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9002 usage_insns if it is not NULL.
9003 (spill_hard_reg_in_range): New function.
9004 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9005 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9006 function prototypes.
9007 (lra_assign): Change prototype.
9008 * lra.c (lra): Add code to deal with fails by splitting hard reg
9011 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
9014 * config/i386/i386.opt: Replace Enforcment with Enforcement.
9016 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
9019 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
9020 (dw_val_node): Add val_symbolic_view.
9021 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
9022 (symview_upper_bound): New.
9023 (new_line_info_table): Initialize symviews_since_reset.
9024 (dwarf2out_source_line): Count symviews_since_reset and set
9025 symview_upper_bound.
9026 (dw_val_equal_p): Handle symview.
9027 (add_AT_symview): New.
9028 (print_dw_val): Handle symview.
9029 (attr_checksum, attr_checksum_ordered): Likewise.
9030 (same_dw_val_p, size_of_die): Likewise.
9031 (value_format, output_die): Likewise.
9032 (add_high_low_attributes): Use add_AT_symview for entry_view.
9033 (dwarf2out_finish): Reset symview_upper_bound, clear
9036 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
9039 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
9040 Add strict argument and use it.
9041 (rs6000_split_multireg_move): Update for new strict argument.
9042 (mem_operand_gpr): Disallow all non-offsettable addresses.
9043 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
9045 2018-03-09 Jakub Jelinek <jakub@redhat.com>
9048 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
9049 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
9050 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
9053 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
9054 decl, use remap_type if we want to use the type.
9056 2018-03-09 Martin Sebor <msebor@redhat.com>
9058 PR tree-optimization/84526
9059 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9061 (builtin_access::generic_overlap): Be prepared to handle non-array
9064 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
9066 PR rtl-optimization/84682
9067 * lra-constraints.c (process_address_1): Check is_address flag
9068 for address constraints.
9069 (process_alt_operands): Likewise.
9070 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
9071 preprocess_constraints.
9072 * recog.h (preprocess_constraints): Add oploc parameter.
9074 * recog.c (preprocess_constraints): Test address_operand for
9075 CT_ADDRESS constraints.
9077 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9080 * lra-assigns.c (assign_by_spills): Return a flag of reload
9081 assignment failure. Do not process the reload assignment
9082 failures. Do not spill other reload pseudos if they has the same
9084 (lra_assign): Add a return arg. Set up from the result of
9085 assign_by_spills call.
9086 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9087 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9088 usage_insns if it is not NULL.
9089 (spill_hard_reg_in_range): New function.
9090 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9091 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9092 function prototypes.
9093 (lra_assign): Change prototype.
9094 * lra.c (lra): Add code to deal with fails by splitting hard reg
9097 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9100 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
9101 Accept complain bool parameter. Only emit errors if it is true.
9102 (arm_parse_cpu_option_name): Likewise.
9103 (arm_target_thumb_only): Adjust callers of the above.
9104 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
9105 prototype to take a default true bool parameter.
9106 (arm_parse_arch_option_name): Likewise.
9108 2018-03-09 David Malcolm <dmalcolm@redhat.com>
9109 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
9113 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
9114 * configure: Regenerate.
9115 * configure.ac ("linker --version-script option"): New.
9116 ("linker soname option"): New.
9118 2018-03-09 Richard Biener <rguenther@suse.de>
9120 PR tree-optimization/84775
9121 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
9122 immediate uses of predicate stmts and mark them modified.
9125 PR tree-optimization/84178
9126 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9128 (version_loop_for_if_conversion): Delay update_ssa call.
9129 (tree_if_conversion): Delay update_ssa until after predicate
9132 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
9135 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
9136 when the function accesses prior frames.
9138 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9141 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
9142 gen_llsym, otherwise call maybe_gen_llsym.
9145 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
9146 has ',' character inside of it.
9148 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9151 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
9152 as clobbering CC_REGNUM.
9154 2018-03-08 Richard Biener <rguenther@suse.de>
9157 * tree-scalar-evolution.c: Include tree-into-ssa.h.
9158 (follow_copies_to_constant): Do not follow SSA names registered
9161 2018-03-08 Richard Biener <rguenther@suse.de>
9163 PR tree-optimization/84178
9164 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9166 (version_loop_for_if_conversion): Delay update_ssa call.
9167 (tree_if_conversion): Delay update_ssa until after predicate
9170 2018-03-08 David Malcolm <dmalcolm@redhat.com>
9172 PR tree-optimization/84178
9173 * tree-if-conv.c (release_bb_predicate): Remove the
9174 the assertion that the stmts have NULL use_ops.
9175 Discard the statements, asserting that they haven't
9176 yet been added to a BB.
9178 2018-03-08 Richard Biener <rguenther@suse.de>
9180 PR tree-optimization/84746
9181 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
9182 (phi_translate): Pass in destination ANTIC_OUT set.
9183 (phi_translate_1): Likewise. For a simplified result lookup
9184 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
9185 (phi_translate_set): Adjust.
9186 (do_pre_regular_insertion): Likewise.
9187 (do_pre_partial_partial_insertion): Likewise.
9189 2018-03-08 Martin Liska <mliska@suse.cz>
9191 PR gcov-profile/84735
9192 * doc/gcov.texi: Document usage of profile files.
9193 * gcov-io.h: Document changes in the format.
9195 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
9199 * dwarf2out.c (struct dw_line_info_table): Update comments for
9201 (FORCE_RESET_NEXT_VIEW): New.
9202 (FORCE_RESETTING_VIEW_P): New.
9203 (RESETTING_VIEW_P): Check for -1 too.
9204 (ZERO_VIEW_P): Likewise.
9205 (new_line_info_table): Force-reset next view.
9206 (dwarf2out_begin_function): Likewise.
9207 (dwarf2out_source_line): Simplify zero_view_p initialization.
9208 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
9209 view directly. Omit view when omitting .loc at line 0.
9211 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9213 PR tree-optimization/84740
9214 * tree-switch-conversion.c (process_switch): Call build_constructors
9215 only if info.phi_count is non-zero.
9217 PR tree-optimization/84739
9218 * tree-tailcall.c (find_tail_calls): Check call arguments against
9219 DECL_ARGUMENTS (current_function_decl) rather than
9220 DECL_ARGUMENTS (func) when checking for tail recursion.
9222 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9224 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
9225 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
9226 Volker Reichelt's entry and add entries for people that perform
9227 GCC fuzzy testing and report numerous bugs.
9229 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
9232 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
9233 readonly data in sdata, if that is disabled.
9234 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
9235 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
9236 -mreadonly-in-sdata option.
9238 2018-03-07 Martin Sebor <msebor@redhat.com>
9240 PR tree-optimization/84468
9241 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
9242 basic block when looking for nul assignment.
9244 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
9247 * except.h (output_function_exception_table): Adjust prototype.
9248 * except.c (output_function_exception_table): Remove FNNAME parameter
9249 and add SECTION parameter. Ouput one part of the table at a time.
9250 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
9251 the first part of the exception table and emit unwind directives.
9252 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
9253 (i386_pe_seh_cold_init): Likewise.
9254 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
9255 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
9256 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
9257 (ix86_output_call_insn): Emit a nop in one more case for SEH.
9258 * config/i386/winnt.c: Include except.h.
9259 (struct seh_frame_state): Add reg_offset, after_prologue and
9260 in_cold_section fields.
9261 (i386_pe_seh_end_prologue): Set seh->after_prologue.
9262 (i386_pe_seh_cold_init): New function.
9263 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
9264 to seh->in_cold_section.
9265 (seh_emit_push): Record the offset of the push.
9266 (seh_emit_save): Record the offet of the save.
9267 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
9268 Test seh->after_prologue to disregard the epilogue.
9269 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
9270 (i386_pe_end_cold_function): New function.
9272 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9275 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
9276 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
9279 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
9281 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
9282 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
9285 * multiple_target.c: Include tree-inline.h and intl.h.
9286 (expand_target_clones): Diagnose and fail if node->definition and
9287 !tree_versionable_function_p (node->decl).
9289 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
9291 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
9293 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
9294 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
9295 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
9297 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9300 * combine.c (try_combine): Use reg_or_subregno instead of handling
9301 just paradoxical SUBREGs and REGs.
9303 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
9305 * config/arc/arc.c (arc_finalize_pic): Remove function.
9306 (arc_must_save_register): We use single base PIC register, remove
9307 checks to save/restore the PIC register.
9308 (arc_expand_prologue): Likewise.
9309 * config/arc/arc-protos.h (arc_set_default_type_attributes):
9311 (arc_verify_short): Likewise.
9312 (arc_attr_type): Likewise.
9313 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
9314 (walk_stores): Likewise.
9315 (arc_address_cost): Make it static.
9316 (arc_verify_short): Likewise.
9317 (branch_dest): Likewise.
9318 (arc_attr_type): Likewise.
9319 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
9320 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
9321 (arc_final_prescan_insn): Remove inserting the nops due to
9322 hardware hazards. It is done in reorg step.
9323 (insn_length_variant_t): Remove.
9324 (insn_length_parameters_t): Likewise.
9325 (arc_insn_length_parameters): Likewise.
9326 (arc_get_insn_variants): Likewise.
9327 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
9329 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9332 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
9335 PR tree-optimization/84687
9336 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
9338 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
9340 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9342 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
9343 Rename to ppc_speculation_barrier.
9344 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
9345 __builtin_ppc_speculation_barrier.
9347 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9350 * combine.c (combine_simplify_rtx): Don't try to simplify if
9351 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
9354 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
9356 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
9357 to 32 bytes when compiling for POWER9.
9359 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9362 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
9363 regparm >= 3 with no arg reg available also for calls with
9364 flag_force_indirect_call. Pass decl to ix86_function_regparm.
9367 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
9369 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
9371 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
9374 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
9376 2018-03-05 Richard Biener <rguenther@suse.de>
9378 PR tree-optimization/84486
9379 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
9380 When inserting a __builtin_assume_aligned call set the LHS
9381 SSA name alignment info accordingly.
9383 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
9385 PR tree-optimization/84114
9386 * config/aarch64/aarch64.c (aarch64_reassociation_width)
9387 Avoid reassociation of FLOAT_MODE addition.
9389 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
9391 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
9392 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
9393 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
9394 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
9395 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
9396 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
9397 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
9399 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
9400 __builtin_ia32_wbinvd): New builtins.
9401 (SPECIAL_ARGS2): New.
9402 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
9403 (SPECIAL_ARGS2): New.
9404 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
9405 (ix86_valid_target_attribute_inner_p): Ditto.
9406 (ix86_init_mmx_sse_builtins): Add special_args2.
9407 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
9408 TARGET_WBNOINVD_P): New.
9409 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
9410 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
9411 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
9412 * config/i386/immintrin.h (_wbinvd): New intrinsic.
9413 * config/i386/pconfigintrin.h: New file.
9414 * config/i386/wbnoinvdintrin.h: Ditto.
9415 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
9417 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
9419 2018-03-05 Richard Biener <rguenther@suse.de>
9421 PR tree-optimization/84670
9422 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9424 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9425 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9426 (compute_antic_aux): Only assert the number of values in ANTIC_IN
9427 doesn't grow if all successors (recursively) were visited at least
9430 2018-03-05 Richard Biener <rguenther@suse.de>
9432 PR tree-optimization/84650
9433 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
9434 if executed in the loop pipeline.
9436 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
9438 * doc/configfiles.texi (Configuration Files): Move info about
9439 conditionalizing $target-protos.h to...
9440 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
9441 differs from $target-protos.h.
9443 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
9444 Chung-Ju Wu <jasonwucj@gmail.com>
9446 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
9447 * config/nds32/nds32-multiple.md (setmemsi): Define.
9448 * config/nds32/nds32-memory-manipulation.c
9449 (nds32_gen_dup_4_byte_to_word_value): New.
9450 (emit_setmem_word_loop): New.
9451 (emit_setmem_byte_loop): New.
9452 (nds32_expand_setmem_loop): New.
9453 (nds32_expand_setmem_loop_v3m): New.
9454 (nds32_expand_setmem_unroll): New.
9455 (nds32_expand_setmem): New.
9457 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9458 Chung-Ju Wu <jasonwucj@gmail.com>
9460 * config/nds32/nds32-memory-manipulation.c
9461 (nds32_emit_load_store): New.
9462 (nds32_emit_post_inc_load_store): New.
9463 (nds32_emit_mem_move): New.
9464 (nds32_emit_mem_move_block): New.
9465 (nds32_expand_movmemsi_loop_unknown_size): New.
9466 (nds32_expand_movmemsi_loop_known_size): New.
9467 (nds32_expand_movmemsi_loop): New.
9468 (nds32_expand_movmemsi_unroll): New.
9469 (nds32_expand_movmemqi): Rename ...
9470 (nds32_expand_movmemsi): ... to this.
9471 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
9472 (movmemsi): ... to this.
9473 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
9474 (nds32_expand_movmemsi): ... to this.
9476 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9477 Monk Chiang <sh.chiang04@gmail.com>
9478 Chung-Ju Wu <jasonwucj@gmail.com>
9480 * config/nds32/nds32-protos.h
9481 (nds32_expand_load_multiple): New arguments.
9482 (nds32_expand_store_multiple): Ditto.
9483 (nds32_valid_multiple_load_store): Rename ...
9484 (nds32_valid_multiple_load_store_p): ... to this.
9485 * config/nds32/nds32-memory-manipulation.c
9486 (nds32_expand_load_multiple): Refine implementation.
9487 (nds32_expand_store_multiple): Ditto.
9488 * config/nds32/nds32-multiple.md
9489 (load_multiple): Update nds32_expand_load_multiple interface.
9490 (store_multiple): Update nds32_expand_store_multiple interface.
9491 * config/nds32/nds32-predicates.c
9492 (nds32_valid_multiple_load_store): Rename ...
9493 (nds32_valid_multiple_load_store_p): ... to this and refine
9495 * config/nds32/predicates.md
9496 (nds32_load_multiple_and_update_address_operation): New predicate.
9497 (nds32_store_multiple_and_update_address_operation): New predicate.
9499 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9500 Chung-Ju Wu <jasonwucj@gmail.com>
9502 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
9503 (combo): New attribute.
9504 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
9506 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
9508 * config/nds32/nds32.opt: Change -mcmodel= default value.
9510 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
9511 Monk Chiang <sh.chiang04@gmail.com>
9512 Chung-Ju Wu <jasonwucj@gmail.com>
9514 * config/nds32/constants.md (unspec_element): New enum.
9515 * config/nds32/constraints.md (Umw): New constraint.
9516 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
9517 * config/nds32/nds32-intrinsic.md: Likewise.
9518 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
9519 (nds32_valid_smw_lwm_base_p): New.
9520 (nds32_output_smw_single_word): New.
9521 (nds32_output_lmw_single_word): New.
9522 (nds32_expand_unaligned_load): New.
9523 (nds32_expand_unaligned_store): New.
9524 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
9525 (nds32_output_smw_single_word): Declare.
9526 (nds32_output_lmw_single_word): Declare.
9527 (nds32_expand_unaligned_load): Declare.
9528 (nds32_expand_unaligned_store): Declare.
9529 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
9530 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
9531 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
9532 NDS32_BUILTIN_UASTORE_DW.
9533 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
9536 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
9537 Kito Cheng <kito.cheng@gmail.com>
9538 Chung-Ju Wu <jasonwucj@gmail.com>
9540 * config/nds32/nds32-intrinsic.c
9541 (nds32_expand_builtin_null_ftype_reg): Delete.
9542 (nds32_expand_builtin_reg_ftype_imm): Ditto.
9543 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
9544 (nds32_read_argument): New.
9545 (nds32_legitimize_target): Ditto.
9546 (nds32_legitimize_argument): Ditto.
9547 (nds32_check_constant_argument): Ditto.
9548 (nds32_expand_unop_builtin): Ditto.
9549 (nds32_expand_unopimm_builtin): Ditto.
9550 (nds32_expand_binop_builtin): Ditto.
9551 (nds32_builtin_decl_impl): Ditto.
9552 (builtin_description): Ditto.
9553 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
9554 (nds32_init_builtins_impl): Ditto.
9555 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
9556 (nds32_builtin_decl): New.
9557 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
9558 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
9560 2018-03-02 Jeff Law <law@redhat.com>
9562 * reorg.c (stop_search_p): Handle DEBUG_INSN.
9563 (redundant_insn, fill_simple_delay_slots): Likewise.
9564 (fill_slots_from_thread): Likewise.
9565 * resource.c (mark_referenced_resources): Likewise.
9566 (mark_set_resources, find_dead_or_set_registers): Likewise.
9568 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9570 * substring-locations.h (format_warning_va): Formatting fix for
9572 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
9574 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
9575 * substring-locations.c: Include intl.h.
9576 (format_warning_va): Turned into small wrapper around
9577 format_warning_n_va, renamed to ...
9578 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
9579 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
9581 (format_warning_at_substring_n): New function.
9582 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
9583 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
9584 format_warning_at_substring with just a shorter name instead of
9585 const function pointer.
9586 (fmtwarn_n): New function.
9587 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
9588 appropriate, get rid of all the fmtstr temporaries, move conditionals
9589 with G_() wrapped string literals directly into fmtwarn arguments,
9590 cast dir.len to (int), formatting fixes.
9592 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
9594 * doc/invoke.texi: Remove "Cilk Plus" references.
9596 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9597 Richard Biener <rguenther@suse.de>
9600 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
9601 for error or warning attributes if CALL_FROM_THUNK_P is set.
9604 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9607 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
9608 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
9611 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
9612 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
9614 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
9615 instead of -1U in last predictors element's probability member.
9617 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9620 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9621 arguments if they are comparable.
9623 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9625 PR tree-optimization/84634
9626 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
9627 masks and masked_loop_p with a single loop_masks, making sure it's
9628 null for bb vectorization.
9630 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9632 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9633 (vect_analyze_data_ref_access): Use loop->safe_len rather than
9634 loop->force_vectorize to check whether there is no alias.
9636 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9639 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
9641 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
9643 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
9644 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
9645 instead of a loop around prev_real_insn.
9646 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
9650 * config/i386/i386.c (ix86_print_operand): Use conditional
9651 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
9654 2018-03-02 Richard Biener <rguenther@suse.de>
9656 PR tree-optimization/84427
9657 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
9658 (bitmap_set_subtract_values): Rewrite to handle multiple
9661 (prune_clobbered_mems): Likewise.
9662 (phi_translate): Take edge instead of pred/phiblock.
9663 (phi_translate_1): Likewise.
9664 (phi_translate_set): Likewise. Insert all translated
9665 exprs for a value into the set, keeping possibly multiple
9666 expressions per value.
9667 (compute_antic_aux): Adjust for phi_translate changes.
9668 When intersecting union the expressions and prune those
9669 not in the final value set, keeping possibly multiple
9670 expressions per value. Do not use value-insertion
9671 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
9672 all expressions. Add verification that the value-sets
9673 only shrink during iteration.
9674 (compute_partial_antic_aux): Adjust for the phi_translate changes.
9675 (do_pre_regular_insertion): Likewise.
9676 (do_pre_partial_partial_insertion): Likewise.
9678 2018-03-02 Richard Biener <rguenther@suse.de>
9681 * config/darwin.c (saved_debug_info_level): New static global.
9682 (darwin_asm_lto_start): Disable debug info generation for LTO out.
9683 (darwin_asm_lto_end): Restore debug info generation settings.
9685 2018-03-01 Martin Liska <mliska@suse.cz>
9688 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
9691 2018-03-01 Richard Biener <rguenther@suse.de>
9694 * dwarf2out.c (gen_variable_die): Properly handle late VLA
9695 type annotation with LTO when debug was disabled at compile-time.
9697 2018-03-01 Matthew Fortune <mfortune@gmail.com>
9699 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
9701 (mips_final_postscan_insn): Likewise.
9703 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
9705 PR rtl-optimization/84528
9706 * alias.c (init_alias_target): Add commentary.
9707 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
9708 a unique base value if the frame pointer is not eliminated
9709 to the stack pointer.
9711 2018-03-01 Tom de Vries <tom@codesourcery.com>
9713 PR rtl-optimization/83327
9714 * lra-int.h (hard_regs_spilled_into): Declare.
9715 * lra.c (hard_regs_spilled_into): Define.
9716 (init_reg_info): Init hard_regs_spilled_into.
9717 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
9718 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
9719 (process_bb_lives): Handle hard_regs_spilled_into.
9720 (lra_create_live_ranges_1): Before doing liveness propagation, clear
9721 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
9723 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
9725 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
9726 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
9727 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
9728 * config/rs6000/aix72.h: New file.
9730 2018-02-28 Jakub Jelinek <jakub@redhat.com>
9732 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
9733 instead of warning_at with conditional singular and plural messages
9737 * stor-layout.c (update_alignment_for_field): For
9738 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
9739 && !DECL_PACKED (field), do the alignment update, just use
9740 only desired_align instead of MAX (type_align, desired_align)
9742 (place_field): Don't do known_align < desired_align handling
9743 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
9744 is non-NULL, instead do it after rli->prev_field handling and
9745 only if not within a bitfield word. For DECL_PACKED (field)
9746 use type_align of BITS_PER_UNIT.
9748 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
9750 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
9751 superfluous parentheses and trailing spaces.
9753 2018-02-28 Richard Biener <rguenther@suse.de>
9755 PR tree-optimization/84584
9756 * graphite-scop-detection.c (scop_detection::add_scop): Discard
9757 SCoPs with fake exit edge.
9759 2018-02-28 Martin Liska <mliska@suse.cz>
9762 * timevar.c (timer::print): Fix format to properly print 100%
9765 2018-02-28 Richard Biener <rguenther@suse.de>
9768 * genmatch.c (capture_info::walk_match): Do not mark
9769 captured expressions without operands as expr_p given
9770 they act more like predicates and should be subject to
9771 "lost tail" side-effect preserving.
9773 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
9775 PR rtl-optimization/81611
9776 * auto-inc-dec.c (attempt_change): Move dead note from
9777 mem_insn if it's the next use of regno
9778 (find_address): Take address use of reg holding
9779 non-incremented value. Add parm to limit search to the named
9781 (merge_in_block): Attempt to use a mem insn that is the next
9782 use of the original regno.
9784 2018-02-27 Martin Sebor <msebor@redhat.com>
9787 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
9788 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
9790 2018-02-27 Martin Sebor <msebor@redhat.com>
9792 PR translation/84207
9793 * diagnostic-core.h (warning_n, error_n, inform_n): Change
9794 n argument to unsigned HOST_WIDE_INT.
9795 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
9796 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
9797 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
9798 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
9800 2018-02-27 Richard Biener <rguenther@suse.de>
9802 PR tree-optimization/84512
9803 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
9804 Do not use the estimate returned from record_stmt_cost for
9805 the scalar iteration cost but sum properly using add_stmt_cost.
9807 2018-02-27 Richard Biener <rguenther@suse.de>
9809 PR tree-optimization/84466
9810 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9811 Adjust last change to less strictly validate use operands.
9813 2018-02-27 Martin Liska <mliska@suse.cz>
9815 PR gcov-profile/84548
9816 * gcov.c (process_file): Allow partial overlap and consider it
9817 also as group functions.
9818 (output_lines): Properly calculate range of lines for a group.
9820 2018-02-27 Martin Liska <mliska@suse.cz>
9822 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
9823 'ggc' suffixes. Change first column width.
9824 (timer::print): Fix formatting of the column.
9826 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
9828 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
9829 preserve inline entry blocks for the sake of debug inline
9830 entry point markers alone.
9831 (remove_unused_locals): Suggest in comments a better place to
9832 force the preservation of inline entry blocks that are
9833 otherwise unused, but do not preserve them.
9835 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9837 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
9839 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9842 * config/i386/constraints.md (Bs): Replace
9843 ix86_indirect_branch_register with
9844 TARGET_INDIRECT_BRANCH_REGISTER.
9846 * config/i386/i386.md (indirect_jump): Likewise.
9847 (tablejump): Likewise.
9848 (*sibcall_memory): Likewise.
9849 (*sibcall_value_memory): Likewise.
9850 Peepholes of indirect call and jump via memory: Likewise.
9851 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
9852 (*sibcall_value_GOT_32): Likewise.
9853 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9854 (GOT_memory_operand): Likewise.
9855 (call_insn_operand): Likewise.
9856 (sibcall_insn_operand): Likewise.
9857 (GOT32_symbol_operand): Likewise.
9858 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
9860 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9862 PR rtl-optimization/83496
9863 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
9864 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
9865 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
9866 redundant insn, if any.
9867 (relax_delay_slots): Likewise.
9868 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
9870 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
9872 PR tree-optimization/83965
9873 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
9874 that grouped statements are part of a reduction chain. Return
9875 true if the statement is not marked as a reduction itself but
9877 (vect_recog_dot_prod_pattern): Don't check whether the statement
9878 is part of a group here.
9879 (vect_recog_sad_pattern): Likewise.
9880 (vect_recog_widen_sum_pattern): Likewise.
9882 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9885 * final.c (rest_of_clean_state): Also look for calls inside sequences.
9887 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9890 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
9892 (ix86_output_indirect_function_return): New prototype.
9893 (ix86_split_simple_return_pop_internal): Likewise.
9894 * config/i386/i386.c (indirect_return_via_cx): New.
9895 (indirect_return_via_cx_bnd): Likewise.
9896 (indirect_thunk_name): Handle return va CX_REG.
9897 (output_indirect_thunk_function): Create alias for
9898 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
9899 (ix86_output_indirect_jmp): Remove the bool argument.
9900 (ix86_output_indirect_function_return): New function.
9901 (ix86_split_simple_return_pop_internal): Likewise.
9902 * config/i386/i386.md (*indirect_jump): Don't pass false
9903 to ix86_output_indirect_jmp.
9904 (*tablejump_1): Likewise.
9905 (simple_return_pop_internal): Change it to define_insn_and_split.
9906 Call ix86_split_simple_return_pop_internal to split it for
9908 (simple_return_indirect_internal): Call
9909 ix86_output_indirect_function_return instead of
9910 ix86_output_indirect_jmp.
9912 2018-02-26 Jakub Jelinek <jakub@redhat.com>
9915 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
9916 memset and value initialization afterwards.
9918 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
9920 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
9922 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9925 * common/config/aarch64/aarch64-common.c
9926 (aarch_option_optimization_table[]): Switch
9927 off fomit-frame-pointer
9929 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9930 Chung-Ju Wu <jasonwucj@gmail.com>
9932 * config/nds32/nds32-multiple.md (load_multiple): Disallow
9934 (store_multiple): Ditto.
9936 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9938 * config.gcc: Add --with-cpu support for nds32 target.
9939 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
9940 * config/nds32/nds32.opt: Add -mcpu= option.
9942 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
9944 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
9945 isel=yes): Warn for these deprecated options.
9947 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
9949 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
9950 ISA_2_5_MASKS_EMBEDDED.
9952 2018-02-23 Jakub Jelinek <jakub@redhat.com>
9954 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
9955 p->max as pointers rather than using iterative_hash_expr.
9957 2018-02-23 Carl Love <cel@us.ibm.com>
9959 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
9960 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
9962 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
9963 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
9964 P8V_BUILTIN_VEC_VUNSIGNED2.
9966 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
9969 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
9970 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
9972 * lra-constraints.c (curr_insn_transform): Set up
9973 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
9975 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
9976 * lra-eliminations.c (spill_pseudos): Ditto.
9977 (process_insn_for_elimination): Ditto.
9978 * lra-lives.c (reg_early_clobber_p): Use the new macros.
9979 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
9980 LRA_NON_CLOBBERED_ALT.
9982 2018-02-22 Martin Sebor <msebor@redhat.com>
9984 PR tree-optimization/84480
9985 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
9986 to maybe_diag_stxncpy_trunc. Call it.
9987 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
9988 from gimple_fold_builtin_strcpy. Print inlining stack.
9989 (handle_builtin_stxncpy): Print inlining stack.
9990 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
9992 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
9995 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
9996 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
9997 and -fcheck-pointer-bounds are used together.
9998 (indirect_thunk_prefix): New enum.
9999 (indirect_thunk_need_prefix): New function.
10000 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
10001 "_nt" instead of "_bnd" for NOTRACK prefix.
10002 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
10003 (output_indirect_thunk_function): Likewise.
10005 (ix86_code_end): Update output_indirect_thunk_function calls.
10006 (ix86_output_indirect_branch_via_reg): Replace
10007 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
10008 (ix86_output_indirect_branch_via_push): Likewise.
10009 (ix86_output_function_return): Likewise.
10010 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
10011 incompatible with -fcf-protection=branch and
10012 -fcheck-pointer-bounds.
10014 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10017 * config/aarch64/aarch64.c (aarch64_print_address_internal):
10018 Change gcc_assert call to output_operand_lossage.
10020 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10022 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
10024 2018-02-22 DJ Delorie <dj@redhat.com>
10025 Sebastian Perta <sebastian.perta@renesas.com>
10026 Oleg Endo <olegendo@gcc.gnu.org>
10028 * config/rx/rx.c (rx_rtx_costs): New function.
10029 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
10031 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
10033 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
10035 2018-02-22 Martin Liska <mliska@suse.cz>
10038 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
10039 Add "native" as a possible value.
10041 2018-02-22 Martin Liska <mliska@suse.cz>
10044 * config/i386/i386.c (ix86_option_override_internal):
10045 Add "native" as a possible value for -march and -mtune.
10047 2018-02-22 Jakub Jelinek <jakub@redhat.com>
10050 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
10051 to all type variants.
10053 PR tree-optimization/84503
10054 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
10055 width as info->bitpos + info->bitsize - start.
10056 (merged_store_group::merge_overlapping): Simplify width computation.
10057 (check_no_overlap): New function.
10058 (imm_store_chain_info::try_coalesce_bswap): Compute expected
10059 start + width and last_order of the group, fail if check_no_overlap
10061 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
10062 to group if check_no_overlap fails.
10064 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10066 * config/rs6000/altivec.md: Delete contraint arguments to
10067 define_expand, define_split, and define_peephole2, and in
10068 define_insn_and_split if always unused.
10069 * config/rs6000/darwin.md: Ditto.
10070 * config/rs6000/dfp.md: Ditto.
10071 * config/rs6000/rs6000.md: Ditto.
10072 * config/rs6000/sync.md: Ditto.
10073 * config/rs6000/vector.md: Ditto.
10074 * config/rs6000/vsx.md: Ditto.
10076 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10078 * config/rs6000/altivec.md: Write output control strings as braced
10079 blocks instead of double-quoted strings.
10080 * config/rs6000/darwin.md: Ditto.
10081 * config/rs6000/rs6000.md: Ditto.
10082 * config/rs6000/vector.md: Ditto.
10083 * config/rs6000/vsx.md: Ditto.
10085 2018-02-21 Jason Merrill <jason@redhat.com>
10087 PR c++/84314 - ICE with templates and fastcall attribute.
10088 * attribs.c (build_type_attribute_qual_variant): Remove assert.
10090 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10092 * ipa-cp.c (determine_versionability): Fix comment typos.
10094 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10097 * ipa-cp.c (determine_versionability): Do not version functions caling
10100 2018-02-21 Martin Liska <mliska@suse.cz>
10103 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
10104 Add "native" as a possible value.
10105 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
10106 the macro when native cpu detection is available.
10108 2018-02-21 Martin Liska <mliska@suse.cz>
10111 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
10112 Add "native" as a possible value.
10113 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
10114 when native cpu detection is available.
10116 2018-02-21 Jakub Jelinek <jakub@redhat.com>
10117 Martin Sebor <msebor@redhat.com>
10119 PR tree-optimization/84478
10120 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
10122 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
10123 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
10124 support which is conservatively correct, for 2 only stay conservative
10125 for maxlen. Formatting and comment capitalization fixes. Add STRICT
10126 argument to the 2 argument get_range_strlen, adjust 6 arg
10127 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
10129 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
10130 (gimple_fold_builtin_strlen): Pass true as last argument to
10133 2018-02-20 Martin Sebor <msebor@redhat.com>
10135 PR middle-end/84095
10136 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
10137 (builtin_memref::set_base_and_offset): Same. Handle inner references.
10138 (builtin_memref::builtin_memref): Factor out parts into
10139 set_base_and_offset and call it.
10141 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10143 PR middle-end/84406
10144 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
10145 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
10146 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
10147 search at the associated MODE_INT.
10149 2018-02-20 Jeff Law <law@redhat.com>
10151 PR middle-end/82123
10152 PR tree-optimization/81592
10153 PR middle-end/79257
10154 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
10155 for range data rather than using global data.
10156 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
10157 range data rather than using global data.
10158 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
10159 pass it to children as needed.
10160 (struct directive::fmtresult): Similarly.
10161 (struct directive::set_width): Similarly.
10162 (struct directive::set_precision): Similarly.
10163 (format_integer, format_directive, parse_directive): Similarly.
10164 (format_none): Accept unnamed vr_values parameter.
10165 (format_percent, format_floating, format_character): Similarly.
10166 (format_string, format_plain): Similarly.
10167 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
10168 the EVRP range analyzer for range data rather than using global data.
10169 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
10170 gimple-ssa-evrp-analyze.h
10171 (class sprintf_dom_walker): Add after_dom_children member function.
10172 Add evrp_range_analyzer member.
10173 (sprintf_dom_walker::before_dom_children): Call into the EVRP
10174 range analyzer as needed.
10175 (sprintf_dom_walker::after_dom_children): New member function.
10176 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
10178 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
10179 (evrp_range_analyzer::pop_to_marker): Likewise.
10181 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10183 PR tree-optimization/84419
10184 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
10185 with the required type if its current type is compatible but
10188 2018-02-20 Jakub Jelinek <jakub@redhat.com>
10190 PR middle-end/82004
10191 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
10192 after vectorization.
10194 2018-02-20 Martin Liska <mliska@suse.cz>
10197 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
10198 possible values if we don't have a hint.
10200 2018-02-20 Martin Liska <mliska@suse.cz>
10204 * final.c (shorten_branches): Build align_tab array with one
10206 * opts.c (finish_options): Add alignment option limit check.
10207 (MAX_CODE_ALIGN): Likewise.
10208 (MAX_CODE_ALIGN_VALUE): Likewise.
10209 * doc/invoke.texi: Document maximum allowed option value for
10210 all -falign-* options.
10212 2018-02-19 Jakub Jelinek <jakub@redhat.com>
10215 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
10216 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
10217 * var-tracking.c (emit_note_insn_var_location): Remove all references
10218 to NOTE_INSN_CALL_ARG_LOCATION.
10219 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
10220 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
10221 Use copy_rtx_if_shared.
10222 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
10223 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
10224 (dwarf2out_var_location): Remove handling of
10225 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
10227 * final.c (final_scan_insn): Remove all references to
10228 NOTE_INSN_CALL_ARG_LOCATION.
10229 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
10230 before dumping final insns.
10231 * except.c (emit_note_eh_region_end): Remove all references to
10232 NOTE_INSN_CALL_ARG_LOCATION.
10233 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
10234 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
10235 * config/arc/arc.c (hwloop_optimize): Likewise.
10236 * config/arm/arm.c (create_fix_barrier): Likewise.
10237 * config/s390/s390.c (s390_chunkify_start): Likewise.
10238 * config/sh/sh.c (find_barrier): Likewise.
10239 * config/i386/i386.c (rest_of_insert_endbranch,
10240 ix86_seh_fixup_eh_fallthru): Likewise.
10241 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
10242 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
10243 * config/frv/frv.c (frv_function_prologue): Likewise.
10244 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
10246 (note_outside_basic_block_p): Remove all references to
10247 NOTE_INSN_CALL_ARG_LOCATION.
10248 * gengtype.c (adjust_field_rtx_def): Likewise.
10249 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
10251 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
10252 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
10255 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
10258 PR tree-optimization/84452
10259 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
10260 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
10263 2018-02-19 Martin Liska <mliska@suse.cz>
10266 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
10268 2018-02-19 Martin Liska <mliska@suse.cz>
10269 Richard Sandiford <richard.sandiford@linaro.org>
10271 PR tree-optimization/82491
10272 * gimple-fold.c (get_base_constructor): Make earlier bail out
10275 2018-02-19 Carl Love <cel@us.ibm.com>
10277 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
10278 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
10280 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
10281 P8V_BUILTIN_VEC_NEG.
10283 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
10285 * config/rl78/rl78.md (movdf): New define expand.
10287 2018-02-19 Martin Liska <mliska@suse.cz>
10290 * doc/invoke.texi: Fix typo.
10291 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
10293 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
10295 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
10296 handle rs6000_single_float and rs6000_double_float specially for
10299 2018-02-16 Jeff Law <law@redhat.com>
10301 * config/rx/rx.c (add_pop_cfi_notes): New function.;
10302 (pop_regs): Use it.
10304 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10307 * ipa-inline.c (inline_small_functions): Fix a typo.
10309 2018-02-16 Nathan Sidwell <nathan@acm.org>
10311 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
10313 2018-02-16 Carl Love <cel@us.ibm.com>
10315 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
10316 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
10317 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
10318 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
10319 expansion to P8V_BUILTIN_VEC_FLOAT2.
10321 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
10323 PR rtl-optimization/70023
10324 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
10325 src_regno into account.
10327 2018-02-16 Carl Love <cel@us.ibm.com>
10329 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
10330 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
10331 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
10332 * config/rs6000/rs6000.c: Remove case statements for
10333 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
10334 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
10335 and P9V_BUILTIN_VEC_VINSERT4B.
10336 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
10337 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
10338 * config/rs6000/vsx.md:
10339 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
10342 2018-02-16 Carl Love <cel@us.ibm.com>
10344 * config/rs6000/altivec.h: Add builtin names vec_extract4b
10346 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
10348 * config/rs6000/rs6000-c.c: Add the definitions for
10349 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
10350 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
10351 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
10352 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
10353 definition for insert4b and define insn *insert3b_internal.
10354 * doc/extend.texi: Add documentation for vec_extract4b.
10356 2018-02-16 Nathan Sidwell <nathan@acm.org>
10358 * doc/extend.texi (Backwards Compatibility): Mention friend
10359 injection. Note for-scope is deprecated.
10360 * doc/invoke.texi (-ffriend-injection): Deprecate.
10362 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
10364 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
10365 that moved to I2, also allow destinations that are a paradoxical
10366 subreg (instead of a normal reg).
10368 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
10371 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
10374 2018-02-16 Richard Biener <rguenther@suse.de>
10376 PR tree-optimization/84037
10377 PR tree-optimization/84016
10379 * config/i386/i386.c (ix86_builtin_vectorization_cost):
10380 Adjust vec_construct for the fact we need additional higher latency
10381 128bit inserts for AVX256 and AVX512 vector builds.
10382 (ix86_add_stmt_cost): Scale vector construction cost for
10385 2018-02-16 Richard Biener <rguenther@suse.de>
10387 PR tree-optimization/84417
10388 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
10389 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
10390 (non_rewritable_lvalue_p): Likewise, use poly-ints.
10392 2018-02-16 Martin Liska <mliska@suse.cz>
10395 * internal-fn.def (ASAN_CHECK): Set proper flags.
10396 (ASAN_MARK): Likewise.
10398 2018-02-16 Julia Koval <julia.koval@intel.com>
10400 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
10401 from PTA_CANNONLAKE.
10403 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10406 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
10407 Use ++iter rather than iter++ for std::list iterators.
10408 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
10409 defer deleting them until all nodes in the forest are processed. Do
10410 free even leaf nodes. Change to_process into auto_vec.
10413 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
10414 * vec.h (vec_default_construct): Use memset instead of placement new
10415 if BROKEN_VALUE_INITIALIZATION is defined.
10416 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
10417 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
10420 PR rtl-optimization/83723
10421 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
10422 * lra.c (lra_substitute_pseudo): Likewise. If true, use
10423 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
10425 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
10427 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
10429 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
10431 PR rtl-optimization/81443
10432 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
10433 from inner REGs to paradoxical SUBREGs.
10435 2018-02-16 Richard Biener <rguenther@suse.de>
10437 PR tree-optimization/84399
10438 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10439 For operands we can analyze at their definition make sure we can
10440 analyze them at each use as well.
10442 2018-02-16 Richard Biener <rguenther@suse.de>
10444 PR tree-optimization/84190
10445 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
10446 volatile accesses if the decl isn't volatile.
10448 2018-02-15 Jason Merrill <jason@redhat.com>
10450 PR c++/84314 - ICE with templates and fastcall attribute.
10451 * attribs.c (build_type_attribute_qual_variant): Don't clobber
10452 TYPE_CANONICAL on an existing type.
10454 2018-02-15 Jakub Jelinek <jakub@redhat.com>
10456 PR tree-optimization/84383
10457 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
10458 dstoff nor call operand_equal_p if dstbase is NULL.
10460 PR tree-optimization/84334
10461 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
10462 also a CONSTANT_CLASS_P, punt.
10464 2018-02-14 Jim Wilson <jimw@sifive.com>
10466 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
10467 first SMALL_OPERAND check. New local min_second_step. Move assert
10468 to where locals are set. Add TARGET_RVC support.
10469 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
10471 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
10473 * doc/invoke.texi: Correct -Wformat-overflow code sample.
10475 2018-02-14 Martin Sebor <msebor@redhat.com>
10477 PR tree-optimization/83698
10478 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
10479 arrays constrain the offset range to their bounds.
10480 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
10481 (builtin_access::overlap): Avoid setting the size of overlap if it's
10483 (maybe_diag_overlap): Also consider arrays when deciding what values
10484 of offsets to include in diagnostics.
10486 2018-02-14 Martin Sebor <msebor@redhat.com>
10489 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
10490 that correspond to the kind of a declaration.
10492 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
10495 * config/pa/pa.md: Load address of PIC label using the linkage table
10496 if the label is nonlocal.
10498 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
10500 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
10501 warning message if user requests -maltivec=be.
10502 * doc/invoke.texi: Document deprecation of -maltivec=be.
10504 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
10507 * config/rs6000/rs6000-c.c: Update definitions for
10508 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
10509 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
10511 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10514 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
10515 add _get_ssp intrinsics. Remove argument from
10516 __builtin_ia32_rdssp[d|q].
10517 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
10518 * config/i386/i386-builtin.def: Remove argument from
10519 __builtin_ia32_rdssp[d|q].
10520 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
10521 ix86_expand_special_args_builtin for _rdssp[d|q].
10522 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
10523 Clear register before usage.
10524 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
10525 Add documentation for new _get_ssp and _inc_ssp intrinsics.
10527 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
10529 PR tree-optimization/84357
10530 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
10531 operand 1 of an ARRAY_REF too.
10533 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
10536 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
10537 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10539 (set_of_reg): New struct.
10540 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
10541 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
10542 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10544 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
10545 Split into bitclr, bitset, bitinvert patterns if appropriate.
10546 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
10547 use rx_fuse_in_memory_bitop.
10548 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
10549 to named insn, correct maximum insn length.
10551 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
10554 * machmode.def: Define a complex mode for PARTIAL_INT.
10555 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
10557 * doc/rtl.texi: Document CSPImode.
10558 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
10560 (msp430_hard_regno_nregs_with_padding): Likewise.
10562 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
10565 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
10567 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
10569 PR rtl-optimization/84169
10570 * combine.c (try_combine): New variable split_i2i3. Set it to true if
10571 we generated a parallel as new i3 and we split that to new i2 and i3
10572 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
10573 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
10574 those to i2, not i1. Partially rewrite this scan code.
10576 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10579 * stor-layout.c (place_field): For variable length fields, adjust
10580 offset_align afterwards not just based on the field's alignment,
10581 but also on the size.
10583 PR middle-end/84309
10584 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
10585 of exps and logs in the use_exp2 case.
10587 2018-02-13 Jeff Law <law@redhat.com>
10589 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
10590 entry for "vector".
10592 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
10595 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
10599 * common.opt (gas-loc-support, gas-locview-support): New.
10600 (ginline-points, ginternal-reset-location-views): New.
10601 * doc/invoke.texi: Document them. Use @itemx where intended.
10602 (gvariable-location-views): Adjust.
10603 * target.def (reset_location_view): New.
10604 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
10605 (TARGET_RESET_LOCATION_VIEW): New.
10606 * doc/tm.texi: Rebuilt.
10607 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
10608 (dwarf2out_default_as_locview_support): New.
10609 (output_asm_line_debug_info): Use option variables.
10610 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
10611 (output_loc_list): Likewise.
10612 (add_high_low_attributes): Check option variables.
10613 Don't output entry view attribute in strict mode.
10614 (gen_inlined_subroutine_die): Check option variables.
10615 (dwarf2out_inline_entry): Likewise.
10616 (init_sections_and_labels): Likewise.
10617 (dwarf2out_early_finish): Likewise.
10618 (maybe_reset_location_view): New, from...
10619 (dwarf2out_var_location): ... here. Call it.
10620 * debug.h (dwarf2out_default_as_loc_support): Declare.
10621 (dwarf2out_default_as_locview_support): Declare.
10622 * hooks.c (hook_int_rtx_insn_0): New.
10623 * hooks.h (hook_int_rtx_insn_0): Declare.
10624 * toplev.c (process_options): Take -gas-loc-support and
10625 -gas-locview-support from dwarf2out. Enable
10626 -gvariable-location-views by default only with locview
10627 assembler support. Enable -ginternal-reset-location-views by
10628 default only if the target defines the corresponding hook.
10629 Enable -ginline-points by default if location views are
10630 enabled; force it disabled if statement frontiers are
10632 * tree-inline.c (expand_call_inline): Check option variables.
10633 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
10635 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
10637 PR tree-optimization/84321
10638 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
10639 handling. Also check whether the anti-range contains any values
10640 that satisfy the mask; switch to a VR_RANGE if not.
10642 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
10645 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
10647 2018-02-13 Martin Jambor <mjambor@suse.cz>
10650 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
10651 of call statements, also set location of a load to a temporary.
10653 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
10655 * config/rl78/rl78.c (add_vector_labels): New function.
10656 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
10657 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
10658 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
10659 which checks that no arguments are passed.
10660 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
10661 * doc/extend.texi: Documentation for the new attribute.
10663 2018-02-13 Andreas Schwab <schwab@suse.de>
10665 * config/riscv/linux.h (CPP_SPEC): Define.
10667 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10670 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
10671 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
10672 OPTION_MASK_ISA_AES as first argument to def_builtin_const
10673 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
10674 instead of OPTION_MASK_ISA_PCLMUL as first argument to
10675 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
10676 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
10677 temporarily for AES and PCLMUL builtins.
10679 PR tree-optimization/84339
10680 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
10681 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
10684 PR middle-end/84309
10685 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
10686 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
10687 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
10689 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
10691 * omp-simd-clone.h: New file.
10692 * omp-simd-clone.c: Include omp-simd-clone.h.
10693 (expand_simd_clones): No longer static.
10694 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
10695 cgraph.h and omp-simd-clone.h.
10696 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
10697 (vect_recog_widen_shift_pattern): Formatting fix.
10698 (vect_pattern_recog_1): Don't check optab for calls.
10701 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
10702 operands[2] into a REG before using gen_lowpart on it.
10704 2018-02-12 Jeff Law <law@redhat.com>
10707 * config/sh/sh.c (find_barrier): Consider a sibling call
10710 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
10711 successfully back substituting a reg.
10713 2018-02-12 Richard Biener <rguenther@suse.de>
10715 PR tree-optimization/84037
10716 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
10717 parameter, move visited init to caller.
10718 (vect_slp_analyze_operations): Separate cost from validity
10719 check, initialize visited once for all instances.
10720 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
10722 * tree-vect-stmts.c (vect_model_simple_cost): Make early
10724 (vect_model_promotion_demotion_cost): Likewise.
10725 (vectorizable_bswap): Guard cost modeling with !slp_node
10726 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
10728 (vectorizable_call): Likewise.
10729 (vectorizable_conversion): Likewise.
10730 (vectorizable_assignment): Likewise.
10731 (vectorizable_shift): Likewise.
10732 (vectorizable_operation): Likewise.
10733 (vectorizable_store): Likewise.
10734 (vectorizable_load): Likewise.
10735 (vectorizable_condition): Likewise.
10736 (vectorizable_comparison): Likewise.
10738 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
10741 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
10742 (ASAN_MARK): Fix fnspec to account for return value, change pointer
10743 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
10745 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10747 PR middle-end/83665
10748 * params.def (inline-min-speedup): Increase from 8 to 15.
10749 (max-inline-insns-auto): Decrease from 40 to 30.
10750 * ipa-split.c (consider_split): Add some buffer for function to
10751 be considered inlining candidate.
10752 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
10755 2018-02-12 Richard Biener <rguenther@suse.de>
10757 PR tree-optimization/84037
10758 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
10759 matched stmts if we cannot swap the non-matched ones.
10761 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10763 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
10764 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
10765 _mm_maskz_scalef_round_ss): New intrinsics.
10766 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
10767 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
10768 __builtin_ia32_scalefss_round): Remove.
10769 (__builtin_ia32_scalefsd_mask_round,
10770 __builtin_ia32_scalefss_mask_round): New intrinsics.
10771 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
10772 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
10773 ((match_operand:VF_128 2 "<round_nimm_predicate>"
10774 "<round_constraint>")): Changed to ...
10775 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
10776 "<round_scalar_constraint>")): ... this.
10777 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
10778 %0, %1, %2<round_op3>}"): Changed to ...
10779 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
10780 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
10781 %2<round_scalar_mask_op3>}"): ... this.
10782 * config/i386/subst.md (round_scalar_nimm_predicate): New.
10784 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10786 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
10787 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
10788 (_mm_maskz_sqrt_round_ss): New intrinsics.
10789 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
10790 (__builtin_ia32_sqrtsd_mask_round)
10791 (__builtin_ia32_sqrtss_mask_round): New builtins.
10792 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
10793 (__builtin_ia32_sqrtss_round): Remove.
10794 (__builtin_ia32_sqrtsd_mask_round)
10795 (__builtin_ia32_sqrtss_mask_round): New builtins.
10796 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
10797 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
10798 ((match_operand:VF_128 1 "vector_operand"
10799 "xBm,<round_constraint>")): Changed to ...
10800 ((match_operand:VF_128 1 "vector_operand"
10801 "xBm,<round_scalar_constraint>")): ... this.
10802 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
10803 %0, %2, %<iptr>1<round_op3>}): Changed to ...
10804 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
10805 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
10806 %<iptr>1<round_scalar_mask_op3>}): ... this.
10807 ((set_attr "prefix" "<round_prefix>")): Changed to ...
10808 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
10810 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
10813 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
10814 Cast vec_cmpeq result to correct type.
10815 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
10816 Cast vec_cmpgt result to correct type.
10818 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
10820 * final.c (final_scan_insn_1): Renamed from...
10821 (final_scan_insn): ... this. New wrapper, to recover
10822 seen from the outermost call in recursive ones.
10823 * config/sparc/sparc.c (output_return): Drop seen from call.
10824 (output_sibcall): Likewise.
10825 * config/visium/visium.c (output_branch): Likewise.
10827 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
10829 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
10832 2018-02-10 Alan Modra <amodra@gmail.com>
10835 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
10836 Specify LR as an input.
10838 2018-02-10 Jakub Jelinek <jakub@redhat.com>
10841 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
10842 remove_member_access_dummy_vars): New functions.
10843 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
10844 lower_omp_1, execute_lower_omp): Use them.
10846 PR rtl-optimization/84308
10847 * shrink-wrap.c (spread_components): Release todo vector.
10849 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
10851 PR rtl-optimization/57193
10852 * ira-color.c (struct allocno_color_data): Add member
10853 conflict_allocno_hard_prefs.
10854 (update_conflict_allocno_hard_prefs): New.
10855 (bucket_allocno_compare_func): Add a preference based on
10856 conflict_allocno_hard_prefs.
10857 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
10858 (color_allocnos): Remove a dead code. Initiate
10859 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
10861 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10864 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
10865 constraint from =wa to wa. Avoid a subreg on the output operand,
10866 instead use a pseudo and subreg it in a move.
10867 (p9_xxbrd_<mode>): Changed to ...
10868 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
10869 (p9_xxbrd_v2df): New expander.
10870 (p9_xxbrw_<mode>): Changed to ...
10871 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
10872 (p9_xxbrw_v4sf): New expander.
10874 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10876 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
10878 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
10881 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
10882 multiply in 32-bit mode.
10883 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
10884 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
10887 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10889 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
10890 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
10891 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
10892 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
10894 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
10897 * dwarf2out.c (is_trivial_indirect_ref): New function.
10898 (dwarf2out_late_global_decl): Do not generate a location
10899 attribute for variables that have a non-trivial DECL_VALUE_EXPR
10900 and that are not defined in the current unit.
10902 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10904 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
10905 instead of a libcall for UNORDERED.
10907 2018-02-09 Tamar Christina <tamar.christina@arm.com>
10910 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
10911 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
10913 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10916 * config/s390/s390.c (s390_set_current_function): Invoke
10917 s390_indirect_branch_settings also if fndecl didn't change.
10919 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10921 * config/rs6000/rs6000.md (blockage): Set length to zero.
10923 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10925 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
10927 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10930 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
10931 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
10935 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
10936 PARALLEL incoming that failed vt_get_decl_and_offset check.
10938 PR middle-end/84237
10939 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
10940 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
10942 (get_variable_section): For decls in named .bss* sections pass true as
10943 second argument to bss_initializer_p.
10945 2018-02-09 Marek Polacek <polacek@redhat.com>
10946 Jakub Jelinek <jakub@redhat.com>
10949 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
10950 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
10951 Sync some changes from cxx_fold_indirect_ref.
10953 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10955 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
10957 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
10958 (BLOCK_INLINE_ENTRY_LABEL): New.
10959 (dwarf2out_var_location): Disregard inline entry markers.
10960 (inline_entry_data): New struct.
10961 (inline_entry_data_hasher): New hashtable type.
10962 (inline_entry_data_hasher::hash): New.
10963 (inline_entry_data_hasher::equal): New.
10964 (inline_entry_data_table): New variable.
10965 (add_high_low_attributes): Add DW_AT_entry_pc and
10966 DW_AT_GNU_entry_view attributes if a pending entry is found
10967 in inline_entry_data_table. Add old entry_pc attribute only
10968 if debug nonbinding markers are disabled.
10969 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
10970 markers are enabled.
10971 (block_within_block_p, dwarf2out_inline_entry): New.
10972 (dwarf2out_finish): Check that no entries remained in
10973 inline_entry_data_table.
10974 * final.c (reemit_insn_block_notes): Handle inline entry notes.
10975 (final_scan_insn, notice_source_line): Likewise.
10976 (rest_of_clean_state): Skip inline entry markers.
10977 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
10979 * gimple.c (gimple_build_debug_inline_entry): New.
10980 * gimple.h (enum gimple_debug_subcode): Add
10981 GIMPLE_DEBUG_INLINE_ENTRY.
10982 (gimple_build_debug_inline_entry): Declare.
10983 (gimple_debug_inline_entry_p): New.
10984 (gimple_debug_nonbind_marker_p): Adjust.
10985 * insn-notes.def (INLINE_ENTRY): New.
10986 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
10987 inline entry marker notes.
10988 (print_insn): Likewise.
10989 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
10990 (INSN_DEBUG_MARKER_KIND): Likewise.
10991 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
10992 * tree-inline.c (expand_call_inline): Build and insert
10993 debug_inline_entry stmt.
10994 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
10995 inline entry blocks early, if nonbind markers are enabled.
10996 (dump_scope_block): Dump fragment info.
10997 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
10998 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
10999 (gimple_build_debug_inline_entry): New.
11000 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
11001 Enable/disable inline entry points too.
11002 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
11003 (DEBUG_INSN): Describe inline entry markers.
11005 * common.opt (gvariable-location-views): New.
11006 (gvariable-location-views=incompat5): New.
11007 * config.in: Rebuilt.
11008 * configure: Rebuilt.
11009 * configure.ac: Test assembler for view support.
11010 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
11011 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
11012 * dwarf2out.c (var_loc_view): New typedef.
11013 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
11014 (dwarf2out_locviews_in_attribute): New.
11015 (dwarf2out_locviews_in_loclist): New.
11016 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
11017 (enum dw_line_info_opcode): Add LI_adv_address.
11018 (struct dw_line_info_table): Add view.
11019 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
11020 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
11021 (zero_view_p): New variable.
11022 (ZERO_VIEW_P): New macro.
11023 (output_asm_line_debug_info): New.
11024 (struct var_loc_node): Add view.
11025 (add_AT_view_list, AT_loc_list): New.
11026 (add_var_loc_to_decl): Add view param. Test it against last.
11027 (new_loc_list): Add view params. Record them.
11028 (AT_loc_list_ptr): Handle loc and view lists.
11029 (view_list_to_loc_list_val_node): New.
11030 (print_dw_val): Handle dw_val_class_view_list.
11031 (size_of_die): Likewise.
11032 (value_format): Likewise.
11033 (loc_list_has_views): New.
11034 (gen_llsym): Set vl_symbol too.
11035 (maybe_gen_llsym, skip_loc_list_entry): New.
11036 (dwarf2out_maybe_output_loclist_view_pair): New.
11037 (output_loc_list): Output view list or entries too.
11038 (output_view_list_offset): New.
11039 (output_die): Handle dw_val_class_view_list.
11040 (output_dwarf_version): New.
11041 (output_compilation_unit_header): Use it.
11042 (output_skeleton_debug_sections): Likewise.
11043 (output_rnglists, output_line_info): Likewise.
11044 (output_pubnames, output_aranges): Update version comments.
11045 (output_one_line_info_table): Output view numbers in asm comments.
11046 (dw_loc_list): Determine current endview, pass it to new_loc_list.
11047 Call maybe_gen_llsym.
11048 (loc_list_from_tree_1): Adjust.
11049 (add_AT_location_description): Create view list attribute if
11050 needed, check it's absent otherwise.
11051 (convert_cfa_to_fb_loc_list): Adjust.
11052 (maybe_emit_file): Call output_asm_line_debug_info for test.
11053 (dwarf2out_var_location): Reset views as needed. Precompute
11054 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
11055 attribute. Set view.
11056 (new_line_info_table): Reset next view.
11057 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
11058 (dwarf2out_source_line): Likewise. Output view resets and labels to
11059 the assembler, or select appropriate line info opcodes.
11060 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
11061 (optimize_string_length): Catch it. Adjust.
11062 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
11063 dw_val_class_view_list, and remove it if no longer needed.
11064 (hash_loc_list): Hash view numbers.
11065 (loc_list_hasher::equal): Compare them.
11066 (optimize_location_lists): Check whether a view list symbol is
11067 needed, and whether the locview attribute is present, and
11068 whether they match. Remove the locview attribute if no longer
11070 (index_location_lists): Call skip_loc_list_entry for test.
11071 (dwarf2out_finish): Call output_asm_line_debug_info for test.
11072 Use output_dwarf_version.
11073 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
11074 (struct dw_val_node): Add val_view_list.
11075 * final.c (SEEN_NEXT_VIEW): New.
11076 (set_next_view_needed): New.
11077 (clear_next_view_needed): New.
11078 (maybe_output_next_view): New.
11079 (final_start_function): Rename to...
11080 (final_start_function_1): ... this. Take pointer to FIRST,
11081 add SEEN parameter. Emit param bindings in the initial view.
11082 (final_start_function): Reintroduce SEEN-less interface.
11083 (final): Rename to...
11084 (final_1): ... this. Take SEEN parameter. Output final pending
11085 next view at the end.
11086 (final): Reintroduce seen-less interface.
11087 (final_scan_insn): Output pending next view before switching
11088 sections or ending a block. Mark the next view as needed when
11089 outputting variable locations. Notify debug backend of section
11090 changes, and of location view changes.
11091 (rest_of_handle_final): Adjust.
11092 * toplev.c (process_options): Autodetect value for debug variable
11093 location views option. Warn on incompat5 without -gdwarf-5.
11094 * doc/invoke.texi (gvariable-location-views): New.
11095 (gvariable-location-views=incompat5): New.
11096 (gno-variable-location-views): New.
11098 2018-02-08 David Malcolm <dmalcolm@redhat.com>
11100 PR tree-optimization/84136
11101 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
11102 that the result of find_edge is non-NULL.
11104 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11107 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
11108 storing integer register in SImode. Fix cost of 256 and 512
11109 byte aligned SSE register store.
11111 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11113 * config/i386/i386.c (ix86_multiplication_cost): Fix
11114 multiplication cost for TARGET_AVX512DQ.
11116 2018-02-08 Marek Polacek <polacek@redhat.com>
11118 PR tree-optimization/84238
11119 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
11122 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11124 PR tree-optimization/84265
11125 * tree-vect-stmts.c (vectorizable_store): Don't treat
11126 VMAT_CONTIGUOUS accesses as grouped.
11127 (vectorizable_load): Likewise.
11129 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11131 PR tree-optimization/81635
11132 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
11133 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
11134 (test_round_for_mask): New functions.
11135 (wide_int_cc_tests): Call test_round_for_mask.
11136 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
11137 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
11138 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
11139 range returned by get_range_info.
11141 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11144 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
11145 * symtab.c: Include builtins.h
11146 (symtab_node::output_to_lto_symbol_table_p): Move here
11147 from lto-streamer-out.c:output_symbol_p.
11148 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
11149 (output_symbol_p): Move all logic to symtab.c
11150 (produce_symtab): Update.
11152 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11154 * config/s390/s390-opts.h (enum indirect_branch): Define.
11155 * config/s390/s390-protos.h (s390_return_addr_from_memory)
11156 (s390_indirect_branch_via_thunk)
11157 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
11158 (enum s390_indirect_branch_type): Define.
11159 * config/s390/s390.c (struct s390_frame_layout, struct
11160 machine_function): Remove.
11161 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
11162 (indirect_branch_table_label_no, indirect_branch_table_name):
11164 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
11165 (enum s390_indirect_branch_option): Define.
11166 (s390_return_addr_from_memory): New function.
11167 (s390_handle_string_attribute): New function.
11168 (s390_attribute_table): Add new attribute handler.
11169 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
11170 (s390_indirect_branch_via_thunk): New function.
11171 (s390_indirect_branch_via_inline_thunk): New function.
11172 (s390_function_ok_for_sibcall): When jumping via thunk disallow
11173 sibling call optimization for non z10 compiles.
11174 (s390_emit_call): Force indirect branch target to be a single
11175 register. Add r1 clobber for non-z10 compiles.
11176 (s390_emit_epilogue): Emit return jump via return_use expander.
11177 (s390_reorg): Handle JUMP_INSNs as execute targets.
11178 (s390_option_override_internal): Perform validity checks for the
11179 new command line options.
11180 (s390_indirect_branch_attrvalue): New function.
11181 (s390_indirect_branch_settings): New function.
11182 (s390_set_current_function): Invoke s390_indirect_branch_settings.
11183 (s390_output_indirect_thunk_function): New function.
11184 (s390_code_end): Implement target hook.
11185 (s390_case_values_threshold): Implement target hook.
11186 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
11188 * config/s390/s390.h (struct s390_frame_layout)
11189 (struct machine_function): Move here from s390.c.
11190 (TARGET_INDIRECT_BRANCH_NOBP_RET)
11191 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
11192 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
11193 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
11194 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
11195 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
11196 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
11197 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
11198 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
11199 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
11200 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
11201 (mnemonic attribute): Add values which aren't recognized
11203 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
11204 pattern for branch conversion. Fix mnemonic attribute.
11205 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
11206 indirect branch via thunk if requested.
11207 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
11208 ("*indirect_jump"): Disable for branch conversion using out of
11210 ("indirect_jump_via_thunk<mode>_z10")
11211 ("indirect_jump_via_thunk<mode>")
11212 ("indirect_jump_via_inlinethunk<mode>_z10")
11213 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
11214 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
11215 ("casesi_jump_via_inlinethunk<mode>_z10")
11216 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
11217 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
11218 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
11219 ("*indirect2_jump"): Disable for branch conversion.
11220 ("casesi_jump"): Turn into expander and expand patterns for branch
11222 ("return_use"): New expander.
11223 ("*return"): Emit return via thunk and rename it to ...
11224 ("*return<mode>"): ... this one.
11225 * config/s390/s390.opt: Add new options and and enum for the
11228 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11230 * lra-constraints.c (match_reload): Unconditionally use
11231 gen_lowpart_SUBREG, rather than selecting between that
11232 and equivalent gen_rtx_SUBREG code.
11234 2018-02-08 Richard Biener <rguenther@suse.de>
11236 PR tree-optimization/84233
11237 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
11238 changed flag instead of boguously re-using phi_inserted.
11240 2018-02-08 Martin Jambor <mjambor@suse.cz>
11242 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
11243 static local variables.
11245 2018-02-08 Richard Biener <rguenther@suse.de>
11247 PR tree-optimization/84278
11248 * tree-vect-stmts.c (vectorizable_store): When looking for
11249 smaller vector types to perform grouped strided loads/stores
11250 make sure the mode is supported by the target.
11251 (vectorizable_load): Likewise.
11253 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11255 * config/aarch64/aarch64.c (aarch64_components_for_bb):
11256 Increase LDP/STP opportunities by adding adjacent callee-saves.
11258 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11260 PR rtl-optimization/84068
11261 PR rtl-optimization/83459
11262 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
11264 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
11266 PR tree-optimization/84224
11267 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
11268 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
11269 non-zero arguments.
11271 2018-02-07 Iain Sandoe <iain@codesourcery.com>
11274 * config/rs6000/altivec.md (*restore_world): Remove LR use.
11275 * config/rs6000/predicates.md (restore_world_operation): Adjust op
11276 count, remove one USE.
11278 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
11280 * doc/install.texi (Configuration): Document the
11281 --with-long-double-format={ibm,ieee} PowerPC configuration
11285 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
11286 Convert from define_expand to be define_insn_and_split. Rework
11287 float/double/_Float128 conversions to QI/HI/SImode to work with
11288 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
11289 conversions to QI/HImode types did a store and then a load to
11290 truncate the value. For conversions to VSX registers, don't split
11291 the insn, instead emit the code directly. Use the code iterator
11292 any_fix to combine signed and unsigned conversions.
11293 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
11294 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
11295 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11296 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
11297 (fix_<mode>di2_hw): Likewise.
11298 (fixuns_<mode>di2_hw): Likewise.
11299 (fix_<mode>si2_hw): Likewise.
11300 (fixuns_<mode>si2_hw): Likewise.
11301 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
11302 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11303 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
11304 fix<uns>_trunc<SFDF:mode>si2_p8.
11305 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
11306 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
11307 (fix<uns>_<mode>_mem): Likewise.
11308 (fctiw<u>z_<mode>_mem): Likewise.
11309 (fix<uns>_<mode>_mem): Likewise.
11310 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
11311 the register allocator from doing a direct move to the GPRs to do
11312 a store, and instead use the ISA 3.0 store byte/half-word from
11313 vector register instruction. For IEEE 128-bit floating point,
11314 also optimize stores of 32-bit ints.
11315 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
11317 2018-02-07 Alan Hayward <alan.hayward@arm.com>
11319 * genextract.c (push_pathstr_operand): New function to support
11321 (walk_rtx): Call push_pathstr_operand.
11322 (print_path): Support [a-zA-Z].
11324 2018-02-07 Richard Biener <rguenther@suse.de>
11326 PR tree-optimization/84037
11327 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
11328 (cse_and_gimplify_to_preheader): Declare.
11329 (vect_get_place_in_interleaving_chain): Likewise.
11330 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11332 (_loop_vec_info::~_loop_vec_info): Delete it.
11333 (cse_and_gimplify_to_preheader): New function.
11334 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
11335 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
11336 (vectorizable_load): Likewise. For grouped stores always base
11337 the IV on the first element.
11338 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
11339 condition before gimplifying.
11341 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11343 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
11344 *DIV_EXPR and *MOD_EXPR.
11346 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
11349 * config/i386/i386.c (ix86_option_override_internal): Mask out
11350 the CF_SET bit when checking -fcf-protection.
11352 2018-02-07 Tom de Vries <tom@codesourcery.com>
11355 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
11358 2018-02-07 Richard Biener <rguenther@suse.de>
11360 PR tree-optimization/84204
11361 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
11364 PR tree-optimization/84205
11365 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
11366 special-case isl_ast_op_zdiv_r.
11368 PR tree-optimization/84223
11369 * graphite-scop-detection.c (gather_bbs::before_dom_children):
11370 Only add conditions from within the region.
11371 (gather_bbs::after_dom_children): Adjust.
11373 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
11376 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
11377 * config/avr/avr.md: Only post-reload split REG-REG moves if
11378 either register is GENERAL_REG_P.
11380 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11382 PR tree-optimization/84235
11383 * tree-ssa-scopedtables.c
11384 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
11385 if the subtraction is performed in floating point type where NaNs are
11386 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
11387 build 1. Formatting fix.
11389 2018-02-06 Jakub Jelinek <jakub@redhat.com>
11392 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
11393 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
11394 and skip it regardless of bb boundaries. Use CALL_P macro,
11395 don't test INSN_P (insn) together with CALL_P or JUMP_P check
11396 unnecessarily, formatting fix.
11398 2018-02-06 Michael Collison <michael.collison@arm.com>
11400 * config/arm/thumb2.md:
11401 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
11402 (*thumb_mov_notscc): Ditto.
11404 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
11407 * config/rs6000/rs6000.md (su code attribute): Use "u" for
11408 unsigned_fix, not "s".
11410 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11412 * configure.ac (gcc_fn_eh_frame_ro): New function.
11413 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
11414 correct .eh_frame permissions.
11415 * configure: Regenerate.
11417 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
11419 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
11420 irrelevant options.
11422 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11424 * config/rs6000/rs6000.c (rs6000_option_override_internal):
11425 Display warning message for -mno-speculate-indirect-jumps.
11427 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
11429 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
11431 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
11433 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
11435 PR tree-optimization/84225
11436 * tree-eh.c (find_trapping_overflow): Only call
11437 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
11439 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11442 * config/i386/i386.c: Reimplement the check of possible options
11443 -mibt/-mshstk conbination. Change error messages.
11444 * doc/invoke.texi: Fix a typo: remove extra '='.
11446 2018-02-06 Marek Polacek <polacek@redhat.com>
11448 PR tree-optimization/84228
11449 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
11451 2018-02-06 Tamar Christina <tamar.christina@arm.com>
11454 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
11455 emitted arch directives.
11456 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
11457 __ARM_FEATURE_COPROC before changing architectures.
11459 2018-02-06 Richard Biener <rguenther@suse.de>
11461 * config/i386/i386.c (print_reg): Fix typo.
11462 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
11464 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
11466 * configure: Regenerate.
11468 2018-02-05 Martin Sebor <msebor@redhat.com>
11470 PR tree-optimization/83369
11471 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
11474 2018-02-05 Martin Liska <mliska@suse.cz>
11476 * doc/invoke.texi: Cherry-pick upstream r323995.
11478 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
11480 * ira.c (ira_init_register_move_cost): Adjust comment.
11482 2018-02-05 Martin Liska <mliska@suse.cz>
11484 PR gcov-profile/84137
11485 * doc/gcov.texi: Fix typo in documentation.
11487 2018-02-05 Martin Liska <mliska@suse.cz>
11489 PR gcov-profile/83879
11490 * doc/gcov.texi: Document necessity of --dynamic-list-data when
11491 using dlopen functionality.
11493 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
11495 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
11496 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
11497 _mm_maskz_range_ss, _mm_mask_range_round_ss,
11498 _mm_maskz_range_round_ss): New intrinsics.
11499 (__builtin_ia32_rangesd128_round)
11500 (__builtin_ia32_rangess128_round): Remove.
11501 (__builtin_ia32_rangesd128_mask_round,
11502 __builtin_ia32_rangess128_mask_round): New builtins.
11503 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
11504 __builtin_ia32_rangess128_round): Remove.
11505 (__builtin_ia32_rangesd128_mask_round,
11506 __builtin_ia32_rangess128_mask_round): New builtins.
11507 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
11508 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
11509 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
11510 "<round_saeonly_constraint>")): Changed to ...
11511 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
11512 "<round_saeonly_scalar_constraint>")): ... this.
11513 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
11514 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
11515 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
11516 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
11517 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
11519 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
11521 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
11523 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
11524 Remove all values except native, 8540 and 8548.
11526 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
11528 * config/i386/i386.c (ix86_output_function_return): Pass
11529 INVALID_REGNUM, instead of -1, as invalid register number to
11530 indirect_thunk_name and output_indirect_thunk.
11532 2018-02-02 Julia Koval <julia.koval@intel.com>
11534 * config.gcc: Add -march=icelake.
11535 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
11536 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
11537 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
11538 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
11539 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
11540 (processor_target_table): Add icelake.
11541 (ix86_option_override_internal): Handle new PTAs.
11542 (get_builtin_code_for_version): Handle icelake.
11543 (M_INTEL_COREI7_ICELAKE): New.
11544 (fold_builtin_cpu): Handle icelake.
11545 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
11546 * doc/invoke.texi: Add -march=icelake.
11548 2018-02-02 Julia Koval <julia.koval@intel.com>
11550 * config/i386/i386.c (ix86_option_override_internal): Change flags type
11551 to wide_int_bitmask.
11552 * wide-int-bitmask.h: New.
11554 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11557 * config/i386/i386.md: Replace Pmode with word_mode in
11558 builtin_setjmp_setup and builtin_longjmp to support x32.
11560 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
11564 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
11566 (rs6000_supported_cpu_names): New static variable.
11567 (linux_cpu_translation_table): Likewise.
11568 (elf_platform) <cpu>: Define new static variable and use it.
11569 Translate kernel AT_PLATFORM name to canonical name if needed.
11570 Error if platform name is unknown.
11572 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
11575 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
11577 2018-02-01 Jeff Law <law@redhat.com>
11580 * config/i386/i386.c (release_scratch_register_on_entry): Add new
11581 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
11582 the scratch if RELEASE_VIA_POP is false.
11583 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
11584 If we have to save a temporary register, decrement SIZE appropriately.
11585 Pass new arguments to release_scratch_register_on_entry.
11586 (ix86_adjust_stack_and_probe): Likewise.
11587 (ix86_emit_probe_stack_range): Pass new arguments to
11588 release_scratch_register_on_entry.
11590 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
11592 PR rtl-optimization/84157
11593 * combine.c (change_zero_ext): Use REG_P predicate in
11594 front of HARD_REGISTER_P predicate.
11596 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
11598 * config/avr/avr.c (avr_option_override): Move disabling of
11599 -fdelete-null-pointer-checks to...
11600 * common/config/avr/avr-common.c (avr_option_optimization_table):
11603 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11605 PR tree-optimization/81635
11606 * tree-data-ref.c (split_constant_offset_1): For types that
11607 wrap on overflow, try to use range info to prove that wrapping
11610 2018-02-01 Renlin Li <renlin.li@arm.com>
11613 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
11614 TAILCALL_ADDR_REGS.
11615 (aarch64_register_move_cost): Likewise.
11616 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
11617 TAILCALL_ADDR_REGS.
11618 (REG_CLASS_NAMES): Likewise.
11619 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
11620 TAILCALL_ADDR_REGS. Remove IP registers.
11621 * config/aarch64/aarch64.md (Ucs): Update register constraint.
11623 2018-02-01 Richard Biener <rguenther@suse.de>
11625 * domwalk.h (dom_walker::dom_walker): Add additional constructor
11626 for specifying RPO order and allow NULL for that.
11627 * domwalk.c (dom_walker::dom_walker): Likewise.
11628 (dom_walker::walk): Handle NULL RPO order.
11629 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
11631 (rewrite_update_dom_walker): Likewise.
11632 (mark_def_dom_walker): Likewise.
11634 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11636 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
11637 (aarch64_maybe_expand_sve_subreg_move): Declare.
11638 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
11639 * config/aarch64/predicates.md (aarch64_any_register_operand): New
11641 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
11642 that are semantically a reverse operation.
11643 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
11644 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
11645 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
11647 (aarch64_can_change_mode_class): For big-endian, forbid changes
11648 between two SVE modes if they have different element sizes.
11650 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11652 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
11653 the TImode handling for big-endian targets.
11655 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11657 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
11658 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
11660 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
11661 Remove BSWAP handing for big-endian targets and use the form of
11662 LD1RQ appropariate for the mode.
11664 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11666 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
11667 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
11668 duplicated element.
11670 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11673 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
11674 check for operands that need to go through aarch64_sve_reload_be.
11676 2018-02-01 Jakub Jelinek <jakub@redhat.com>
11678 PR tree-optimization/81661
11679 PR tree-optimization/84117
11680 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
11681 * tree-eh.c: Include gimplify.h.
11682 (find_trapping_overflow, replace_trapping_overflow,
11683 rewrite_to_non_trapping_overflow): New functions.
11684 * tree-vect-loop.c: Include tree-eh.h.
11685 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
11686 * tree-data-ref.c: Include tree-eh.h.
11687 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
11689 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
11691 PR rtl-optimization/84123
11692 * combine.c (change_zero_ext): Check if hard register satisfies
11693 can_change_dest_mode before calling gen_lowpart_SUBREG.
11695 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
11698 * ira.c (ira_init_register_move_cost): Remove assert.
11700 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11702 PR rtl-optimization/84071
11703 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
11704 * doc/tm.texi: Regenerate.
11706 2018-01-31 Richard Biener <rguenther@suse.de>
11708 PR tree-optimization/84132
11709 * tree-data-ref.c (analyze_miv_subscript): Properly
11710 check whether evolution_function_is_affine_multivariate_p
11711 before calling gcd_of_steps_may_divide_p.
11713 2018-01-31 Julia Koval <julia.koval@intel.com>
11716 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
11717 * config/i386/i386.md (rdpid_rex64) New.
11718 (rdpid): Make 32bit only.
11720 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
11723 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
11724 an IDENTIFIER_NODE for FUNCTION_TYPE's.
11726 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11729 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11731 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11733 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11735 PR rtl-optimization/84071
11736 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
11737 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
11739 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11741 * config/arc/arc.c (arc_handle_aux_attribute): New function.
11742 (arc_attribute_table): Add 'aux' attribute.
11743 (arc_in_small_data_p): Consider aux like variables.
11744 (arc_is_aux_reg_p): New function.
11745 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
11746 (arc_get_aux_arg): New function.
11747 (prepare_move_operands): Handle aux-register access.
11748 (arc_handle_aux_attribute): New function.
11749 * doc/extend.texi (ARC Variable attributes): Add subsection.
11751 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11753 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
11754 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
11755 (arc_attribute_table): Add 'uncached' attribute.
11756 (arc_print_operand): Print '.di' flag for uncached memory
11758 (arc_in_small_data_p): Do not consider for small data the uncached
11760 (arc_is_uncached_mem_p): New function.
11761 * config/arc/predicates.md (compact_store_memory_operand): Check
11762 for uncached memory accesses.
11763 (nonvol_nonimm_operand): Likewise.
11764 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
11766 2018-01-31 Jakub Jelinek <jakub@redhat.com>
11769 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
11770 falign-loops=): Add Optimization flag.
11772 2018-01-30 Jeff Law <law@redhat.com>
11775 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
11776 INT_REGISTERS_SAVED. Check it prior to calling
11777 get_scratch_register_on_entry.
11778 (ix86_adjust_stack_and_probe): Similarly.
11779 (ix86_emit_probe_stack_range): Similarly.
11780 (ix86_expand_prologue): Corresponding changes.
11782 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11785 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
11786 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
11788 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
11791 * lra-constraints.c (curr_insn_transform): Process AND in the
11794 2018-01-30 Jakub Jelinek <jakub@redhat.com>
11796 PR rtl-optimization/83986
11797 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
11798 dependence against last_pending_memory_flush in addition to
11799 pending_jump_insns.
11801 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
11803 PR tree-optimization/81611
11804 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
11807 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11810 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
11813 2018-01-30 Richard Biener <rguenther@suse.de>
11814 Jakub Jelinek <jakub@redhat.com>
11816 PR tree-optimization/84111
11817 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
11818 inner loops added during recursion, as they don't have up-to-date
11821 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11824 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
11825 (can_inline_edge_by_limits_p): ... here.
11826 (can_early_inline_edge_p, check_callers,
11827 update_caller_keys, update_callee_keys, recursive_inlining,
11828 add_new_edges_to_heap, speculation_useful_p,
11829 inline_small_functions,
11830 inline_small_functions, flatten_function,
11831 inline_to_all_callers_1): Update.
11833 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11835 * profile-count.c (profile_count::combine_with_ipa_count): Handle
11838 2018-01-30 Richard Biener <rguenther@suse.de>
11840 PR tree-optimization/83008
11841 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
11842 invariant and constant vector uses in stmts when they need
11843 more than one stmt.
11845 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11848 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
11849 * configure: Regenerate.
11851 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11853 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
11855 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
11856 Use gen_rtx_REG rather than gen_lowpart.
11858 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11860 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
11861 rather than 0 when creating partial subregs.
11863 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11865 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
11868 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
11871 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
11872 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
11873 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
11874 flags. This restores the settings used before the 2017-07-24.
11875 Turning off pre increment/decrement/modify allows IVOPTS to
11876 optimize DF/SF loops where the index is an int.
11878 2018-01-29 Richard Biener <rguenther@suse.de>
11879 Kelvin Nilsen <kelvin@gcc.gnu.org>
11882 * tree-vect-stmts.c (vectorizable_call): Don't call
11883 targetm.vectorize_builtin_md_vectorized_function if callee is
11886 2018-01-22 Carl Love <cel@us.ibm.com>
11888 * doc/extend.tex: Fix typo in second arg in
11889 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
11891 2018-01-29 Richard Biener <rguenther@suse.de>
11893 PR tree-optimization/84086
11894 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
11895 (flush_ssaname_freelist): When SSA names were released reset
11896 the SCEV hash table.
11898 2018-01-29 Richard Biener <rguenther@suse.de>
11900 PR tree-optimization/84057
11901 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
11902 removed paths when removing edges.
11904 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
11906 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
11907 -mfunction-return=@var{choice}.
11909 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
11911 PR diagnostic/84034
11912 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
11913 Handle CR like TAB.
11914 (layout::print_source_line): Likewise.
11915 (test_get_line_width_without_trailing_whitespace): Add test cases.
11917 2018-01-27 Jakub Jelinek <jakub@redhat.com>
11919 PR middle-end/84040
11920 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
11923 2018-01-26 Jim Wilson <jimw@sifive.com>
11925 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
11927 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
11930 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11932 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
11933 and CMP + SUB-immediate -> SUBS.
11935 2018-01-26 Martin Sebor <msebor@redhat.com>
11937 PR tree-optimization/83896
11938 * tree-ssa-strlen.c (get_string_len): Rename...
11939 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
11940 Avoid assuming length is constant.
11941 (handle_char_store): Use HOST_WIDE_INT for string length.
11943 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
11946 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
11947 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
11949 2018-01-26 Richard Biener <rguenther@suse.de>
11951 PR rtl-optimization/84003
11952 * dse.c (record_store): Only record redundant stores when
11953 the earlier store aliases at least all accesses the later one does.
11955 2018-01-26 Jakub Jelinek <jakub@redhat.com>
11957 PR rtl-optimization/83985
11958 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
11959 REG_CFA_RESTORE insns.
11960 (delete_unmarked_insns): Don't ignore separate shrink wrapping
11961 REG_CFA_RESTORE insns here.
11964 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
11965 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
11967 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11969 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
11970 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
11971 (arc_init): Likewise.
11972 (arc_override_options): Likewise.
11973 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
11975 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
11977 * config/arc/arc.h (TARGET_DBNZ): Define.
11978 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
11979 properly set the tune attribute.
11980 (dbnz): Use TARGET_DBNZ guard.
11981 * config/arc/arc.opt (mtune): Add core3 option.
11983 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11985 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
11986 recognize new pic like addresses.
11987 (arc_delegitimize_address): Clean up.
11989 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11991 * config/arc/arc-arches.def: Option mrf16 valid for all
11993 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
11994 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
11995 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
11996 * config/arc/arc-tables.opt: Regenerate.
11997 * config/arc/arc.c (arc_conditional_register_usage): Handle
11998 reduced register file case.
11999 (arc_file_start): Set must have build attributes.
12000 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
12001 mrf16 option value.
12002 * config/arc/arc.opt (mrf16): Add new option.
12003 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
12004 * config/arc/genmultilib.awk: Handle new mrf16 option.
12005 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
12006 * config/arc/t-multilib: Regenerate.
12007 * doc/invoke.texi (ARC Options): Document mrf16 option.
12009 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12011 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
12012 * config/arc/arc.c (arc_handle_secure_attribute): New function.
12013 (arc_attribute_table): Add 'secure_call' attribute.
12014 (arc_print_operand): Print secure call operand.
12015 (arc_function_ok_for_sibcall): Don't optimize tail calls when
12017 (arc_is_secure_call_p): New function. * config/arc/arc.md
12018 (call_i): Add support for sjli instruction.
12019 (call_value_i): Likewise.
12020 * config/arc/constraints.md (Csc): New constraint.
12022 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12023 John Eric Martin <John.Martin@emmicro-us.com>
12025 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
12026 * config/arc/arc.c (_arc_jli_section): New struct.
12027 (arc_jli_section): New type.
12028 (rc_jli_sections): New static variable.
12029 (arc_handle_jli_attribute): New function.
12030 (arc_attribute_table): Add jli_always and jli_fixed attribute.
12031 (arc_file_end): New function.
12032 (TARGET_ASM_FILE_END): Define.
12033 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
12034 (arc_add_jli_section): New function.
12035 (jli_call_scan): Likewise.
12036 (arc_reorg): Call jli_call_scan.
12037 (arc_output_addsi): Remove 'S' from printing asm operand.
12038 (arc_is_jli_call_p): New function.
12039 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
12041 (movhi_insn): Likewise.
12042 (movsi_insn): Likewise.
12043 (movsi_set_cc_insn): Likewise.
12044 (loadqi_update): Likewise.
12045 (load_zeroextendqisi_update): Likewise.
12046 (load_signextendqisi_update): Likewise.
12047 (loadhi_update): Likewise.
12048 (load_zeroextendhisi_update): Likewise.
12049 (load_signextendhisi_update): Likewise.
12050 (loadsi_update): Likewise.
12051 (loadsf_update): Likewise.
12052 (movsicc_insn): Likewise.
12053 (bset_insn): Likewise.
12054 (bxor_insn): Likewise.
12055 (bclr_insn): Likewise.
12056 (bmsk_insn): Likewise.
12057 (bicsi3_insn): Likewise.
12058 (cmpsi_cc_c_insn): Likewise.
12059 (movsi_ne): Likewise.
12060 (movsi_cond_exec): Likewise.
12061 (clrsbsi2): Likewise.
12062 (norm_f): Likewise.
12072 (call_i): Remove 'S' asm letter, add jli instruction.
12073 (call_value_i): Likewise.
12074 * config/arc/arc.op (mjli-always): New option.
12075 * config/arc/constraints.md (Cji): New constraint.
12076 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
12078 (subsf3_fpx): Likewise.
12079 (mulsf3_fpx): Likewise.
12080 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
12082 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
12083 function attrbutes.
12084 * doc/invoke.texi (ARC): Document mjli-always option.
12086 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
12088 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
12089 avoid addition with 0 and use incw and decw where possible.
12091 2018-01-26 Richard Biener <rguenther@suse.de>
12093 PR tree-optimization/81082
12094 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
12095 association if it requires casting to unsigned.
12096 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
12097 from fold_plusminus_mult_expr to catch important cases late when
12098 range info is available.
12100 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12102 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
12103 * configure.ac (hidden_linkonce): New test.
12104 * configure: Regenerate.
12105 * config.in: Regenerate.
12107 2018-01-26 Julia Koval <julia.koval@intel.com>
12109 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
12110 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
12111 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
12112 _mm_mask_bitshuffle_epi64_mask): Fix type.
12113 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
12114 USI_FTYPE_V4DI_V4DI_USI): Remove.
12115 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
12116 __builtin_ia32_vpshufbitqmb256_mask,
12117 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
12118 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
12119 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
12121 2018-01-26 Alan Modra <amodra@gmail.com>
12124 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
12125 UNSPEC_VBPERMQ. Sort other unspecs.
12127 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
12129 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
12131 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
12133 PR middle-end/83055
12134 * predict.c (drop_profile): Do not push/pop cfun; update also
12136 (handle_missing_profiles): Fix logic looking for zero profiles.
12138 2018-01-25 Jakub Jelinek <jakub@redhat.com>
12140 PR middle-end/83977
12141 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
12142 on functions with #pragma omp declare simd or functions with simd
12144 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
12145 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
12146 Remove trailing \n from warning_at calls.
12148 2018-01-25 Tom de Vries <tom@codesourcery.com>
12151 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12152 for neutered workers.
12154 2018-01-24 Joseph Myers <joseph@codesourcery.com>
12157 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
12158 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
12160 2018-01-24 Jeff Law <law@redhat.com>
12163 * i386.c (get_probe_interval): Move to earlier point.
12164 (ix86_compute_frame_layout): If -fstack-clash-protection and
12165 the frame is larger than the probe interval, then use pushes
12166 to save registers rather than reg->mem moves.
12167 (ix86_expand_prologue): Remove conditional for int_registers_saved
12170 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
12173 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
12174 min/max for never referenced object.
12176 2018-01-24 Jakub Jelinek <jakub@redhat.com>
12178 PR middle-end/83977
12179 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
12181 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
12182 attributes from DECL_ATTRIBUTES (decl) without affecting
12183 DECL_ATTRIBUTES (current_function_decl).
12184 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
12185 functions with non-NULL DECL_ABSTRACT_ORIGIN.
12187 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
12189 PR tree-optimization/83979
12190 * fold-const.c (fold_comparison): Use constant_boolean_node
12191 instead of boolean_{true,false}_node.
12193 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
12195 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
12198 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12200 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12201 Simplify the clause that sets the length attribute.
12202 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12203 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
12204 clause that sets the length attribute.
12205 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12207 2018-01-24 Tom de Vries <tom@codesourcery.com>
12210 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
12211 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
12212 Add strict parameter.
12213 (prevent_branch_around_nothing): Insert dummy insn between branch to
12214 label and label with no ptx insn inbetween.
12215 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
12217 2018-01-24 Tom de Vries <tom@codesourcery.com>
12220 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12221 for neutered threads in warp.
12222 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
12224 2018-01-24 Richard Biener <rguenther@suse.de>
12226 PR tree-optimization/83176
12227 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
12230 2018-01-24 Richard Biener <rguenther@suse.de>
12232 PR tree-optimization/82819
12233 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
12234 code generating pluses that are no-ops in the target precision.
12236 2018-01-24 Richard Biener <rguenther@suse.de>
12238 PR middle-end/84000
12239 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
12241 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12243 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
12244 to merge probabilities.
12245 * predict.c (probably_never_executed): Also mark as cold functions
12246 with global 0 profile and guessed local profile.
12247 * profile-count.c (profile_probability::combine_with_count): New
12249 * profile-count.h (profile_probability::operator*,
12250 profile_probability::operator*=, profile_probability::operator/,
12251 profile_probability::operator/=): Reduce precision to adjusted
12252 and set value to guessed on contradictory divisions.
12253 (profile_probability::combine_with_freq): Remove.
12254 (profile_probability::combine_wiht_count): Declare.
12255 (profile_count::force_nonzero):: Set to adjusted.
12256 (profile_count::probability_in):: Set quality to adjusted.
12257 * tree-ssa-tail-merge.c (replace_block_by): Use
12258 combine_with_count.
12260 2018-01-23 Andrew Waterman <andrew@sifive.com>
12261 Jim Wilson <jimw@sifive.com>
12263 * config/riscv/riscv.c (riscv_stack_boundary): New.
12264 (riscv_option_override): Set riscv_stack_boundary. Handle
12265 riscv_preferred_stack_boundary_arg.
12266 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
12267 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
12268 (STACK_BOUNDARY): Set to riscv_stack_boundary.
12269 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
12270 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
12271 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
12273 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
12276 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
12277 of struct ix86_frame.
12278 (ix86_expand_epilogue): Likewise. Add a local variable for
12279 the reg_save_offset field in struct ix86_frame.
12281 2018-01-23 Bin Cheng <bin.cheng@arm.com>
12283 PR tree-optimization/82604
12284 * tree-loop-distribution.c (enum partition_kind): New enum item
12285 PKIND_PARTIAL_MEMSET.
12286 (partition_builtin_p): Support above new enum item.
12287 (generate_code_for_partition): Ditto.
12288 (compute_access_range): Differentiate cases that equality can be
12289 proven at all loops, the innermost loops or no loops.
12290 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
12291 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
12292 (finalize_partitions, distribute_loop): Don't fuse partition of
12293 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
12294 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
12295 parloop is enabled.
12297 2018-01-23 Martin Liska <mliska@suse.cz>
12299 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
12300 order to ignore the predictor.
12301 (PRED_POLYMORPHIC_CALL): Likewise.
12302 (PRED_RECURSIVE_CALL): Likewise.
12304 2018-01-23 Martin Liska <mliska@suse.cz>
12306 * tree-profile.c (tree_profiling): Print function header to
12307 aware reader which function we are working on.
12308 * value-prof.c (gimple_find_values_to_profile): Do not print
12309 not interesting value histograms.
12311 2018-01-23 Martin Liska <mliska@suse.cz>
12313 * profile-count.h (enum profile_quality): Add
12314 profile_uninitialized as the first value. Do not number values
12315 as they are zero based.
12316 (profile_count::verify): Update sanity check.
12317 (profile_probability::verify): Likewise.
12319 2018-01-23 Nathan Sidwell <nathan@acm.org>
12321 * doc/invoke.texi (ffor-scope): Deprecate.
12323 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12325 PR tree-optimization/83510
12326 * domwalk.c (set_all_edges_as_executable): New function.
12327 (dom_walker::dom_walker): Convert bool param
12328 "skip_unreachable_blocks" to enum reachability. Move setup of
12329 edge flags to set_all_edges_as_executable and only do it when
12330 reachability is REACHABLE_BLOCKS.
12331 * domwalk.h (enum dom_walker::reachability): New enum.
12332 (dom_walker::dom_walker): Convert bool param
12333 "skip_unreachable_blocks" to enum reachability.
12334 (set_all_edges_as_executable): New decl.
12335 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
12336 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
12338 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
12339 but converting true to REACHABLE_BLOCKS.
12340 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
12342 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
12343 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
12344 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
12346 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
12347 if check_all_array_refs will be called.
12349 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12351 * tree.c (selftest::test_location_wrappers): Add more test
12354 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12356 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
12357 (selftest::test_bit_in_range): Likewise.
12359 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12362 * doc/sourcebuild.texi (vect_float): Say that the selector
12363 only describes the situation when -funsafe-math-optimizations is on.
12364 (vect_float_strict): Document.
12366 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12368 PR tree-optimization/83965
12369 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
12370 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
12371 instead of checking only for a reduction.
12372 (vect_recog_widen_sum_pattern): Likewise.
12374 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12376 * predict.c (probably_never_executed): Only use precise profile info.
12377 (compute_function_frequency): Skip after inlining hack since we now
12378 have quality checking.
12380 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12382 * profile-count.h (profile_probability::very_unlikely,
12383 profile_probability::unlikely, profile_probability::even): Set
12384 precision to guessed.
12386 2018-01-23 Richard Biener <rguenther@suse.de>
12388 PR tree-optimization/83963
12389 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
12390 Properly terminate dominator walk when crossing the exit edge not
12391 when visiting its source block.
12393 2018-01-23 Jakub Jelinek <jakub@redhat.com>
12396 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
12397 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
12399 2018-01-22 Jakub Jelinek <jakub@redhat.com>
12401 PR tree-optimization/83957
12402 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
12403 semicolon after for body surrounded by braces.
12405 PR tree-optimization/83081
12406 * profile-count.h (profile_probability::split): New method.
12407 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
12408 Use profile_probability::split.
12409 (do_compare_rtx_and_jump): Fix adjustment of probabilities
12410 when splitting a single conditional jump into 2.
12412 2018-01-22 David Malcolm <dmalcolm@redhat.com>
12414 PR tree-optimization/69452
12415 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
12418 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12420 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
12421 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
12422 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
12424 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12426 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
12428 * config/rl78/rl78.md (movdi): New define_expand.
12429 * config/rl78/rl78.c (rl78_split_movdi): New function.
12431 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
12434 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
12436 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
12437 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
12438 128-bit to produce an UNSPEC move to get the double word with the
12439 signbit and then a shift directly to do signbit.
12440 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
12441 implementation with a new version that just does either a direct
12442 move or a regular move. Move memory interface to separate insns.
12443 Move insns so they are next to the expander.
12444 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
12445 with signbit move. Split big and little endian case.
12446 (signbit<mode>2_dm_mem_le): Likewise.
12447 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
12448 (signbit<mode>2_dm2): Likewise.
12450 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12452 * config/rl78/rl78.md (anddi3): New define_expand.
12454 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12456 * config/rl78/rl78.md (umindi3): New define_expand.
12458 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12460 * config/rl78/rl78.md (smindi3): New define_expand.
12462 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12464 * config/rl78/rl78.md (smaxdi3): New define_expand.
12466 2018-01-22 Carl Love <cel@us.ibm.com>
12468 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
12469 LVX_V1TI): Add macro expansion.
12470 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
12471 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
12472 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
12473 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
12474 Change check to determine if the instruction is a byte reversing
12475 entry. Fix typo in comment.
12476 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
12477 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
12478 Add def_builtin calls for new builtins.
12479 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
12480 Add define_insn expansion.
12482 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12484 * config/rl78/rl78.md (umaxdi3): New define_expand.
12486 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12488 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
12489 for non-QImode registers.
12491 2018-01-22 Richard Biener <rguenther@suse.de>
12493 PR tree-optimization/83963
12494 * graphite-scop-detection.c (scop_detection::get_sese): Delay
12495 including the loop exit block.
12496 (scop_detection::merge_sese): Likewise.
12497 (scop_detection::add_scop): Do it here instead.
12499 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12501 * doc/sourcebuild.texi (arm_softfloat): Document.
12503 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
12506 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
12507 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
12508 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
12510 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12511 David Edelsohn <dje.gcc@gmail.com>
12514 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12515 Change "crset eq" to "crset 2".
12516 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12517 (*call_indirect_aix<mode>_nospec): Likewise.
12518 (*call_value_indirect_aix<mode>_nospec): Likewise.
12519 (*call_indirect_elfv2<mode>_nospec): Likewise.
12520 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
12521 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
12522 change assembly output from . to $.
12523 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12524 (indirect_jump<mode>_nospec): Change assembly output from . to $.
12525 (*tablejump<mode>_internal1_nospec): Likewise.
12527 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
12530 * config/sh/sh_optimize_sett_clrt.cc:
12531 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
12533 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
12535 PR tree-optimization/83940
12536 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
12537 offset_dt to vect_constant_def rather than vect_unknown_def_type.
12538 (vect_check_load_store_mask): Add a mask_dt_out parameter and
12539 use it to pass back the definition type.
12540 (vect_check_store_rhs): Likewise rhs_dt_out.
12541 (vect_build_gather_load_calls): Add a mask_dt argument and use
12542 it instead of a call to vect_is_simple_use.
12543 (vectorizable_store): Update calls to vect_check_load_store_mask
12544 and vect_check_store_rhs. Use the dt returned by the latter instead
12545 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
12546 instead of calls to vect_is_simple_use. Pass the scalar rather
12547 than the vector operand to vect_is_simple_use when handling
12548 second and subsequent copies of an rhs value.
12549 (vectorizable_load): Update calls to vect_check_load_store_mask
12550 and vect_build_gather_load_calls. Use the cached mask_dt and
12551 gs_info.offset_dt instead of calls to vect_is_simple_use.
12553 2018-01-20 Jakub Jelinek <jakub@redhat.com>
12555 PR middle-end/83945
12556 * tree-emutls.c: Include gimplify.h.
12557 (lower_emutls_2): New function.
12558 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
12559 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
12560 it before further processing.
12563 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
12564 UINTVAL (trueop1) instead of INTVAL (op1).
12566 2018-01-19 Jakub Jelinek <jakub@redhat.com>
12570 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
12571 INCOMING_FRAME_SP_OFFSET if not defined.
12572 (scan_trace): Add ENTRY argument. If true and
12573 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
12574 emit a note to adjust the CFA offset.
12575 (create_cfi_notes): Adjust scan_trace callers.
12576 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
12577 INCOMING_FRAME_SP_OFFSET in the CIE.
12578 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
12579 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
12581 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
12582 * doc/tm.texi: Regenerated.
12584 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12586 PR rtl-optimization/83147
12587 * lra-constraints.c (remove_inheritance_pseudos): Use
12588 lra_substitute_pseudo_within_insn.
12590 2018-01-19 Tom de Vries <tom@codesourcery.com>
12591 Cesar Philippidis <cesar@codesourcery.com>
12594 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
12596 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
12599 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
12600 spaces for function labels.
12602 2018-01-19 Martin Liska <mliska@suse.cz>
12604 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
12605 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
12606 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
12607 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
12608 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
12609 (PRED_CONST_RETURN): Change from 69 to 65.
12610 (PRED_NULL_RETURN): Change from 91 to 71.
12611 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
12612 (PRED_LOOP_GUARD): Change from 66 to 73.
12614 2018-01-19 Martin Liska <mliska@suse.cz>
12616 * predict.c (predict_insn_def): Add new assert.
12617 (struct branch_predictor): Change type to signed integer.
12618 (test_prediction_value_range): Amend test to cover
12619 PROB_UNINITIALIZED.
12620 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
12621 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
12622 (PRED_LOOP_ITERATIONS_MAX): Likewise.
12623 (PRED_LOOP_IV_COMPARE): Likewise.
12624 * predict.h (PROB_UNINITIALIZED): Define new constant.
12626 2018-01-19 Martin Liska <mliska@suse.cz>
12628 * predict.c (dump_prediction): Add new format for
12629 analyze_brprob.py script which is enabled with -details
12631 * profile-count.h (precise_p): New function.
12633 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12635 PR tree-optimization/83922
12636 * tree-vect-loop.c (vect_verify_full_masking): Return false if
12637 there are no statements that need masking.
12638 (vect_active_double_reduction_p): New function.
12639 (vect_analyze_loop_operations): Use it when handling phis that
12640 are not in the loop header.
12642 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12644 PR tree-optimization/83914
12645 * tree-vect-loop.c (vectorizable_induction): Don't convert
12646 init_expr or apply the peeling adjustment for inductions
12647 that are nested within the vectorized loop.
12649 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12651 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
12654 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12658 * function.h (gimplify_parameters): Add gimple_seq * argument.
12659 * function.c: Include gimple.h and options.h.
12660 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
12661 for the added local temporaries if needed.
12662 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
12663 if there are any parameter cleanups, wrap whole body into a
12664 try/finally with the cleanups.
12666 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
12669 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
12670 Use GET_MODE_CLASS for scalar floating point.
12672 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12676 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
12677 Fix call of call_cgraph_insertion_hooks.
12679 2018-01-18 Martin Sebor <msebor@redhat.com>
12681 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
12683 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12686 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
12689 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
12692 * common.opt: (-ffile-prefix-map): New option.
12693 * opts.c (common_handle_option): Defer it.
12694 * opts-global.c (handle_common_deferred_options): Handle it.
12695 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
12696 * file-prefix-map.h: New file.
12697 (remap_debug_filename, add_debug_prefix_map): ...here.
12698 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
12699 * final.c (debug_prefix_map, add_debug_prefix_map
12700 remap_debug_filename): Move to...
12701 * file-prefix-map.c: New file.
12702 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
12703 generalize, get rid of alloca(), use strrchr() instead of strchr().
12704 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
12705 Implement in terms of add_prefix_map().
12706 (remap_macro_filename, remap_debug_filename): Implement in term of
12708 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
12709 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
12710 * dbxout.c: Include file-prefix-map.h.
12711 * varasm.c: Likewise.
12712 * vmsdbgout.c: Likewise.
12713 * xcoffout.c: Likewise.
12714 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
12715 * doc/cppopts.texi (-fmacro-prefix-map): Document.
12716 * doc/invoke.texi (-ffile-prefix-map): Document.
12717 (-fdebug-prefix-map): Update description.
12719 2018-01-18 Martin Liska <mliska@suse.cz>
12721 * config/i386/i386.c (indirect_thunk_name): Document that also
12723 (output_indirect_thunk): Document why both instructions
12724 (pause and lfence) are generated.
12726 2018-01-18 Richard Biener <rguenther@suse.de>
12728 PR tree-optimization/83887
12729 * graphite-scop-detection.c
12730 (scop_detection::get_nearest_dom_with_single_entry): Remove.
12731 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
12732 (scop_detection::merge_sese): Re-implement with a flood-fill
12733 algorithm that properly finds a SESE region if it exists.
12735 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12738 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
12739 pointer_diff optimizations use view_convert instead of convert.
12741 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12743 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12744 Generate different code for -mno-speculate-indirect-jumps.
12745 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12746 (*call_indirect_aix<mode>): Disable for
12747 -mno-speculate-indirect-jumps.
12748 (*call_indirect_aix<mode>_nospec): New define_insn.
12749 (*call_value_indirect_aix<mode>): Disable for
12750 -mno-speculate-indirect-jumps.
12751 (*call_value_indirect_aix<mode>_nospec): New define_insn.
12752 (*sibcall_nonlocal_sysv<mode>): Generate different code for
12753 -mno-speculate-indirect-jumps.
12754 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12756 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12758 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
12759 long double type, set the flags for noting the default long double
12760 type, even if we don't pass or return a long double type.
12762 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
12765 * ipa-inline.c (flatten_function): Do not overwrite final inlining
12768 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
12770 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
12771 support for merge[hl].
12772 (fold_mergehl_helper): New helper function.
12773 (tree-vector-builder.h): New #include for tree_vector_builder usage.
12774 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
12775 (altivec_vmrglw_direct): Add xxmrglw insn.
12777 2018-01-17 Andrew Waterman <andrew@sifive.com>
12779 * config/riscv/riscv.c (riscv_conditional_register_usage): If
12780 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
12782 2018-01-17 David Malcolm <dmalcolm@redhat.com>
12785 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
12786 call the lto_location_cache before reading the
12787 DECL_SOURCE_LOCATION of the types.
12789 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
12790 Richard Sandiford <richard.sandiford@linaro.org>
12792 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
12793 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
12794 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
12795 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
12796 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
12798 * config/aarch64/constraints.md (aarch64_movti_operand):
12800 * config/aarch64/predicates.md (Uti): Add new constraint.
12802 2018-01-17 Carl Love <cel@us.ibm.com>
12804 * config/rs6000/vsx.md (define_expand xl_len_r,
12805 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
12806 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
12808 (define_expand, define_insn): Move the shift left from the
12809 define_insn to the define_expand for lxvl and stxvl instructions.
12810 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
12811 and XL_LEN_R definitions to PURE.
12813 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
12815 * config/i386/i386.c (indirect_thunk_name): Declare regno
12816 as unsigned int. Compare regno with INVALID_REGNUM.
12817 (output_indirect_thunk): Ditto.
12818 (output_indirect_thunk_function): Ditto.
12819 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
12820 in the call to output_indirect_thunk_function.
12822 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
12824 PR middle-end/83884
12825 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
12826 rather than the size of inner_type to determine the stack slot size
12827 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
12829 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
12832 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
12835 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
12837 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
12838 endian Linux systems to optionally enable multilibs for selecting
12839 the long double type if the user configured an explicit type.
12840 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
12841 have no long double multilibs if not defined.
12842 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
12843 warn if the user used -mabi={ieee,ibm}longdouble and we built
12844 multilibs for long double.
12845 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
12846 appropriate multilib option.
12847 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
12849 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
12850 for building long double multilibs.
12851 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
12853 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
12855 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
12858 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
12860 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
12863 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
12866 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
12869 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12871 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
12872 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
12874 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
12876 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
12877 different rtl trees depending on TARGET_64BIT.
12878 (rs6000_gen_lvx): Likewise.
12880 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12882 * config/visium/visium.md (nop): Tweak comment.
12883 (hazard_nop): Likewise.
12885 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12887 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
12888 -mspeculate-indirect-jumps.
12889 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
12890 for -mno-speculate-indirect-jumps.
12891 (*call_indirect_elfv2<mode>_nospec): New define_insn.
12892 (*call_value_indirect_elfv2<mode>): Disable for
12893 -mno-speculate-indirect-jumps.
12894 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
12895 (indirect_jump): Emit different RTL for
12896 -mno-speculate-indirect-jumps.
12897 (*indirect_jump<mode>): Disable for
12898 -mno-speculate-indirect-jumps.
12899 (*indirect_jump<mode>_nospec): New define_insn.
12900 (tablejump): Emit different RTL for
12901 -mno-speculate-indirect-jumps.
12902 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
12903 (tablejumpsi_nospec): New define_expand.
12904 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
12905 (tablejumpdi_nospec): New define_expand.
12906 (*tablejump<mode>_internal1): Disable for
12907 -mno-speculate-indirect-jumps.
12908 (*tablejump<mode>_internal1_nospec): New define_insn.
12909 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
12912 2018-01-16 Artyom Skrobov tyomitch@gmail.com
12914 * caller-save.c (insert_save): Drop unnecessary parameter. All
12917 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12918 Richard Biener <rguenth@suse.de>
12921 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
12922 return early, inline manually is_gimple_sizepos. Make sure if we
12923 call gimplify_expr we don't end up with a gimple constant.
12924 * tree.c (variably_modified_type_p): Don't return true for
12925 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
12926 * gimplify.h (is_gimple_sizepos): Remove.
12928 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12930 PR tree-optimization/83857
12931 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
12932 vectorizable_live_operation for pure SLP statements.
12933 (vectorizable_live_operation): Handle PHIs.
12935 2018-01-16 Richard Biener <rguenther@suse.de>
12937 PR tree-optimization/83867
12938 * tree-vect-stmts.c (vect_transform_stmt): Precompute
12939 nested_in_vect_loop_p since the scalar stmt may get invalidated.
12941 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12944 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
12945 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
12946 If off is not INTEGER_CST, issue a may not be aligned warning
12947 rather than isn't aligned. Use isn%'t rather than isn't.
12948 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
12950 <case MULT_EXPR>: Improve the case when bottom and one of the
12951 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
12952 operand, in that case check if the other operand is multiple of
12953 bottom divided by the INTEGER_CST operand.
12955 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12958 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
12959 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
12960 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
12961 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
12962 * config/pa/pa.c (pa_function_arg_advance): Likewise.
12963 (pa_function_arg, pa_arg_partial_bytes): Likewise.
12964 (pa_function_arg_size): New function.
12966 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12968 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
12969 in a separate statement.
12971 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12973 PR tree-optimization/83847
12974 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
12975 group gathers and scatters.
12977 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12979 PR rtl-optimization/86620
12980 * params.def (max-sched-ready-insns): Bump minimum value to 1.
12982 PR rtl-optimization/83213
12983 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
12984 to last if both are JUMP_INSNs.
12986 PR tree-optimization/83843
12987 * gimple-ssa-store-merging.c
12988 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
12989 store_immediate_info for bswap/nop orig_stores.
12991 2018-01-15 Andrew Waterman <andrew@sifive.com>
12993 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
12995 <UDIV>: Increase cost if !TARGET_DIV.
12997 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
12999 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
13000 (define_attr "cr_logical_3op"): New.
13001 (cceq_ior_compare): Adjust.
13002 (cceq_ior_compare_complement): Adjust.
13003 (*cceq_rev_compare): Adjust.
13004 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
13005 (is_cracked_insn): Adjust.
13006 (insn_must_be_first_in_group): Adjust.
13007 * config/rs6000/40x.md: Adjust.
13008 * config/rs6000/440.md: Adjust.
13009 * config/rs6000/476.md: Adjust.
13010 * config/rs6000/601.md: Adjust.
13011 * config/rs6000/603.md: Adjust.
13012 * config/rs6000/6xx.md: Adjust.
13013 * config/rs6000/7450.md: Adjust.
13014 * config/rs6000/7xx.md: Adjust.
13015 * config/rs6000/8540.md: Adjust.
13016 * config/rs6000/cell.md: Adjust.
13017 * config/rs6000/e300c2c3.md: Adjust.
13018 * config/rs6000/e500mc.md: Adjust.
13019 * config/rs6000/e500mc64.md: Adjust.
13020 * config/rs6000/e5500.md: Adjust.
13021 * config/rs6000/e6500.md: Adjust.
13022 * config/rs6000/mpc.md: Adjust.
13023 * config/rs6000/power4.md: Adjust.
13024 * config/rs6000/power5.md: Adjust.
13025 * config/rs6000/power6.md: Adjust.
13026 * config/rs6000/power7.md: Adjust.
13027 * config/rs6000/power8.md: Adjust.
13028 * config/rs6000/power9.md: Adjust.
13029 * config/rs6000/rs64.md: Adjust.
13030 * config/rs6000/titan.md: Adjust.
13032 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13034 * config/i386/predicates.md (indirect_branch_operand): Rewrite
13035 ix86_indirect_branch_register logic.
13037 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13039 * config/i386/constraints.md (Bs): Update
13040 ix86_indirect_branch_register check. Don't check
13041 ix86_indirect_branch_register with GOT_memory_operand.
13043 * config/i386/predicates.md (GOT_memory_operand): Don't check
13044 ix86_indirect_branch_register here.
13045 (GOT32_symbol_operand): Likewise.
13047 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13049 * config/i386/predicates.md (constant_call_address_operand):
13050 Rewrite ix86_indirect_branch_register logic.
13051 (sibcall_insn_operand): Likewise.
13053 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13055 * config/i386/constraints.md (Bs): Replace
13056 ix86_indirect_branch_thunk_register with
13057 ix86_indirect_branch_register.
13059 * config/i386/i386.md (indirect_jump): Likewise.
13060 (tablejump): Likewise.
13061 (*sibcall_memory): Likewise.
13062 (*sibcall_value_memory): Likewise.
13063 Peepholes of indirect call and jump via memory: Likewise.
13064 * config/i386/i386.opt: Likewise.
13065 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13066 (GOT_memory_operand): Likewise.
13067 (call_insn_operand): Likewise.
13068 (sibcall_insn_operand): Likewise.
13069 (GOT32_symbol_operand): Likewise.
13071 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13073 PR middle-end/83837
13074 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
13075 type rather than type addr's type points to.
13076 (expand_omp_atomic_mutex): Likewise.
13077 (expand_omp_atomic): Likewise.
13079 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13082 * config/i386/i386.c (output_indirect_thunk_function): Use
13083 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
13084 for __x86_return_thunk.
13086 2018-01-15 Richard Biener <rguenther@suse.de>
13088 PR middle-end/83850
13089 * expmed.c (extract_bit_field_1): Fix typo.
13091 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13094 * config/arm/iterators.md (VF): New mode iterator.
13095 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
13096 Remove integer-related logic from pattern.
13097 (neon_vabd<mode>_3): Likewise.
13099 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13101 PR middle-end/82694
13102 * common.opt (fstrict-overflow): No longer an alias.
13103 (fwrapv-pointer): New option.
13104 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
13105 also for pointer types based on flag_wrapv_pointer.
13106 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
13107 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
13108 opts->x_flag_wrapv got set.
13109 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
13110 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
13111 POINTER_TYPE_OVERFLOW_UNDEFINED.
13112 * match.pd: Likewise in address comparison pattern.
13113 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
13115 2018-01-15 Richard Biener <rguenther@suse.de>
13118 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
13119 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
13120 Reset type names to their identifier if their TYPE_DECL doesn't
13121 have linkage (and thus is used for ODR and devirt).
13122 (save_debug_info_for_decl): Remove.
13123 (save_debug_info_for_type): Likewise.
13124 (add_tree_to_fld_list): Adjust.
13125 * tree-pretty-print.c (dump_generic_node): Make dumping of
13126 type names more robust.
13128 2018-01-15 Richard Biener <rguenther@suse.de>
13130 * BASE-VER: Bump to 8.0.1.
13132 2018-01-14 Martin Sebor <msebor@redhat.com>
13135 * builtins.c (check_access): Avoid warning when the no-warning bit
13138 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
13140 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
13141 * ira-color (allocno_hard_regs_compare): Likewise.
13143 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
13146 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
13147 Use .pushsection/.popsection.
13149 2018-01-14 Martin Sebor <msebor@redhat.com>
13152 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
13154 2018-01-14 Jakub Jelinek <jakub@redhat.com>
13156 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
13157 entry from extra_headers.
13158 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
13159 extra_headers, make the list bitwise identical to the i?86-*-* one.
13161 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13163 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
13164 -mcmodel=large with -mindirect-branch=thunk,
13165 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
13166 -mfunction-return=thunk-extern.
13167 * doc/invoke.texi: Document -mcmodel=large is incompatible with
13168 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
13169 -mfunction-return=thunk and -mfunction-return=thunk-extern.
13171 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13173 * config/i386/i386.c (print_reg): Print the name of the full
13174 integer register without '%'.
13175 (ix86_print_operand): Handle 'V'.
13176 * doc/extend.texi: Document 'V' modifier.
13178 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13180 * config/i386/constraints.md (Bs): Disallow memory operand for
13181 -mindirect-branch-register.
13183 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13184 (GOT_memory_operand): Likewise.
13185 (call_insn_operand): Likewise.
13186 (sibcall_insn_operand): Likewise.
13187 (GOT32_symbol_operand): Likewise.
13188 * config/i386/i386.md (indirect_jump): Call convert_memory_address
13189 for -mindirect-branch-register.
13190 (tablejump): Likewise.
13191 (*sibcall_memory): Likewise.
13192 (*sibcall_value_memory): Likewise.
13193 Disallow peepholes of indirect call and jump via memory for
13194 -mindirect-branch-register.
13195 (*call_pop): Replace m with Bw.
13196 (*call_value_pop): Likewise.
13197 (*sibcall_pop_memory): Replace m with Bs.
13198 * config/i386/i386.opt (mindirect-branch-register): New option.
13199 * doc/invoke.texi: Document -mindirect-branch-register option.
13201 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13203 * config/i386/i386-protos.h (ix86_output_function_return): New.
13204 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
13205 set function_return_type.
13206 (indirect_thunk_name): Add ret_p to indicate thunk for function
13208 (output_indirect_thunk_function): Pass false to
13209 indirect_thunk_name.
13210 (ix86_output_indirect_branch_via_reg): Likewise.
13211 (ix86_output_indirect_branch_via_push): Likewise.
13212 (output_indirect_thunk_function): Create alias for function
13213 return thunk if regno < 0.
13214 (ix86_output_function_return): New function.
13215 (ix86_handle_fndecl_attribute): Handle function_return.
13216 (ix86_attribute_table): Add function_return.
13217 * config/i386/i386.h (machine_function): Add
13218 function_return_type.
13219 * config/i386/i386.md (simple_return_internal): Use
13220 ix86_output_function_return.
13221 (simple_return_internal_long): Likewise.
13222 * config/i386/i386.opt (mfunction-return=): New option.
13223 (indirect_branch): Mention -mfunction-return=.
13224 * doc/extend.texi: Document function_return function attribute.
13225 * doc/invoke.texi: Document -mfunction-return= option.
13227 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13229 * config/i386/i386-opts.h (indirect_branch): New.
13230 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
13231 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
13232 with local indirect jump when converting indirect call and jump.
13233 (ix86_set_indirect_branch_type): New.
13234 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
13235 (indirectlabelno): New.
13236 (indirect_thunk_needed): Likewise.
13237 (indirect_thunk_bnd_needed): Likewise.
13238 (indirect_thunks_used): Likewise.
13239 (indirect_thunks_bnd_used): Likewise.
13240 (INDIRECT_LABEL): Likewise.
13241 (indirect_thunk_name): Likewise.
13242 (output_indirect_thunk): Likewise.
13243 (output_indirect_thunk_function): Likewise.
13244 (ix86_output_indirect_branch_via_reg): Likewise.
13245 (ix86_output_indirect_branch_via_push): Likewise.
13246 (ix86_output_indirect_branch): Likewise.
13247 (ix86_output_indirect_jmp): Likewise.
13248 (ix86_code_end): Call output_indirect_thunk_function if needed.
13249 (ix86_output_call_insn): Call ix86_output_indirect_branch if
13251 (ix86_handle_fndecl_attribute): Handle indirect_branch.
13252 (ix86_attribute_table): Add indirect_branch.
13253 * config/i386/i386.h (machine_function): Add indirect_branch_type
13254 and has_local_indirect_jump.
13255 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
13257 (tablejump): Likewise.
13258 (*indirect_jump): Use ix86_output_indirect_jmp.
13259 (*tablejump_1): Likewise.
13260 (simple_return_indirect_internal): Likewise.
13261 * config/i386/i386.opt (mindirect-branch=): New option.
13262 (indirect_branch): New.
13265 (thunk-inline): Likewise.
13266 (thunk-extern): Likewise.
13267 * doc/extend.texi: Document indirect_branch function attribute.
13268 * doc/invoke.texi: Document -mindirect-branch= option.
13270 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
13273 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
13275 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
13277 * ipa-inline.c (want_inline_small_function_p): Return false if
13278 inlining has already failed with CIF_FINAL_ERROR.
13279 (update_caller_keys): Call want_inline_small_function_p before
13281 (update_callee_keys): Likewise.
13283 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
13285 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
13287 (rs6000_quadword_masked_address_p): Likewise.
13288 (quad_aligned_load_p): Likewise.
13289 (quad_aligned_store_p): Likewise.
13290 (const_load_sequence_p): Add comment to describe the outer-most loop.
13291 (mimic_memory_attributes_and_flags): New function.
13292 (rs6000_gen_stvx): Likewise.
13293 (replace_swapped_aligned_store): Likewise.
13294 (rs6000_gen_lvx): Likewise.
13295 (replace_swapped_aligned_load): Likewise.
13296 (replace_swapped_load_constant): Capitalize argument name in
13297 comment describing this function.
13298 (rs6000_analyze_swaps): Add a third pass to search for vector loads
13299 and stores that access quad-word aligned addresses and replace
13300 with stvx or lvx instructions when appropriate.
13301 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
13302 New function prototype.
13303 (rs6000_quadword_masked_address_p): Likewise.
13304 (rs6000_gen_lvx): Likewise.
13305 (rs6000_gen_stvx): Likewise.
13306 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
13307 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
13308 when memory address is aligned.
13309 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
13310 this split to select lvx instruction when memory address is aligned.
13311 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
13312 instruction when memory address is aligned.
13313 (*vsx_le_perm_load_v16qi): Likewise.
13314 (four unnamed splitters): Modify to select the stvx instruction
13315 when memory is aligned.
13317 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
13319 * predict.c (determine_unlikely_bbs): Handle correctly BBs
13320 which appears in the queue multiple times.
13322 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13323 Alan Hayward <alan.hayward@arm.com>
13324 David Sherwood <david.sherwood@arm.com>
13326 * tree-vectorizer.h (vec_lower_bound): New structure.
13327 (_loop_vec_info): Add check_nonzero and lower_bounds.
13328 (LOOP_VINFO_CHECK_NONZERO): New macro.
13329 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
13330 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
13331 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
13332 fields. Make seg_len the distance travelled, not including the
13334 (dr_direction_indicator): Declare.
13335 (dr_zero_step_indicator): Likewise.
13336 (dr_known_forward_stride_p): Likewise.
13337 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
13339 (runtime_alias_check_p): Allow runtime alias checks with
13341 (operator ==): Compare access_size and align.
13342 (prune_runtime_alias_test_list): Rework for new distinction between
13343 the access_size and seg_len.
13344 (create_intersect_range_checks_index): Likewise. Cope with polynomial
13346 (get_segment_min_max): New function.
13347 (create_intersect_range_checks): Use it.
13348 (dr_step_indicator): New function.
13349 (dr_direction_indicator): Likewise.
13350 (dr_zero_step_indicator): Likewise.
13351 (dr_known_forward_stride_p): Likewise.
13352 * tree-loop-distribution.c (data_ref_segment_size): Return
13353 DR_STEP * (niters - 1).
13354 (compute_alias_check_pairs): Update call to the dr_with_seg_len
13356 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
13357 (vect_preserves_scalar_order_p): New function, split out from...
13358 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
13359 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
13360 (vect_vfa_access_size): New function.
13361 (vect_vfa_align): Likewise.
13362 (vect_compile_time_alias): Take access_size_a and access_b arguments.
13363 (dump_lower_bound): New function.
13364 (vect_check_lower_bound): Likewise.
13365 (vect_small_gap_p): Likewise.
13366 (vectorizable_with_step_bound_p): Likewise.
13367 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
13368 depencies if the vectorization factor is 1. Convert the checks
13369 for nonzero steps into checks on the bounds of DR_STEP. Try using
13370 a bunds check for variable steps if the minimum required step is
13371 relatively small. Update calls to the dr_with_seg_len
13372 constructor and to vect_compile_time_alias.
13373 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
13375 (vect_loop_versioning): Call it.
13376 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
13378 (vect_estimate_min_profitable_iters): Account for any bounds checks.
13380 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13381 Alan Hayward <alan.hayward@arm.com>
13382 David Sherwood <david.sherwood@arm.com>
13384 * doc/sourcebuild.texi (vect_scatter_store): Document.
13385 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
13387 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
13389 * genopinit.c (main): Add supports_vec_scatter_store and
13390 supports_vec_scatter_store_cached to target_optabs.
13391 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
13392 IFN_MASK_SCATTER_STORE.
13393 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
13395 * internal-fn.h (internal_store_fn_p): Declare.
13396 (internal_fn_stored_value_index): Likewise.
13397 * internal-fn.c (scatter_store_direct): New macro.
13398 (expand_scatter_store_optab_fn): New function.
13399 (direct_scatter_store_optab_supported_p): New macro.
13400 (internal_store_fn_p): New function.
13401 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
13402 IFN_MASK_SCATTER_STORE.
13403 (internal_fn_mask_index): Likewise.
13404 (internal_fn_stored_value_index): New function.
13405 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
13406 for scatter stores.
13407 * optabs-query.h (supports_vec_scatter_store_p): Declare.
13408 * optabs-query.c (supports_vec_scatter_store_p): New function.
13409 * tree-vectorizer.h (vect_get_store_rhs): Declare.
13410 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
13411 true for scatter stores.
13412 (vect_gather_scatter_fn_p): Handle scatter stores too.
13413 (vect_check_gather_scatter): Consider using scatter stores if
13414 supports_vec_scatter_store_p.
13415 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
13416 scatter stores too.
13417 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13418 internal_fn_stored_value_index.
13419 (check_load_store_masking): Handle scatter stores too.
13420 (vect_get_store_rhs): Make public.
13421 (vectorizable_call): Use internal_store_fn_p.
13422 (vectorizable_store): Handle scatter store internal functions.
13423 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
13424 when deciding whether the end of the group has been reached.
13425 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
13426 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
13427 (mask_scatter_store<mode>): New insns.
13429 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13430 Alan Hayward <alan.hayward@arm.com>
13431 David Sherwood <david.sherwood@arm.com>
13433 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
13434 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
13435 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
13437 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
13438 Use vect_truncate_gather_scatter_offset if we can't treat the
13439 operation as a normal gather load or scatter store.
13440 (get_group_load_store_type): Take the gather_scatter_info
13441 as argument. Try using a gather load or scatter store for
13442 single-element groups.
13443 (get_load_store_type): Update calls to get_group_load_store_type
13444 and vect_use_strided_gather_scatters_p.
13446 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13447 Alan Hayward <alan.hayward@arm.com>
13448 David Sherwood <david.sherwood@arm.com>
13450 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
13451 optional tree argument.
13452 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
13454 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
13455 but continue to use the current value as a fallback.
13456 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
13457 to compare the updates.
13458 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
13459 (get_load_store_type): Use it when handling a strided access.
13460 (vect_get_strided_load_store_ops): New function.
13461 (vect_get_data_ptr_increment): Likewise.
13462 (vectorizable_load): Handle strided gather loads. Always pass
13463 a step to vect_create_data_ref_ptr and bump_vector_ptr.
13465 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13466 Alan Hayward <alan.hayward@arm.com>
13467 David Sherwood <david.sherwood@arm.com>
13469 * doc/md.texi (gather_load@var{m}): Document.
13470 (mask_gather_load@var{m}): Likewise.
13471 * genopinit.c (main): Add supports_vec_gather_load and
13472 supports_vec_gather_load_cached to target_optabs.
13473 * optabs-tree.c (init_tree_optimization_optabs): Use
13474 ggc_cleared_alloc to allocate target_optabs.
13475 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
13476 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
13478 * internal-fn.h (internal_load_fn_p): Declare.
13479 (internal_gather_scatter_fn_p): Likewise.
13480 (internal_fn_mask_index): Likewise.
13481 (internal_gather_scatter_fn_supported_p): Likewise.
13482 * internal-fn.c (gather_load_direct): New macro.
13483 (expand_gather_load_optab_fn): New function.
13484 (direct_gather_load_optab_supported_p): New macro.
13485 (direct_internal_fn_optab): New function.
13486 (internal_load_fn_p): Likewise.
13487 (internal_gather_scatter_fn_p): Likewise.
13488 (internal_fn_mask_index): Likewise.
13489 (internal_gather_scatter_fn_supported_p): Likewise.
13490 * optabs-query.c (supports_at_least_one_mode_p): New function.
13491 (supports_vec_gather_load_p): Likewise.
13492 * optabs-query.h (supports_vec_gather_load_p): Declare.
13493 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
13494 and memory_type field.
13495 (NUM_PATTERNS): Bump to 15.
13496 * tree-vect-data-refs.c: Include internal-fn.h.
13497 (vect_gather_scatter_fn_p): New function.
13498 (vect_describe_gather_scatter_call): Likewise.
13499 (vect_check_gather_scatter): Try using internal functions for
13500 gather loads. Recognize existing calls to a gather load function.
13501 (vect_analyze_data_refs): Consider using gather loads if
13502 supports_vec_gather_load_p.
13503 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
13504 (vect_get_gather_scatter_offset_type): Likewise.
13505 (vect_convert_mask_for_vectype): Likewise.
13506 (vect_add_conversion_to_patterm): Likewise.
13507 (vect_try_gather_scatter_pattern): Likewise.
13508 (vect_recog_gather_scatter_pattern): New pattern recognizer.
13509 (vect_vect_recog_func_ptrs): Add it.
13510 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13511 internal_fn_mask_index and internal_gather_scatter_fn_p.
13512 (check_load_store_masking): Take the gather_scatter_info as an
13513 argument and handle gather loads.
13514 (vect_get_gather_scatter_ops): New function.
13515 (vectorizable_call): Check internal_load_fn_p.
13516 (vectorizable_load): Likewise. Handle gather load internal
13518 (vectorizable_store): Update call to check_load_store_masking.
13519 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
13520 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
13521 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
13522 (aarch64_gather_scale_operand_d): New predicates.
13523 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
13524 (mask_gather_load<mode>): New insns.
13526 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13527 Alan Hayward <alan.hayward@arm.com>
13528 David Sherwood <david.sherwood@arm.com>
13530 * optabs.def (fold_left_plus_optab): New optab.
13531 * doc/md.texi (fold_left_plus_@var{m}): Document.
13532 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
13533 * internal-fn.c (fold_left_direct): Define.
13534 (expand_fold_left_optab_fn): Likewise.
13535 (direct_fold_left_optab_supported_p): Likewise.
13536 * fold-const-call.c (fold_const_fold_left): New function.
13537 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
13538 * tree-parloops.c (valid_reduction_p): New function.
13539 (gather_scalar_reductions): Use it.
13540 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
13541 (vect_finish_replace_stmt): Declare.
13542 * tree-vect-loop.c (fold_left_reduction_fn): New function.
13543 (needs_fold_left_reduction_p): New function, split out from...
13544 (vect_is_simple_reduction): ...here. Accept reductions that
13545 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
13546 (vect_force_simple_reduction): Also store the reduction type in
13547 the assignment's STMT_VINFO_REDUC_TYPE.
13548 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
13549 (merge_with_identity): New function.
13550 (vect_expand_fold_left): Likewise.
13551 (vectorize_fold_left_reduction): Likewise.
13552 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
13553 scalar phi in place for it. Check for target support and reject
13554 cases that would reassociate the operation. Defer the transform
13555 phase to vectorize_fold_left_reduction.
13556 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
13557 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
13558 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
13560 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13562 * tree-if-conv.c (predicate_mem_writes): Remove redundant
13563 call to ifc_temp_var.
13565 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13566 Alan Hayward <alan.hayward@arm.com>
13567 David Sherwood <david.sherwood@arm.com>
13569 * target.def (legitimize_address_displacement): Take the original
13570 offset as a poly_int.
13571 * targhooks.h (default_legitimize_address_displacement): Update
13573 * targhooks.c (default_legitimize_address_displacement): Likewise.
13574 * doc/tm.texi: Regenerate.
13575 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
13576 as an argument, moving assert of ad->disp == ad->disp_term to...
13577 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
13578 Try calling targetm.legitimize_address_displacement before expanding
13579 the address rather than afterwards, and adjust for the new interface.
13580 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
13581 Match the new hook interface. Handle SVE addresses.
13582 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
13583 new hook interface.
13585 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13587 * Makefile.in (OBJS): Add early-remat.o.
13588 * target.def (select_early_remat_modes): New hook.
13589 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
13590 * doc/tm.texi: Regenerate.
13591 * targhooks.h (default_select_early_remat_modes): Declare.
13592 * targhooks.c (default_select_early_remat_modes): New function.
13593 * timevar.def (TV_EARLY_REMAT): New timevar.
13594 * passes.def (pass_early_remat): New pass.
13595 * tree-pass.h (make_pass_early_remat): Declare.
13596 * early-remat.c: New file.
13597 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
13599 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
13601 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13602 Alan Hayward <alan.hayward@arm.com>
13603 David Sherwood <david.sherwood@arm.com>
13605 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
13606 vfm1 with a bound_epilog parameter.
13607 (vect_do_peeling): Update calls accordingly, and move the prologue
13608 call earlier in the function. Treat the base bound_epilog as 0 for
13609 fully-masked loops and retain vf - 1 for other loops. Add 1 to
13610 this base when peeling for gaps.
13611 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
13612 with fully-masked loops.
13613 (vect_estimate_min_profitable_iters): Handle the single peeled
13614 iteration in that case.
13616 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13617 Alan Hayward <alan.hayward@arm.com>
13618 David Sherwood <david.sherwood@arm.com>
13620 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
13621 single-element interleaving even if the size is not a power of 2.
13622 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
13623 accesses for single-element interleaving if the group size is
13626 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13627 Alan Hayward <alan.hayward@arm.com>
13628 David Sherwood <david.sherwood@arm.com>
13630 * doc/md.texi (fold_extract_last_@var{m}): Document.
13631 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
13632 * optabs.def (fold_extract_last_optab): New optab.
13633 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
13634 * internal-fn.c (fold_extract_direct): New macro.
13635 (expand_fold_extract_optab_fn): Likewise.
13636 (direct_fold_extract_optab_supported_p): Likewise.
13637 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
13638 * tree-vect-loop.c (vect_model_reduction_cost): Handle
13639 EXTRACT_LAST_REDUCTION.
13640 (get_initial_def_for_reduction): Do not create an initial vector
13641 for EXTRACT_LAST_REDUCTION reductions.
13642 (vectorizable_reduction): Leave the scalar phi in place for
13643 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
13644 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
13645 epilogue code for EXTRACT_LAST_REDUCTION and defer the
13646 transform phase to vectorizable_condition.
13647 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
13649 (vect_finish_stmt_generation): ...here.
13650 (vect_finish_replace_stmt): New function.
13651 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
13652 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
13654 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
13656 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13657 Alan Hayward <alan.hayward@arm.com>
13658 David Sherwood <david.sherwood@arm.com>
13660 * doc/md.texi (extract_last_@var{m}): Document.
13661 * optabs.def (extract_last_optab): New optab.
13662 * internal-fn.def (EXTRACT_LAST): New internal function.
13663 * internal-fn.c (cond_unary_direct): New macro.
13664 (expand_cond_unary_optab_fn): Likewise.
13665 (direct_cond_unary_optab_supported_p): Likewise.
13666 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
13667 loops using EXTRACT_LAST.
13668 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
13669 (extract_last_<mode>): ...this optab.
13670 (vec_extract<mode><Vel>): Update accordingly.
13672 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13673 Alan Hayward <alan.hayward@arm.com>
13674 David Sherwood <david.sherwood@arm.com>
13676 * target.def (empty_mask_is_expensive): New hook.
13677 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
13678 * doc/tm.texi: Regenerate.
13679 * targhooks.h (default_empty_mask_is_expensive): Declare.
13680 * targhooks.c (default_empty_mask_is_expensive): New function.
13681 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
13682 if the target says that empty masks are expensive.
13683 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
13685 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
13687 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13688 Alan Hayward <alan.hayward@arm.com>
13689 David Sherwood <david.sherwood@arm.com>
13691 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
13692 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
13693 (vect_use_loop_mask_for_alignment_p): New function.
13694 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
13695 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
13696 niters_skip argument. Make sure that the first niters_skip elements
13697 of the first iteration are inactive.
13698 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
13699 Update call to vect_set_loop_masks_directly.
13700 (get_misalign_in_elems): New function, split out from...
13701 (vect_gen_prolog_loop_niters): ...here.
13702 (vect_update_init_of_dr): Take a code argument that specifies whether
13703 the adjustment should be added or subtracted.
13704 (vect_update_init_of_drs): Likewise.
13705 (vect_prepare_for_masked_peels): New function.
13706 (vect_do_peeling): Skip prologue peeling if we're using a mask
13707 instead. Update call to vect_update_inits_of_drs.
13708 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13710 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
13711 alignment. Do not include the number of peeled iterations in
13712 the minimum threshold in that case.
13713 (vectorizable_induction): Adjust the start value down by
13714 LOOP_VINFO_MASK_SKIP_NITERS iterations.
13715 (vect_transform_loop): Call vect_prepare_for_masked_peels.
13716 Take the number of skipped iterations into account when calculating
13718 * tree-vect-stmts.c (vect_gen_while_not): New function.
13720 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13721 Alan Hayward <alan.hayward@arm.com>
13722 David Sherwood <david.sherwood@arm.com>
13724 * doc/sourcebuild.texi (vect_fully_masked): Document.
13725 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
13726 default value to 0.
13727 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
13729 (vect_analyze_loop_2): ...here. Don't check the vectorization
13730 factor against the number of loop iterations if the loop is
13733 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13734 Alan Hayward <alan.hayward@arm.com>
13735 David Sherwood <david.sherwood@arm.com>
13737 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
13738 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
13739 (dump_groups): Update accordingly.
13740 (iv_use::mem_type): New member variable.
13741 (address_p): New function.
13742 (record_use): Add a mem_type argument and initialize the new
13744 (record_group_use): Add a mem_type argument. Use address_p.
13745 Remove obsolete null checks of base_object. Update call to record_use.
13746 (find_interesting_uses_op): Update call to record_group_use.
13747 (find_interesting_uses_cond): Likewise.
13748 (find_interesting_uses_address): Likewise.
13749 (get_mem_type_for_internal_fn): New function.
13750 (find_address_like_use): Likewise.
13751 (find_interesting_uses_stmt): Try find_address_like_use before
13752 calling find_interesting_uses_op.
13753 (addr_offset_valid_p): Use the iv mem_type field as the type
13754 of the addressed memory.
13755 (add_autoinc_candidates): Likewise.
13756 (get_address_cost): Likewise.
13757 (split_small_address_groups_p): Use address_p.
13758 (split_address_groups): Likewise.
13759 (add_iv_candidate_for_use): Likewise.
13760 (autoinc_possible_for_pair): Likewise.
13761 (rewrite_groups): Likewise.
13762 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
13763 (determine_group_iv_cost): Update after split of USE_ADDRESS.
13764 (get_alias_ptr_type_for_ptr_address): New function.
13765 (rewrite_use_address): Rewrite address uses in calls that were
13766 identified by find_address_like_use.
13768 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13769 Alan Hayward <alan.hayward@arm.com>
13770 David Sherwood <david.sherwood@arm.com>
13772 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
13774 * gimple-expr.h (is_gimple_addressable: Likewise.
13775 * gimple-expr.c (is_gimple_address): Likewise.
13776 * internal-fn.c (expand_call_mem_ref): New function.
13777 (expand_mask_load_optab_fn): Use it.
13778 (expand_mask_store_optab_fn): Likewise.
13780 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13781 Alan Hayward <alan.hayward@arm.com>
13782 David Sherwood <david.sherwood@arm.com>
13784 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
13785 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
13786 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
13787 (cond_umax@var{mode}): Document.
13788 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
13789 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
13790 (cond_umin_optab, cond_umax_optab): New optabs.
13791 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
13792 (COND_IOR, COND_XOR): New internal functions.
13793 * internal-fn.h (get_conditional_internal_fn): Declare.
13794 * internal-fn.c (cond_binary_direct): New macro.
13795 (expand_cond_binary_optab_fn): Likewise.
13796 (direct_cond_binary_optab_supported_p): Likewise.
13797 (get_conditional_internal_fn): New function.
13798 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
13799 Cope with reduction statements that are vectorized as calls rather
13801 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
13802 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
13803 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
13804 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13805 (UNSPEC_COND_EOR): New unspecs.
13806 (optab): Add mappings for them.
13807 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
13808 (sve_int_op, sve_fp_op): New int attributes.
13810 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13811 Alan Hayward <alan.hayward@arm.com>
13812 David Sherwood <david.sherwood@arm.com>
13814 * optabs.def (while_ult_optab): New optab.
13815 * doc/md.texi (while_ult@var{m}@var{n}): Document.
13816 * internal-fn.def (WHILE_ULT): New internal function.
13817 * internal-fn.h (direct_internal_fn_supported_p): New override
13818 that takes two types as argument.
13819 * internal-fn.c (while_direct): New macro.
13820 (expand_while_optab_fn): New function.
13821 (convert_optab_supported_p): Likewise.
13822 (direct_while_optab_supported_p): New macro.
13823 * wide-int.h (wi::udiv_ceil): New function.
13824 * tree-vectorizer.h (rgroup_masks): New structure.
13825 (vec_loop_masks): New typedef.
13826 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
13827 and fully_masked_p.
13828 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
13829 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
13830 (vect_max_vf): New function.
13831 (slpeel_make_loop_iterate_ntimes): Delete.
13832 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
13833 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
13834 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
13835 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
13836 internal-fn.h, stor-layout.h and optabs-query.h.
13837 (vect_set_loop_mask): New function.
13838 (add_preheader_seq): Likewise.
13839 (add_header_seq): Likewise.
13840 (interleave_supported_p): Likewise.
13841 (vect_maybe_permute_loop_masks): Likewise.
13842 (vect_set_loop_masks_directly): Likewise.
13843 (vect_set_loop_condition_masked): Likewise.
13844 (vect_set_loop_condition_unmasked): New function, split out from
13845 slpeel_make_loop_iterate_ntimes.
13846 (slpeel_make_loop_iterate_ntimes): Rename to..
13847 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
13848 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
13849 (vect_do_peeling): Update call accordingly.
13850 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
13852 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13853 mask_compare_type, can_fully_mask_p and fully_masked_p.
13854 (release_vec_loop_masks): New function.
13855 (_loop_vec_info): Use it to free the loop masks.
13856 (can_produce_all_loop_masks_p): New function.
13857 (vect_get_max_nscalars_per_iter): Likewise.
13858 (vect_verify_full_masking): Likewise.
13859 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
13860 retries, and free the mask rgroups before retrying. Check loop-wide
13861 reasons for disallowing fully-masked loops. Make the final decision
13862 about whether use a fully-masked loop or not.
13863 (vect_estimate_min_profitable_iters): Do not assume that peeling
13864 for the number of iterations will be needed for fully-masked loops.
13865 (vectorizable_reduction): Disable fully-masked loops.
13866 (vectorizable_live_operation): Likewise.
13867 (vect_halve_mask_nunits): New function.
13868 (vect_double_mask_nunits): Likewise.
13869 (vect_record_loop_mask): Likewise.
13870 (vect_get_loop_mask): Likewise.
13871 (vect_transform_loop): Handle the case in which the final loop
13872 iteration might handle a partial vector. Call vect_set_loop_condition
13873 instead of slpeel_make_loop_iterate_ntimes.
13874 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
13875 (check_load_store_masking): New function.
13876 (prepare_load_store_mask): Likewise.
13877 (vectorizable_store): Handle fully-masked loops.
13878 (vectorizable_load): Likewise.
13879 (supportable_widening_operation): Use vect_halve_mask_nunits for
13881 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
13882 (vect_gen_while): New function.
13883 * config/aarch64/aarch64.md (umax<mode>3): New expander.
13884 (aarch64_uqdec<mode>): New insn.
13886 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13887 Alan Hayward <alan.hayward@arm.com>
13888 David Sherwood <david.sherwood@arm.com>
13890 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
13891 (reduc_xor_scal_optab): New optabs.
13892 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
13893 (reduc_xor_scal_@var{m}): Document.
13894 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
13895 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
13896 internal functions.
13897 * fold-const-call.c (fold_const_call): Handle them.
13898 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
13899 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
13900 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
13901 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
13902 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
13903 (UNSPEC_XORV): New unspecs.
13904 (optab): Add entries for them.
13905 (BITWISEV): New int iterator.
13906 (bit_reduc_op): New int attributes.
13908 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13909 Alan Hayward <alan.hayward@arm.com>
13910 David Sherwood <david.sherwood@arm.com>
13912 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
13913 * internal-fn.def (VEC_SHL_INSERT): New internal function.
13914 * optabs.def (vec_shl_insert_optab): New optab.
13915 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
13916 (duplicate_and_interleave): Likewise.
13917 * tree-vect-loop.c: Include internal-fn.h.
13918 (neutral_op_for_slp_reduction): New function, split out from
13919 get_initial_defs_for_reduction.
13920 (get_initial_def_for_reduction): Handle option 2 for variable-length
13921 vectors by loading the neutral value into a vector and then shifting
13922 the initial value into element 0.
13923 (get_initial_defs_for_reduction): Replace the code argument with
13924 the neutral value calculated by neutral_op_for_slp_reduction.
13925 Use gimple_build_vector for constant-length vectors.
13926 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
13927 but the first group_size elements have a neutral value.
13928 Use duplicate_and_interleave otherwise.
13929 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
13930 Update call to get_initial_defs_for_reduction. Handle SLP
13931 reductions for variable-length vectors by creating one vector
13932 result for each scalar result, with the elements associated
13933 with other scalar results stubbed out with the neutral value.
13934 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
13935 Require IFN_VEC_SHL_INSERT for double reductions on
13936 variable-length vectors, or SLP reductions that have
13937 a neutral value. Require can_duplicate_and_interleave_p
13938 support for variable-length unchained SLP reductions if there
13939 is no neutral value, such as for MIN/MAX reductions. Also require
13940 the number of vector elements to be a multiple of the number of
13941 SLP statements when doing variable-length unchained SLP reductions.
13942 Update call to vect_create_epilog_for_reduction.
13943 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
13944 and remove initial values.
13945 (duplicate_and_interleave): Make public.
13946 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
13947 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
13949 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13950 Alan Hayward <alan.hayward@arm.com>
13951 David Sherwood <david.sherwood@arm.com>
13953 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
13954 (can_duplicate_and_interleave_p): New function.
13955 (vect_get_and_check_slp_defs): Take the vector of statements
13956 rather than just the current one. Remove excess parentheses.
13957 Restriction rejectinon of vect_constant_def and vect_external_def
13958 for variable-length vectors to boolean types, or types for which
13959 can_duplicate_and_interleave_p is false.
13960 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
13961 (duplicate_and_interleave): New function.
13962 (vect_get_constant_vectors): Use gimple_build_vector for
13963 constant-length vectors and suitable variable-length constant
13964 vectors. Use duplicate_and_interleave for other variable-length
13965 vectors. Don't defer the update when inserting new statements.
13967 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13968 Alan Hayward <alan.hayward@arm.com>
13969 David Sherwood <david.sherwood@arm.com>
13971 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
13972 min_profitable_iters doesn't go negative.
13974 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13975 Alan Hayward <alan.hayward@arm.com>
13976 David Sherwood <david.sherwood@arm.com>
13978 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
13979 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
13980 * optabs.def (vec_mask_load_lanes_optab): New optab.
13981 (vec_mask_store_lanes_optab): Likewise.
13982 * internal-fn.def (MASK_LOAD_LANES): New internal function.
13983 (MASK_STORE_LANES): Likewise.
13984 * internal-fn.c (mask_load_lanes_direct): New macro.
13985 (mask_store_lanes_direct): Likewise.
13986 (expand_mask_load_optab_fn): Handle masked operations.
13987 (expand_mask_load_lanes_optab_fn): New macro.
13988 (expand_mask_store_optab_fn): Handle masked operations.
13989 (expand_mask_store_lanes_optab_fn): New macro.
13990 (direct_mask_load_lanes_optab_supported_p): Likewise.
13991 (direct_mask_store_lanes_optab_supported_p): Likewise.
13992 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
13994 (vect_load_lanes_supported): Likewise.
13995 * tree-vect-data-refs.c (strip_conversion): New function.
13996 (can_group_stmts_p): Likewise.
13997 (vect_analyze_data_ref_accesses): Use it instead of checking
13998 for a pair of assignments.
13999 (vect_store_lanes_supported): Take a masked_p parameter.
14000 (vect_load_lanes_supported): Likewise.
14001 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
14002 vect_store_lanes_supported and vect_load_lanes_supported.
14003 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14004 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
14005 parameter. Don't allow gaps for masked accesses.
14006 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
14007 and vect_load_lanes_supported.
14008 (get_load_store_type): Take a masked_p parameter and update
14009 call to get_group_load_store_type.
14010 (vectorizable_store): Update call to get_load_store_type.
14011 Handle IFN_MASK_STORE_LANES.
14012 (vectorizable_load): Update call to get_load_store_type.
14013 Handle IFN_MASK_LOAD_LANES.
14015 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14016 Alan Hayward <alan.hayward@arm.com>
14017 David Sherwood <david.sherwood@arm.com>
14019 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
14021 * config/aarch64/aarch64-protos.h
14022 (aarch64_sve_struct_memory_operand_p): Declare.
14023 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
14024 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
14025 (VPRED, vpred): Handle SVE structure modes.
14026 * config/aarch64/constraints.md (Utx): New constraint.
14027 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
14028 (aarch64_sve_struct_nonimmediate_operand): New predicates.
14029 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
14030 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
14031 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
14032 structure modes. Split into pieces after RA.
14033 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
14034 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
14036 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
14037 SVE structure modes.
14038 (aarch64_classify_address): Likewise.
14039 (sizetochar): Move earlier in file.
14040 (aarch64_print_operand): Handle SVE register lists.
14041 (aarch64_array_mode): New function.
14042 (aarch64_sve_struct_memory_operand_p): Likewise.
14043 (TARGET_ARRAY_MODE): Redefine.
14045 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14046 Alan Hayward <alan.hayward@arm.com>
14047 David Sherwood <david.sherwood@arm.com>
14049 * target.def (array_mode): New target hook.
14050 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
14051 * doc/tm.texi: Regenerate.
14052 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
14053 * hooks.c (hook_optmode_mode_uhwi_none): New function.
14054 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
14055 targetm.array_mode.
14056 * stor-layout.c (mode_for_array): Likewise. Support polynomial
14059 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14060 Alan Hayward <alan.hayward@arm.com>
14061 David Sherwood <david.sherwood@arm.com>
14063 * fold-const.c (fold_binary_loc): Check the argument types
14064 rather than the result type when testing for a vector operation.
14066 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14068 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
14069 * doc/tm.texi: Regenerate.
14071 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14072 Alan Hayward <alan.hayward@arm.com>
14073 David Sherwood <david.sherwood@arm.com>
14075 * doc/invoke.texi (-msve-vector-bits=): Document new option.
14076 (sve): Document new AArch64 extension.
14077 * doc/md.texi (w): Extend the description of the AArch64
14078 constraint to include SVE vectors.
14079 (Upl, Upa): Document new AArch64 predicate constraints.
14080 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
14082 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
14083 (msve-vector-bits=): New option.
14084 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
14085 SVE when these are disabled.
14086 (sve): New extension.
14087 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
14088 modes. Adjust their number of units based on aarch64_sve_vg.
14089 (MAX_BITSIZE_MODE_ANY_MODE): Define.
14090 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
14091 aarch64_addr_query_type.
14092 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
14093 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
14094 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
14095 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
14096 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
14097 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
14098 (aarch64_simd_imm_zero_p): Delete.
14099 (aarch64_check_zero_based_sve_index_immediate): Declare.
14100 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14101 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14102 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14103 (aarch64_sve_float_mul_immediate_p): Likewise.
14104 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14105 rather than an rtx.
14106 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
14107 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
14108 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
14109 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
14110 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
14111 (aarch64_regmode_natural_size): Likewise.
14112 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
14113 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
14115 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
14116 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
14117 for VG and the SVE predicate registers.
14118 (V_ALIASES): Add a "z"-prefixed alias.
14119 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
14120 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
14121 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
14122 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
14123 (REG_CLASS_NAMES): Add entries for them.
14124 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
14125 and the predicate registers.
14126 (aarch64_sve_vg): Declare.
14127 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
14128 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
14129 (REGMODE_NATURAL_SIZE): Define.
14130 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
14132 * config/aarch64/aarch64.c: Include cfgrtl.h.
14133 (simd_immediate_info): Add a constructor for series vectors,
14134 and an associated step field.
14135 (aarch64_sve_vg): New variable.
14136 (aarch64_dbx_register_number): Handle VG and the predicate registers.
14137 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
14138 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
14139 (VEC_ANY_DATA, VEC_STRUCT): New constants.
14140 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
14141 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
14142 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
14143 (aarch64_get_mask_mode): New functions.
14144 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
14145 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
14146 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
14147 predicate modes and predicate registers. Explicitly restrict
14148 GPRs to modes of 16 bytes or smaller. Only allow FP registers
14149 to store a vector mode if it is recognized by
14150 aarch64_classify_vector_mode.
14151 (aarch64_regmode_natural_size): New function.
14152 (aarch64_hard_regno_caller_save_mode): Return the original mode
14154 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
14155 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
14156 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
14157 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
14159 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
14160 does not overlap dest if the function is frame-related. Handle
14162 (aarch64_split_add_offset): New function.
14163 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
14164 them aarch64_add_offset.
14165 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
14166 and update call to aarch64_sub_sp.
14167 (aarch64_add_cfa_expression): New function.
14168 (aarch64_expand_prologue): Pass extra temporary registers to the
14169 functions above. Handle the case in which we need to emit new
14170 DW_CFA_expressions for registers that were originally saved
14171 relative to the stack pointer, but now have to be expressed
14172 relative to the frame pointer.
14173 (aarch64_output_mi_thunk): Pass extra temporary registers to the
14175 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
14176 IP0 and IP1 values for SVE frames.
14177 (aarch64_expand_vec_series): New function.
14178 (aarch64_expand_sve_widened_duplicate): Likewise.
14179 (aarch64_expand_sve_const_vector): Likewise.
14180 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
14181 Handle SVE constants. Use emit_move_insn to move a force_const_mem
14182 into the register, rather than emitting a SET directly.
14183 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
14184 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
14185 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
14186 (offset_9bit_signed_scaled_p): New functions.
14187 (aarch64_replicate_bitmask_imm): New function.
14188 (aarch64_bitmask_imm): Use it.
14189 (aarch64_cannot_force_const_mem): Reject expressions involving
14190 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
14191 (aarch64_classify_index): Handle SVE indices, by requiring
14192 a plain register index with a scale that matches the element size.
14193 (aarch64_classify_address): Handle SVE addresses. Assert that
14194 the mode of the address is VOIDmode or an integer mode.
14195 Update call to aarch64_classify_symbol.
14196 (aarch64_classify_symbolic_expression): Update call to
14197 aarch64_classify_symbol.
14198 (aarch64_const_vec_all_in_range_p): New function.
14199 (aarch64_print_vector_float_operand): Likewise.
14200 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
14201 "vN" for FP registers with SVE modes. Handle (const ...) vectors
14202 and the FP immediates 1.0 and 0.5.
14203 (aarch64_print_address_internal): Handle SVE addresses.
14204 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
14205 (aarch64_regno_regclass): Handle predicate registers.
14206 (aarch64_secondary_reload): Handle big-endian reloads of SVE
14208 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
14209 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
14210 (aarch64_convert_sve_vector_bits): New function.
14211 (aarch64_override_options): Use it to handle -msve-vector-bits=.
14212 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14213 rather than an rtx.
14214 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
14215 Handle SVE vector and predicate modes. Accept VL-based constants
14216 that need only one temporary register, and VL offsets that require
14217 no temporary registers.
14218 (aarch64_conditional_register_usage): Mark the predicate registers
14219 as fixed if SVE isn't available.
14220 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
14221 Return true for SVE vector and predicate modes.
14222 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
14223 rather than an unsigned int. Handle SVE modes.
14224 (aarch64_preferred_simd_mode): Update call accordingly. Handle
14226 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
14228 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14229 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14230 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14231 (aarch64_sve_float_mul_immediate_p): New functions.
14232 (aarch64_sve_valid_immediate): New function.
14233 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
14234 Explicitly reject structure modes. Check for INDEX constants.
14235 Handle PTRUE and PFALSE constants.
14236 (aarch64_check_zero_based_sve_index_immediate): New function.
14237 (aarch64_simd_imm_zero_p): Delete.
14238 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
14239 vector modes. Accept constants in the range of CNT[BHWD].
14240 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
14241 ask for an Advanced SIMD mode.
14242 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
14243 (aarch64_simd_vector_alignment): Handle SVE predicates.
14244 (aarch64_vectorize_preferred_vector_alignment): New function.
14245 (aarch64_simd_vector_alignment_reachable): Use it instead of
14247 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
14248 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
14250 (MAX_VECT_LEN): Delete.
14251 (expand_vec_perm_d): Add a vec_flags field.
14252 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
14253 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14254 (aarch64_evpc_ext): Don't apply a big-endian lane correction
14256 (aarch64_evpc_rev): Rename to...
14257 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
14258 (aarch64_evpc_rev_global): New function.
14259 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
14260 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
14262 (aarch64_evpc_sve_tbl): New function.
14263 (aarch64_expand_vec_perm_const_1): Update after rename of
14264 aarch64_evpc_rev. Handle SVE permutes too, trying
14265 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
14266 than aarch64_evpc_tbl.
14267 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
14268 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
14269 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
14270 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
14271 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
14272 (aarch64_expand_sve_vcond): New functions.
14273 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
14274 of aarch64_vector_mode_p.
14275 (aarch64_dwarf_poly_indeterminate_value): New function.
14276 (aarch64_compute_pressure_classes): Likewise.
14277 (aarch64_can_change_mode_class): Likewise.
14278 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
14279 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
14280 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
14281 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
14282 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
14283 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
14284 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
14285 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
14287 (Dn, Dl, Dr): Accept const as well as const_vector.
14288 (Dz): Likewise. Compare against CONST0_RTX.
14289 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
14290 of "vector" where appropriate.
14291 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
14292 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
14293 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
14294 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
14295 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
14296 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
14297 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
14298 (v_int_equiv): Extend to SVE modes.
14299 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
14301 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
14302 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
14303 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
14304 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
14305 (SVE_COND_FP_CMP): New int iterators.
14306 (perm_hilo): Handle the new unpack unspecs.
14307 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
14309 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
14310 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
14311 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
14312 (aarch64_equality_operator, aarch64_constant_vector_operand)
14313 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
14314 (aarch64_sve_nonimmediate_operand): Likewise.
14315 (aarch64_sve_general_operand): Likewise.
14316 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
14317 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
14318 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
14319 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
14320 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
14321 (aarch64_sve_float_arith_immediate): Likewise.
14322 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
14323 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
14324 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
14325 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
14326 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
14327 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
14328 (aarch64_sve_float_arith_operand): Likewise.
14329 (aarch64_sve_float_arith_with_sub_operand): Likewise.
14330 (aarch64_sve_float_mul_operand): Likewise.
14331 (aarch64_sve_vec_perm_operand): Likewise.
14332 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
14333 (aarch64_mov_operand): Accept const_poly_int and const_vector.
14334 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
14335 as well as const_vector.
14336 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
14337 in file. Use CONST0_RTX and CONSTM1_RTX.
14338 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
14339 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
14340 Use aarch64_simd_imm_zero.
14341 * config/aarch64/aarch64-sve.md: New file.
14342 * config/aarch64/aarch64.md: Include it.
14343 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
14344 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
14345 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
14346 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
14347 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
14348 (sve): New attribute.
14349 (enabled): Disable instructions with the sve attribute unless
14351 (movqi, movhi): Pass CONST_POLY_INT operaneds through
14352 aarch64_expand_mov_immediate.
14353 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
14354 CNT[BHSD] immediates.
14355 (movti): Split CONST_POLY_INT moves into two halves.
14356 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
14357 Split additions that need a temporary here if the destination
14358 is the stack pointer.
14359 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
14360 (*add<mode>3_poly_1): New instruction.
14361 (set_clobber_cc): New expander.
14363 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14365 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
14366 parameter and use it instead of GET_MODE_SIZE (innermode). Use
14367 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
14368 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
14369 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
14370 Change innermode from fixed_mode_size to machine_mode.
14371 (simplify_subreg): Update call accordingly. Handle a constant-sized
14372 subreg of a variable-length CONST_VECTOR.
14374 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14375 Alan Hayward <alan.hayward@arm.com>
14376 David Sherwood <david.sherwood@arm.com>
14378 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
14379 (add_offset_to_base): New function, split out from...
14380 (create_mem_ref): ...here. When handling a scale other than 1,
14381 check first whether the address is valid without the offset.
14382 Add it into the base if so, leaving the index and scale as-is.
14384 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14387 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
14388 fold_for_warn before checking if arg2 is INTEGER_CST.
14390 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
14392 * config/rs6000/predicates.md (load_multiple_operation): Delete.
14393 (store_multiple_operation): Delete.
14394 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
14395 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
14396 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
14397 guarded by TARGET_STRING.
14398 (rs6000_output_load_multiple): Delete.
14399 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
14400 OPTION_MASK_STRING / TARGET_STRING handling.
14401 (print_operand) <'N', 'O'>: Add comment that these are unused now.
14402 (const rs6000_opt_masks) <"string">: Change mask to 0.
14403 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
14404 (MASK_STRING): Delete.
14405 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
14407 (load_multiple): Delete.
14414 (store_multiple): Delete.
14421 (movmemsi_8reg): Delete.
14422 (corresponding unnamed define_insn): Delete.
14423 (movmemsi_6reg): Delete.
14424 (corresponding unnamed define_insn): Delete.
14425 (movmemsi_4reg): Delete.
14426 (corresponding unnamed define_insn): Delete.
14427 (movmemsi_2reg): Delete.
14428 (corresponding unnamed define_insn): Delete.
14429 (movmemsi_1reg): Delete.
14430 (corresponding unnamed define_insn): Delete.
14431 * config/rs6000/rs6000.opt (mno-string): New.
14432 (mstring): Replace by deprecation warning stub.
14433 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
14435 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14437 * regrename.c (regrename_do_replace): If replacing the same
14438 reg multiple times, try to reuse last created gen_raw_REG.
14441 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
14442 main to workaround a bug in GDB.
14444 2018-01-12 Tom de Vries <tom@codesourcery.com>
14447 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
14449 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
14451 PR rtl-optimization/80481
14452 * ira-color.c (get_cap_member): New function.
14453 (allocnos_conflict_by_live_ranges_p): Use it.
14454 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
14455 (setup_slot_coalesced_allocno_live_ranges): Ditto.
14457 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
14460 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
14461 (*saddl_se_1): Ditto.
14462 (*ssubsi_1): Ditto.
14463 (*ssubl_se_1): Ditto.
14465 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14467 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
14468 rather than wi::to_widest for DR_INITs.
14469 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
14470 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
14471 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
14473 (vect_analyze_group_access_1): Note that here.
14475 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14477 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
14478 polynomial type sizes.
14480 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14482 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
14483 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
14484 (gimple_add_tmp_var): Likewise.
14486 2018-01-12 Martin Liska <mliska@suse.cz>
14488 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
14489 (gimple_alloc_sizes): Likewise.
14490 (dump_gimple_statistics): Use PRIu64 in printf format.
14491 * gimple.h: Change uint64_t to int.
14493 2018-01-12 Martin Liska <mliska@suse.cz>
14495 * tree-core.h: Use uint64_t instead of int.
14496 * tree.c (tree_node_counts): Likewise.
14497 (tree_node_sizes): Likewise.
14498 (dump_tree_statistics): Use PRIu64 in printf format.
14500 2018-01-12 Martin Liska <mliska@suse.cz>
14502 * Makefile.in: As qsort_chk is implemented in vec.c, add
14503 vec.o to linkage of gencfn-macros.
14504 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
14505 passing the info to record_node_allocation_statistics.
14506 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
14508 * ggc-common.c (struct ggc_usage): Add operator== and use
14509 it in operator< and compare function.
14510 * mem-stats.h (struct mem_usage): Likewise.
14511 * vec.c (struct vec_usage): Remove operator< and compare
14512 function. Can be simply inherited.
14514 2018-01-12 Martin Jambor <mjambor@suse.cz>
14517 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
14518 * tree-ssa-math-opts.c: Include domwalk.h.
14519 (convert_mult_to_fma_1): New function.
14520 (fma_transformation_info): New type.
14521 (fma_deferring_state): Likewise.
14522 (cancel_fma_deferring): New function.
14523 (result_of_phi): Likewise.
14524 (last_fma_candidate_feeds_initial_phi): Likewise.
14525 (convert_mult_to_fma): Added deferring logic, split actual
14526 transformation to convert_mult_to_fma_1.
14527 (math_opts_dom_walker): New type.
14528 (math_opts_dom_walker::after_dom_children): New method, body moved
14529 here from pass_optimize_widening_mul::execute, added deferring logic
14531 (pass_optimize_widening_mul::execute): Moved most of code to
14532 math_opts_dom_walker::after_dom_children.
14533 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
14534 * config/i386/i386.c (ix86_option_override_internal): Added
14535 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
14537 2018-01-12 Richard Biener <rguenther@suse.de>
14540 * dwarf2out.c (gen_variable_die): Do not reset old_die for
14541 inline instance vars.
14543 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
14546 * config/rx/rx.c (rx_is_restricted_memory_address):
14547 Handle SUBREG case.
14549 2018-01-12 Richard Biener <rguenther@suse.de>
14551 PR tree-optimization/80846
14552 * target.def (split_reduction): New target hook.
14553 * targhooks.c (default_split_reduction): New function.
14554 * targhooks.h (default_split_reduction): Declare.
14555 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
14556 target requests first reduce vectors by combining low and high
14558 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
14559 (get_vectype_for_scalar_type_and_size): Export.
14560 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
14561 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
14562 * doc/tm.texi: Regenerate.
14563 * config/i386/i386.c (ix86_split_reduction): Implement
14564 TARGET_VECTORIZE_SPLIT_REDUCTION.
14566 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14569 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
14570 in PIC mode except for TARGET_VXWORKS_RTP.
14571 * config/sparc/sparc.c: Include cfgrtl.h.
14572 (TARGET_INIT_PIC_REG): Define.
14573 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
14574 (sparc_pic_register_p): New predicate.
14575 (sparc_legitimate_address_p): Use it.
14576 (sparc_legitimize_pic_address): Likewise.
14577 (sparc_delegitimize_address): Likewise.
14578 (sparc_mode_dependent_address_p): Likewise.
14579 (gen_load_pcrel_sym): Remove 4th parameter.
14580 (load_got_register): Adjust call to above. Remove obsolete stuff.
14581 (sparc_expand_prologue): Do not call load_got_register here.
14582 (sparc_flat_expand_prologue): Likewise.
14583 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
14584 (sparc_use_pseudo_pic_reg): New function.
14585 (sparc_init_pic_reg): Likewise.
14586 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14587 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
14589 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
14591 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
14592 Add item for branch_cost.
14594 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14596 PR rtl-optimization/83565
14597 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
14598 not extend the result to a larger mode for rotate operations.
14599 (num_sign_bit_copies1): Likewise.
14601 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14604 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
14606 Use values-Xc.o for -pedantic.
14607 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
14609 2018-01-12 Martin Liska <mliska@suse.cz>
14612 * ipa-devirt.c (final_warning_record::grow_type_warnings):
14614 (possible_polymorphic_call_targets): Use it.
14615 (ipa_devirt): Likewise.
14617 2018-01-12 Martin Liska <mliska@suse.cz>
14619 * profile-count.h (enum profile_quality): Use 0 as invalid
14620 enum value of profile_quality.
14622 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
14624 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
14625 -mext-string options.
14627 2018-01-12 Richard Biener <rguenther@suse.de>
14629 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
14630 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
14631 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
14633 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
14635 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
14637 * configure.ac (--with-long-double-format): Add support for the
14638 configuration option to change the default long double format on
14640 * config.gcc (powerpc*-linux*-*): Likewise.
14641 * configure: Regenerate.
14642 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
14643 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
14644 used without modification.
14646 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14648 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
14649 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
14650 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
14651 MISC_BUILTIN_SPEC_BARRIER.
14652 (rs6000_init_builtins): Likewise.
14653 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
14655 (speculation_barrier): New define_insn.
14656 * doc/extend.texi: Document __builtin_speculation_barrier.
14658 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14661 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
14662 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
14663 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
14665 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
14666 integral modes instead of "ss" and "sd".
14667 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
14668 vectors with 32-bit and 64-bit elements.
14669 (vecdupssescalarmodesuffix): New mode attribute.
14670 (vec_dup<mode>): Use it.
14672 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
14675 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
14676 frame if argument is passed on stack.
14678 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14681 * ree.c (combine_reaching_defs): Optimize also
14682 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
14683 reg2=any_extend(exp); reg1=reg2;, formatting fix.
14685 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14687 PR middle-end/83189
14688 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
14690 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14692 PR middle-end/83718
14693 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
14694 after they are computed.
14696 2018-01-11 Bin Cheng <bin.cheng@arm.com>
14698 PR tree-optimization/83695
14699 * gimple-loop-linterchange.cc
14700 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
14701 reset cached scev information after interchange.
14702 (pass_linterchange::execute): Remove call to scev_reset_htab.
14704 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14706 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
14707 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
14708 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
14709 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
14710 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
14711 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
14712 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
14713 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
14714 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
14715 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
14716 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
14717 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
14718 (V_lane_reg): Likewise.
14719 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
14721 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
14722 (vfmal_lane_low<mode>_intrinsic,
14723 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
14724 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
14725 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
14726 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
14727 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
14728 vfmsl_lane_high<mode>_intrinsic): New define_insns.
14730 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14732 * config/arm/arm-cpus.in (fp16fml): New feature.
14733 (ALL_SIMD): Add fp16fml.
14734 (armv8.2-a): Add fp16fml as an option.
14735 (armv8.3-a): Likewise.
14736 (armv8.4-a): Add fp16fml as part of fp16.
14737 * config/arm/arm.h (TARGET_FP16FML): Define.
14738 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
14740 * config/arm/arm-modes.def (V2HF): Define.
14741 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
14742 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
14743 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
14744 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
14745 vfmsl_low, vfmsl_high): New set of builtins.
14746 * config/arm/iterators.md (PLUSMINUS): New code iterator.
14747 (vfml_op): New code attribute.
14748 (VFMLHALVES): New int iterator.
14749 (VFML, VFMLSEL): New mode attributes.
14750 (V_reg): Define mapping for V2HF.
14751 (V_hi, V_lo): New mode attributes.
14752 (VF_constraint): Likewise.
14753 (vfml_half, vfml_half_selector): New int attributes.
14754 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
14756 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
14757 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
14759 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
14760 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
14761 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
14762 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
14764 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
14765 Document new effective target and option set.
14767 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14769 * config/arm/arm-cpus.in (armv8_4): New feature.
14770 (ARMv8_4a): New fgroup.
14771 (armv8.4-a): New arch.
14772 * config/arm/arm-tables.opt: Regenerate.
14773 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
14774 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
14775 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
14776 Add matching rules for -march=armv8.4-a and extensions.
14777 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
14779 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
14782 * config/rx/rx.md (BW): New mode attribute.
14783 (sync_lock_test_and_setsi): Add mode suffix to insn output.
14785 2018-01-11 Richard Biener <rguenther@suse.de>
14787 PR tree-optimization/83435
14788 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
14789 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
14790 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
14792 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14793 Alan Hayward <alan.hayward@arm.com>
14794 David Sherwood <david.sherwood@arm.com>
14796 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
14798 (aarch64_classify_address): Initialize it. Track polynomial offsets.
14799 (aarch64_print_address_internal): Use it to check for a zero offset.
14801 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14802 Alan Hayward <alan.hayward@arm.com>
14803 David Sherwood <david.sherwood@arm.com>
14805 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
14806 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
14807 Return a poly_int64 rather than a HOST_WIDE_INT.
14808 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
14809 rather than a HOST_WIDE_INT.
14810 * config/aarch64/aarch64.h (aarch64_frame): Protect with
14811 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
14812 hard_fp_offset, frame_size, initial_adjust, callee_offset and
14813 final_offset from HOST_WIDE_INT to poly_int64.
14814 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
14815 to_constant when getting the number of units in an Advanced SIMD
14817 (aarch64_builtin_vectorized_function): Check for a constant number
14819 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
14821 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
14822 attribute instead of GET_MODE_NUNITS.
14823 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
14824 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
14825 GET_MODE_SIZE for fixed-size registers.
14826 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
14827 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
14828 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
14829 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
14830 (aarch64_print_operand, aarch64_print_address_internal)
14831 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
14832 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
14833 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
14834 Handle polynomial GET_MODE_SIZE.
14835 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
14836 wider than SImode without modification.
14837 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
14838 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
14839 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
14840 passing and returning SVE modes.
14841 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
14842 rather than GEN_INT.
14843 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
14844 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
14845 (aarch64_allocate_and_probe_stack_space): Likewise.
14846 (aarch64_layout_frame): Cope with polynomial offsets.
14847 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
14848 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
14849 polynomial offsets.
14850 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
14851 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
14852 poly_int64 rather than a HOST_WIDE_INT.
14853 (aarch64_get_separate_components, aarch64_process_components)
14854 (aarch64_expand_prologue, aarch64_expand_epilogue)
14855 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
14856 (aarch64_anchor_offset): New function, split out from...
14857 (aarch64_legitimize_address): ...here.
14858 (aarch64_builtin_vectorization_cost): Handle polynomial
14859 TYPE_VECTOR_SUBPARTS.
14860 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
14862 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
14863 number of elements from the PARALLEL rather than the mode.
14864 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
14865 rather than GET_MODE_BITSIZE.
14866 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
14867 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
14868 (aarch64_expand_vec_perm_const_1): Handle polynomial
14869 d->perm.length () and d->perm elements.
14870 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
14871 Apply to_constant to d->perm elements.
14872 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
14873 polynomial CONST_VECTOR_NUNITS.
14874 (aarch64_move_pointer): Take amount as a poly_int64 rather
14876 (aarch64_progress_pointer): Avoid temporary variable.
14877 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
14878 the mode attribute instead of GET_MODE.
14880 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14881 Alan Hayward <alan.hayward@arm.com>
14882 David Sherwood <david.sherwood@arm.com>
14884 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
14885 x exists before using it.
14886 (aarch64_add_constant_internal): Rename to...
14887 (aarch64_add_offset_1): ...this. Replace regnum with separate
14888 src and dest rtxes. Handle the case in which they're different,
14889 including when the offset is zero. Replace scratchreg with an rtx.
14890 Use 2 additions if there is no spare register into which we can
14891 move a 16-bit constant.
14892 (aarch64_add_constant): Delete.
14893 (aarch64_add_offset): Replace reg with separate src and dest
14894 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
14895 Use aarch64_add_offset_1.
14896 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
14897 an rtx rather than an int. Take the delta as a poly_int64
14898 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
14899 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
14900 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
14901 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
14902 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
14903 and aarch64_add_sp.
14904 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
14905 aarch64_add_constant.
14907 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14909 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
14910 Use scalar_float_mode.
14912 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14914 * config/aarch64/aarch64-simd.md
14915 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
14916 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
14917 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
14918 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
14919 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
14920 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
14921 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
14922 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
14923 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
14924 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
14926 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14929 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
14930 targ_options->x_arm_arch_string is non NULL.
14932 2018-01-11 Tamar Christina <tamar.christina@arm.com>
14934 * config/aarch64/aarch64.h
14935 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
14937 2018-01-11 Sudakshina Das <sudi.das@arm.com>
14940 * expmed.c (emit_store_flag_force): Swap if const op0
14941 and change VOIDmode to mode of op0.
14943 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14945 PR rtl-optimization/83761
14946 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
14947 than bytes to mode_for_size.
14949 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14951 PR middle-end/83189
14952 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
14953 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
14956 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14958 PR middle-end/83575
14959 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
14960 when in layout mode.
14961 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
14962 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
14965 2018-01-10 Michael Collison <michael.collison@arm.com>
14967 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
14968 * config/aarch64/aarch64-option-extension.def: Add
14969 AARCH64_OPT_EXTENSION of 'fp16fml'.
14970 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14971 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
14972 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
14973 * config/aarch64/constraints.md (Ui7): New constraint.
14974 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
14975 (VFMLA_SEL_W): Ditto.
14978 (VFMLA16_LOW): New int iterator.
14979 (VFMLA16_HIGH): Ditto.
14980 (UNSPEC_FMLAL): New unspec.
14981 (UNSPEC_FMLSL): Ditto.
14982 (UNSPEC_FMLAL2): Ditto.
14983 (UNSPEC_FMLSL2): Ditto.
14984 (f16mac): New code attribute.
14985 * config/aarch64/aarch64-simd-builtins.def
14986 (aarch64_fmlal_lowv2sf): Ditto.
14987 (aarch64_fmlsl_lowv2sf): Ditto.
14988 (aarch64_fmlalq_lowv4sf): Ditto.
14989 (aarch64_fmlslq_lowv4sf): Ditto.
14990 (aarch64_fmlal_highv2sf): Ditto.
14991 (aarch64_fmlsl_highv2sf): Ditto.
14992 (aarch64_fmlalq_highv4sf): Ditto.
14993 (aarch64_fmlslq_highv4sf): Ditto.
14994 (aarch64_fmlal_lane_lowv2sf): Ditto.
14995 (aarch64_fmlsl_lane_lowv2sf): Ditto.
14996 (aarch64_fmlal_laneq_lowv2sf): Ditto.
14997 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
14998 (aarch64_fmlalq_lane_lowv4sf): Ditto.
14999 (aarch64_fmlsl_lane_lowv4sf): Ditto.
15000 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
15001 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
15002 (aarch64_fmlal_lane_highv2sf): Ditto.
15003 (aarch64_fmlsl_lane_highv2sf): Ditto.
15004 (aarch64_fmlal_laneq_highv2sf): Ditto.
15005 (aarch64_fmlsl_laneq_highv2sf): Ditto.
15006 (aarch64_fmlalq_lane_highv4sf): Ditto.
15007 (aarch64_fmlsl_lane_highv4sf): Ditto.
15008 (aarch64_fmlalq_laneq_highv4sf): Ditto.
15009 (aarch64_fmlsl_laneq_highv4sf): Ditto.
15010 * config/aarch64/aarch64-simd.md:
15011 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
15012 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15013 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
15014 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15015 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
15016 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
15017 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
15018 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
15019 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
15020 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
15021 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
15022 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
15023 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
15024 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
15025 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
15026 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
15027 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
15028 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
15029 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
15030 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
15031 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
15032 (vfmlsl_low_u32): Ditto.
15033 (vfmlalq_low_u32): Ditto.
15034 (vfmlslq_low_u32): Ditto.
15035 (vfmlal_high_u32): Ditto.
15036 (vfmlsl_high_u32): Ditto.
15037 (vfmlalq_high_u32): Ditto.
15038 (vfmlslq_high_u32): Ditto.
15039 (vfmlal_lane_low_u32): Ditto.
15040 (vfmlsl_lane_low_u32): Ditto.
15041 (vfmlal_laneq_low_u32): Ditto.
15042 (vfmlsl_laneq_low_u32): Ditto.
15043 (vfmlalq_lane_low_u32): Ditto.
15044 (vfmlslq_lane_low_u32): Ditto.
15045 (vfmlalq_laneq_low_u32): Ditto.
15046 (vfmlslq_laneq_low_u32): Ditto.
15047 (vfmlal_lane_high_u32): Ditto.
15048 (vfmlsl_lane_high_u32): Ditto.
15049 (vfmlal_laneq_high_u32): Ditto.
15050 (vfmlsl_laneq_high_u32): Ditto.
15051 (vfmlalq_lane_high_u32): Ditto.
15052 (vfmlslq_lane_high_u32): Ditto.
15053 (vfmlalq_laneq_high_u32): Ditto.
15054 (vfmlslq_laneq_high_u32): Ditto.
15055 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
15056 (AARCH64_FL_FOR_ARCH8_4): New.
15057 (AARCH64_ISA_F16FML): New ISA flag.
15058 (TARGET_F16FML): New feature flag for fp16fml.
15059 (doc/invoke.texi): Document new fp16fml option.
15061 2018-01-10 Michael Collison <michael.collison@arm.com>
15063 * config/aarch64/aarch64-builtins.c:
15064 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
15065 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15066 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
15067 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
15068 (AARCH64_ISA_SHA3): New ISA flag.
15069 (TARGET_SHA3): New feature flag for sha3.
15070 * config/aarch64/iterators.md (sha512_op): New int attribute.
15071 (CRYPTO_SHA512): New int iterator.
15072 (UNSPEC_SHA512H): New unspec.
15073 (UNSPEC_SHA512H2): Ditto.
15074 (UNSPEC_SHA512SU0): Ditto.
15075 (UNSPEC_SHA512SU1): Ditto.
15076 * config/aarch64/aarch64-simd-builtins.def
15077 (aarch64_crypto_sha512hqv2di): New builtin.
15078 (aarch64_crypto_sha512h2qv2di): Ditto.
15079 (aarch64_crypto_sha512su0qv2di): Ditto.
15080 (aarch64_crypto_sha512su1qv2di): Ditto.
15081 (aarch64_eor3qv8hi): Ditto.
15082 (aarch64_rax1qv2di): Ditto.
15083 (aarch64_xarqv2di): Ditto.
15084 (aarch64_bcaxqv8hi): Ditto.
15085 * config/aarch64/aarch64-simd.md:
15086 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
15087 (aarch64_crypto_sha512su0qv2di): Ditto.
15088 (aarch64_crypto_sha512su1qv2di): Ditto.
15089 (aarch64_eor3qv8hi): Ditto.
15090 (aarch64_rax1qv2di): Ditto.
15091 (aarch64_xarqv2di): Ditto.
15092 (aarch64_bcaxqv8hi): Ditto.
15093 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
15094 (vsha512h2q_u64): Ditto.
15095 (vsha512su0q_u64): Ditto.
15096 (vsha512su1q_u64): Ditto.
15097 (veor3q_u16): Ditto.
15098 (vrax1q_u64): Ditto.
15099 (vxarq_u64): Ditto.
15100 (vbcaxq_u16): Ditto.
15101 * config/arm/types.md (crypto_sha512): New type attribute.
15102 (crypto_sha3): Ditto.
15103 (doc/invoke.texi): Document new sha3 option.
15105 2018-01-10 Michael Collison <michael.collison@arm.com>
15107 * config/aarch64/aarch64-builtins.c:
15108 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
15109 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15110 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
15111 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
15112 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
15113 (AARCH64_ISA_SM4): New ISA flag.
15114 (TARGET_SM4): New feature flag for sm4.
15115 * config/aarch64/aarch64-simd-builtins.def
15116 (aarch64_sm3ss1qv4si): Ditto.
15117 (aarch64_sm3tt1aq4si): Ditto.
15118 (aarch64_sm3tt1bq4si): Ditto.
15119 (aarch64_sm3tt2aq4si): Ditto.
15120 (aarch64_sm3tt2bq4si): Ditto.
15121 (aarch64_sm3partw1qv4si): Ditto.
15122 (aarch64_sm3partw2qv4si): Ditto.
15123 (aarch64_sm4eqv4si): Ditto.
15124 (aarch64_sm4ekeyqv4si): Ditto.
15125 * config/aarch64/aarch64-simd.md:
15126 (aarch64_sm3ss1qv4si): Ditto.
15127 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
15128 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
15129 (aarch64_sm4eqv4si): Ditto.
15130 (aarch64_sm4ekeyqv4si): Ditto.
15131 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
15132 (sm3part_op): Ditto.
15133 (CRYPTO_SM3TT): Ditto.
15134 (CRYPTO_SM3PART): Ditto.
15135 (UNSPEC_SM3SS1): New unspec.
15136 (UNSPEC_SM3TT1A): Ditto.
15137 (UNSPEC_SM3TT1B): Ditto.
15138 (UNSPEC_SM3TT2A): Ditto.
15139 (UNSPEC_SM3TT2B): Ditto.
15140 (UNSPEC_SM3PARTW1): Ditto.
15141 (UNSPEC_SM3PARTW2): Ditto.
15142 (UNSPEC_SM4E): Ditto.
15143 (UNSPEC_SM4EKEY): Ditto.
15144 * config/aarch64/constraints.md (Ui2): New constraint.
15145 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
15146 * config/arm/types.md (crypto_sm3): New type attribute.
15147 (crypto_sm4): Ditto.
15148 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
15149 (vsm3tt1aq_u32): Ditto.
15150 (vsm3tt1bq_u32): Ditto.
15151 (vsm3tt2aq_u32): Ditto.
15152 (vsm3tt2bq_u32): Ditto.
15153 (vsm3partw1q_u32): Ditto.
15154 (vsm3partw2q_u32): Ditto.
15155 (vsm4eq_u32): Ditto.
15156 (vsm4ekeyq_u32): Ditto.
15157 (doc/invoke.texi): Document new sm4 option.
15159 2018-01-10 Michael Collison <michael.collison@arm.com>
15161 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
15162 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
15163 (AARCH64_FL_FOR_ARCH8_4): New.
15164 (AARCH64_FL_V8_4): New flag.
15165 (doc/invoke.texi): Document new armv8.4-a option.
15167 2018-01-10 Michael Collison <michael.collison@arm.com>
15169 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15170 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
15171 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
15172 * config/aarch64/aarch64-option-extension.def: Add
15173 AARCH64_OPT_EXTENSION of 'sha2'.
15174 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
15175 (crypto): Disable sha2 and aes if crypto disabled.
15176 (crypto): Enable aes and sha2 if enabled.
15177 (simd): Disable sha2 and aes if simd disabled.
15178 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
15180 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
15181 (TARGET_SHA2): New feature flag for sha2.
15182 (TARGET_AES): New feature flag for aes.
15183 * config/aarch64/aarch64-simd.md:
15184 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
15185 conditional on TARGET_AES.
15186 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
15187 (aarch64_crypto_sha1hsi): Make pattern conditional
15189 (aarch64_crypto_sha1hv4si): Ditto.
15190 (aarch64_be_crypto_sha1hv4si): Ditto.
15191 (aarch64_crypto_sha1su1v4si): Ditto.
15192 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
15193 (aarch64_crypto_sha1su0v4si): Ditto.
15194 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
15195 (aarch64_crypto_sha256su0v4si): Ditto.
15196 (aarch64_crypto_sha256su1v4si): Ditto.
15197 (doc/invoke.texi): Document new aes and sha2 options.
15199 2018-01-10 Martin Sebor <msebor@redhat.com>
15201 PR tree-optimization/83781
15202 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
15205 2018-01-11 Martin Sebor <msebor@gmail.com>
15206 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15208 PR tree-optimization/83501
15209 PR tree-optimization/81703
15211 * tree-ssa-strlen.c (get_string_cst): Rename...
15212 (get_string_len): ...to this. Handle global constants.
15213 (handle_char_store): Adjust.
15215 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
15216 Jim Wilson <jimw@sifive.com>
15218 * config/riscv/riscv-protos.h (riscv_output_return): New.
15219 * config/riscv/riscv.c (struct machine_function): New naked_p field.
15220 (riscv_attribute_table, riscv_output_return),
15221 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
15222 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
15223 (riscv_compute_frame_info): Only compute frame->mask if not a naked
15225 (riscv_expand_prologue): Add early return for naked function.
15226 (riscv_expand_epilogue): Likewise.
15227 (riscv_function_ok_for_sibcall): Return false for naked function.
15228 (riscv_set_current_function): New.
15229 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
15230 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
15231 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
15232 * doc/extend.texi (RISC-V Function Attributes): New.
15234 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
15236 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
15237 check for 128-bit long double before checking TCmode.
15238 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
15239 128-bit long doubles before checking TFmode or TCmode.
15240 (FLOAT128_IBM_P): Likewise.
15242 2018-01-10 Martin Sebor <msebor@redhat.com>
15244 PR tree-optimization/83671
15245 * builtins.c (c_strlen): Unconditionally return zero for the empty
15247 Use -Warray-bounds for warnings.
15248 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
15249 for non-constant array indices with COMPONENT_REF, arrays of
15250 arrays, and pointers to arrays.
15251 (gimple_fold_builtin_strlen): Determine and set length range for
15252 non-constant character arrays.
15254 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
15256 PR middle-end/81897
15257 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
15260 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
15262 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
15264 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15267 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
15268 VECTOR_MEM_ALTIVEC_OR_VSX_P.
15269 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
15270 indexed_or_indirect_operand predicate.
15271 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
15272 (*vsx_le_perm_load_v8hi): Likewise.
15273 (*vsx_le_perm_load_v16qi): Likewise.
15274 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
15275 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
15276 (*vsx_le_perm_store_v8hi): Likewise.
15277 (*vsx_le_perm_store_v16qi): Likewise.
15278 (eight unnamed splitters): Likewise.
15280 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15282 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
15283 * config/rs6000/emmintrin.h: Likewise.
15284 * config/rs6000/mmintrin.h: Likewise.
15285 * config/rs6000/xmmintrin.h: Likewise.
15287 2018-01-10 David Malcolm <dmalcolm@redhat.com>
15290 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
15292 * tree.c (tree_nop_conversion): Return true for location wrapper
15294 (maybe_wrap_with_location): New function.
15295 (selftest::check_strip_nops): New function.
15296 (selftest::test_location_wrappers): New function.
15297 (selftest::tree_c_tests): Call it.
15298 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
15299 (maybe_wrap_with_location): New decl.
15300 (EXPR_LOCATION_WRAPPER_P): New macro.
15301 (location_wrapper_p): New inline function.
15302 (tree_strip_any_location_wrapper): New inline function.
15304 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
15307 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
15308 stack_realign_offset for the largest alignment of stack slot
15310 (ix86_find_max_used_stack_alignment): New function.
15311 (ix86_finalize_stack_frame_flags): Use it. Set
15312 max_used_stack_alignment if we don't realign stack.
15313 * config/i386/i386.h (machine_function): Add
15314 max_used_stack_alignment.
15316 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
15318 * config/arm/arm.opt (-mbranch-cost): New option.
15319 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
15322 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
15325 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
15326 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
15328 2018-01-10 Richard Biener <rguenther@suse.de>
15331 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
15332 early out so it also covers the case where we have a non-NULL
15335 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15337 PR tree-optimization/83753
15338 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
15339 for non-strided grouped accesses if the number of elements is 1.
15341 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15344 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
15345 * i386.h (TARGET_USE_GATHER): Define.
15346 * x86-tune.def (X86_TUNE_USE_GATHER): New.
15348 2018-01-10 Martin Liska <mliska@suse.cz>
15351 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
15352 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
15354 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
15355 CLEANUP_NO_PARTITIONING is not set.
15357 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15359 * doc/rtl.texi: Remove documentation of (const ...) wrappers
15360 for vectors, as a partial revert of r254296.
15361 * rtl.h (const_vec_p): Delete.
15362 (const_vec_duplicate_p): Don't test for vector CONSTs.
15363 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
15364 * expmed.c (make_tree): Likewise.
15367 * common.md (E, F): Use CONSTANT_P instead of checking for
15369 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
15370 checking for CONST_VECTOR.
15372 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15374 PR middle-end/83575
15375 * predict.c (force_edge_cold): Handle in more sane way edges
15376 with no prediction.
15378 2018-01-09 Carl Love <cel@us.ibm.com>
15380 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
15382 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
15383 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
15384 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
15385 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
15386 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
15387 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
15388 * config/rs6000/rs6000-protos.h: Add extern defition for
15389 rs6000_generate_float2_double_code.
15390 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
15392 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
15393 (float2_v2df): Add define_expand.
15395 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
15398 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
15399 op_mode in the force_to_mode call.
15401 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15403 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
15404 instead of checking each element individually.
15405 (aarch64_evpc_uzp): Likewise.
15406 (aarch64_evpc_zip): Likewise.
15407 (aarch64_evpc_ext): Likewise.
15408 (aarch64_evpc_rev): Likewise.
15409 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
15410 instead of checking each element individually. Return true without
15412 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
15413 whether all selected elements come from the same input, instead of
15414 checking each element individually. Remove calls to gen_rtx_REG,
15415 start_sequence and end_sequence and instead assert that no rtl is
15418 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15420 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
15421 order of HIGH and CONST checks.
15423 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15425 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
15426 if the destination isn't an SSA_NAME.
15428 2018-01-09 Richard Biener <rguenther@suse.de>
15430 PR tree-optimization/83668
15431 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
15433 (canonicalize_loop_form): ... here, renamed from ...
15434 (canonicalize_loop_closed_ssa_form): ... this and amended to
15435 swap successor edges for loop exit blocks to make us use
15436 the RPO order we need for initial schedule generation.
15438 2018-01-09 Joseph Myers <joseph@codesourcery.com>
15440 PR tree-optimization/64811
15441 * match.pd: When optimizing comparisons with Inf, avoid
15442 introducing or losing exceptions from comparisons with NaN.
15444 2018-01-09 Martin Liska <mliska@suse.cz>
15447 * asan.c (shadow_mem_size): Add gcc_assert.
15449 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
15451 Don't save registers in main().
15454 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
15455 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
15456 * config/avr/avr.c (avr_set_current_function): Don't error if
15457 naked, OS_task or OS_main are specified at the same time.
15458 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
15460 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
15462 * common/config/avr/avr-common.c (avr_option_optimization_table):
15463 Switch on -mmain-is-OS_task for optimizing compilations.
15465 2018-01-09 Richard Biener <rguenther@suse.de>
15467 PR tree-optimization/83572
15468 * graphite.c: Include cfganal.h.
15469 (graphite_transform_loops): Connect infinite loops to exit
15470 and remove fake edges at the end.
15472 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15474 * ipa-inline.c (edge_badness): Revert accidental checkin.
15476 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15479 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
15480 symbols; not inline clones.
15482 2018-01-09 Jakub Jelinek <jakub@redhat.com>
15485 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
15486 hard registers. Formatting fixes.
15488 PR preprocessor/83722
15489 * gcc.c (try_generate_repro): Pass
15490 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
15491 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
15494 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
15495 Kito Cheng <kito.cheng@gmail.com>
15497 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
15498 (riscv_leaf_function_p): Delete.
15499 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
15501 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15503 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
15505 (do_ifelse): New function.
15506 (do_isel): New function.
15507 (do_sub3): New function.
15508 (do_add3): New function.
15509 (do_load_mask_compare): New function.
15510 (do_overlap_load_compare): New function.
15511 (expand_compare_loop): New function.
15512 (expand_block_compare): Call expand_compare_loop() when appropriate.
15513 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
15514 option description.
15515 (-mblock-compare-inline-loop-limit): New option.
15517 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15520 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
15521 Reverse order of second and third operands in first alternative.
15522 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
15523 of first and second elements in UNSPEC_VPERMR vector.
15524 (altivec_expand_vec_perm_le): Likewise.
15526 2018-01-08 Jeff Law <law@redhat.com>
15528 PR rtl-optimizatin/81308
15529 * tree-switch-conversion.c (cfg_altered): New file scoped static.
15530 (process_switch): If group_case_labels makes a change, then set
15532 (pass_convert_switch::execute): If a switch is converted, then
15533 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
15535 PR rtl-optimization/81308
15536 * recog.c (split_all_insns): Conditionally cleanup the CFG after
15539 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
15541 PR target/83663 - Revert r255946
15542 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
15543 generation for cases where splatting a value is not useful.
15544 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
15545 across a vec_duplicate and a paradoxical subreg forming a vector
15546 mode to a vec_concat.
15548 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15550 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
15551 -march=armv8.3-a variants.
15552 * config/arm/t-multilib: Likewise.
15553 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
15555 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15557 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
15559 (cceq_ior_compare_complement): Give it a name so I can use it, and
15560 change boolean_or_operator predicate to boolean_operator so it can
15561 be used to generate a crand.
15562 (eqne): New code iterator.
15563 (bd/bd_neg): New code_attrs.
15564 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
15565 a single define_insn.
15566 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
15567 decrement (bdnzt/bdnzf/bdzt/bdzf).
15568 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
15569 with the new names of the branch decrement patterns, and added the
15570 names of the branch decrement conditional patterns.
15572 2018-01-08 Richard Biener <rguenther@suse.de>
15574 PR tree-optimization/83563
15575 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
15578 2018-01-08 Richard Biener <rguenther@suse.de>
15580 PR middle-end/83713
15581 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
15583 2018-01-08 Richard Biener <rguenther@suse.de>
15585 PR tree-optimization/83685
15586 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
15587 references to abnormals.
15589 2018-01-08 Richard Biener <rguenther@suse.de>
15592 * dwarf2out.c (output_indirect_strings): Handle empty
15593 skeleton_debug_str_hash.
15594 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
15596 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15598 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
15599 (emit_store_direct): Likewise.
15600 (arc_trampoline_adjust_address): Likewise.
15601 (arc_asm_trampoline_template): New function.
15602 (arc_initialize_trampoline): Use asm_trampoline_template.
15603 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
15604 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
15605 * config/arc/arc.md (flush_icache): Delete pattern.
15607 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15609 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
15610 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
15613 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15616 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
15617 by not USED_FOR_TARGET.
15618 (make_pass_resolve_sw_modes): Likewise.
15620 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15622 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
15625 2018-01-08 Richard Biener <rguenther@suse.de>
15627 PR middle-end/83580
15628 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
15630 2018-01-08 Richard Biener <rguenther@suse.de>
15632 PR middle-end/83517
15633 * match.pd ((t * 2) / 2) -> t): Add missing :c.
15635 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
15637 PR middle-end/81897
15638 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
15639 basic blocks with a small number of successors.
15640 (convert_control_dep_chain_into_preds): Improve handling of
15642 (dump_predicates): Split apart into...
15643 (dump_pred_chain): ...here...
15644 (dump_pred_info): ...and here.
15645 (can_one_predicate_be_invalidated_p): Add debugging printfs.
15646 (can_chain_union_be_invalidated_p): Improve check for invalidation
15648 (uninit_uses_cannot_happen): Avoid unnecessary if
15649 convert_control_dep_chain_into_preds yielded nothing.
15651 2018-01-06 Martin Sebor <msebor@redhat.com>
15653 PR tree-optimization/83640
15654 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
15655 subtracting negative offset from size.
15656 (builtin_access::overlap): Adjust offset bounds of the access to fall
15657 within the size of the object if possible.
15659 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
15661 PR rtl-optimization/83699
15662 * expmed.c (extract_bit_field_1): Restrict the vector usage of
15663 extract_bit_field_as_subreg to cases in which the extracted
15664 value is also a vector.
15666 * lra-constraints.c (process_alt_operands): Test for the equivalence
15667 substitutions when detecting a possible reload cycle.
15669 2018-01-06 Jakub Jelinek <jakub@redhat.com>
15672 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
15673 by default if flag_selective_schedling{,2}. Formatting fixes.
15675 PR rtl-optimization/83682
15676 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
15677 if it has non-VECTOR_MODE element mode.
15678 (vec_duplicate_p): Likewise.
15680 PR middle-end/83694
15681 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
15682 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
15684 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15687 * config/i386/i386-builtin.def
15688 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
15689 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
15690 Require also OPTION_MASK_ISA_AVX512F in addition to
15691 OPTION_MASK_ISA_GFNI.
15692 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
15693 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
15694 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
15695 to OPTION_MASK_ISA_GFNI.
15696 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
15697 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
15698 OPTION_MASK_ISA_AVX512BW.
15699 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
15700 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
15701 addition to OPTION_MASK_ISA_GFNI.
15702 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
15703 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
15704 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
15705 to OPTION_MASK_ISA_GFNI.
15706 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
15707 a requirement for all ISAs rather than any of them with a few
15709 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
15711 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
15712 bitmasks to be enabled with 3 exceptions, instead of requiring any
15713 enabled ISA with lots of exceptions.
15714 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
15715 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
15716 Change avx512bw in isa attribute to avx512f.
15717 * config/i386/sgxintrin.h: Add license boilerplate.
15718 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
15719 to __AVX512F__ and __AVX512VL to __AVX512VL__.
15720 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
15721 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
15723 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
15724 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
15725 temporarily sse2 rather than sse if not enabled already.
15728 * config/i386/sse.md (VI248_VLBW): Rename to ...
15729 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
15730 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
15731 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
15732 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
15733 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
15734 mode iterator instead of VI248_VLBW.
15736 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
15738 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
15739 (record_modified): Skip clobbers; add debug output.
15740 (param_change_prob): Use sreal frequencies.
15742 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15744 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
15745 punt for user-aligned variables.
15747 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15749 * tree-chrec.c (chrec_contains_symbols): Return true for
15752 2018-01-05 Sudakshina Das <sudi.das@arm.com>
15755 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
15756 of (x|y) == x for BICS pattern.
15758 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15760 PR tree-optimization/83605
15761 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
15762 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
15765 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
15767 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
15768 * config/epiphany/rtems.h: New file.
15770 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15771 Uros Bizjak <ubizjak@gmail.com>
15774 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
15775 QIreg_operand instead of register_operand predicate.
15776 * config/i386/i386.c (ix86_rop_should_change_byte_p,
15777 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
15778 comments instead of -fmitigate[-_]rop.
15780 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15783 * cgraphunit.c (symbol_table::compile): Switch to text_section
15784 before calling assembly_start debug hook.
15785 * run-rtl-passes.c (run_rtl_passes): Likewise.
15788 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15790 * tree-vrp.c (extract_range_from_binary_expr_1): Check
15791 range_int_cst_p rather than !symbolic_range_p before calling
15792 extract_range_from_multiplicative_op_1.
15794 2018-01-04 Jeff Law <law@redhat.com>
15796 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
15797 redundant test in assertion.
15799 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15801 * doc/rtl.texi: Document machine_mode wrapper classes.
15803 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15805 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
15806 using tree_to_uhwi.
15808 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15810 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
15811 the VEC_PERM_EXPR fold to fail.
15813 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15816 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
15817 to switched_sections.
15819 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15822 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
15823 test for d.testing.
15825 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
15828 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
15829 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
15831 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15834 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
15835 is BLKmode and bitpos not zero or mode change is needed.
15837 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15840 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
15843 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
15846 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
15847 instead of MULT rtx. Update all corresponding splitters.
15848 (*saddl_se): Ditto.
15849 (*ssub<modesuffix>): Ditto.
15850 (*ssubl_se): Ditto.
15851 (*cmp_sadd_di): Update split patterns.
15852 (*cmp_sadd_si): Ditto.
15853 (*cmp_sadd_sidi): Ditto.
15854 (*cmp_ssub_di): Ditto.
15855 (*cmp_ssub_si): Ditto.
15856 (*cmp_ssub_sidi): Ditto.
15857 * config/alpha/predicates.md (const23_operand): New predicate.
15858 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
15859 Look for ASHIFT, not MULT inner operand.
15860 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
15862 2018-01-04 Martin Liska <mliska@suse.cz>
15864 PR gcov-profile/83669
15865 * gcov.c (output_intermediate_file): Add version to intermediate
15867 * doc/gcov.texi: Document new field 'version' in intermediate
15868 file format. Fix location of '-k' option of gcov command.
15870 2018-01-04 Martin Liska <mliska@suse.cz>
15873 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
15875 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15877 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
15879 2018-01-03 Martin Sebor <msebor@redhat.com>
15881 PR tree-optimization/83655
15882 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
15883 checking calls with invalid arguments.
15885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15887 * tree-vect-stmts.c (vect_get_store_rhs): New function.
15888 (vectorizable_mask_load_store): Delete.
15889 (vectorizable_call): Return false for masked loads and stores.
15890 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
15891 instead of gimple_assign_rhs1.
15892 (vectorizable_load): Handle IFN_MASK_LOAD.
15893 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
15895 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15897 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
15899 (vectorizable_mask_load_store): ...here.
15900 (vectorizable_load): ...and here.
15902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15904 * tree-vect-stmts.c (vect_build_all_ones_mask)
15905 (vect_build_zero_merge_argument): New functions, split out from...
15906 (vectorizable_load): ...here.
15908 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15910 * tree-vect-stmts.c (vect_check_store_rhs): New function,
15912 (vectorizable_mask_load_store): ...here.
15913 (vectorizable_store): ...and here.
15915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15917 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
15919 (vectorizable_mask_load_store): ...here.
15921 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15923 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
15924 (vect_model_store_cost): Take a vec_load_store_type instead of a
15926 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
15927 (vect_model_store_cost): Take a vec_load_store_type instead of a
15929 (vectorizable_mask_load_store): Update accordingly.
15930 (vectorizable_store): Likewise.
15931 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
15933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15935 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
15936 IFN_MASK_LOAD calls here rather than...
15937 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
15939 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15940 Alan Hayward <alan.hayward@arm.com>
15941 David Sherwood <david.sherwood@arm.com>
15943 * expmed.c (extract_bit_field_1): For vector extracts,
15944 fall back to extract_bit_field_as_subreg if vec_extract
15947 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15948 Alan Hayward <alan.hayward@arm.com>
15949 David Sherwood <david.sherwood@arm.com>
15951 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
15952 they are variable or constant sized.
15953 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
15954 slots for constant-sized data.
15956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15957 Alan Hayward <alan.hayward@arm.com>
15958 David Sherwood <david.sherwood@arm.com>
15960 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
15961 handling COND_EXPRs with boolean comparisons, try to find a better
15962 basis for the mask type than the boolean itself.
15964 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15966 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
15967 is calculated and how it can be overridden.
15968 * genmodes.c (max_bitsize_mode_any_mode): New variable.
15969 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
15971 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
15974 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15975 Alan Hayward <alan.hayward@arm.com>
15976 David Sherwood <david.sherwood@arm.com>
15978 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
15979 Remove the mode argument.
15980 (aarch64_simd_valid_immediate): Remove the mode and inverse
15982 * config/aarch64/iterators.md (bitsize): New iterator.
15983 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
15984 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
15985 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
15986 aarch64_simd_valid_immediate.
15987 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
15988 (aarch64_reg_or_bic_imm): Likewise.
15989 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
15990 with an insn_type enum and msl with a modifier_type enum.
15991 Replace element_width with a scalar_mode. Change the shift
15992 to unsigned int. Add constructors for scalar_float_mode and
15993 scalar_int_mode elements.
15994 (aarch64_vect_float_const_representable_p): Delete.
15995 (aarch64_can_const_movi_rtx_p)
15996 (aarch64_simd_scalar_immediate_valid_for_move)
15997 (aarch64_simd_make_constant): Update call to
15998 aarch64_simd_valid_immediate.
15999 (aarch64_advsimd_valid_immediate_hs): New function.
16000 (aarch64_advsimd_valid_immediate): Likewise.
16001 (aarch64_simd_valid_immediate): Remove mode and inverse
16002 arguments. Rewrite to use the above. Use const_vec_duplicate_p
16003 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
16004 and aarch64_float_const_representable_p on the result.
16005 (aarch64_output_simd_mov_immediate): Remove mode argument.
16006 Update call to aarch64_simd_valid_immediate and use of
16007 simd_immediate_info.
16008 (aarch64_output_scalar_simd_mov_immediate): Update call
16011 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16012 Alan Hayward <alan.hayward@arm.com>
16013 David Sherwood <david.sherwood@arm.com>
16015 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
16016 (mode_nunits): Likewise CONST_MODE_NUNITS.
16017 * machmode.def (ADJUST_NUNITS): Document.
16018 * genmodes.c (mode_data::need_nunits_adj): New field.
16019 (blank_mode): Update accordingly.
16020 (adj_nunits): New variable.
16021 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
16023 (emit_mode_size_inline): Set need_bytesize_adj for all modes
16024 listed in adj_nunits.
16025 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
16026 listed in adj_nunits. Don't emit case statements for such modes.
16027 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
16028 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
16029 nothing if adj_nunits is nonnull.
16030 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
16031 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
16032 (emit_mode_fbit): Update use of print_maybe_const_decl.
16033 (emit_move_size): Likewise. Treat the array as non-const
16035 (emit_mode_adjustments): Handle adj_nunits.
16037 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16039 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
16040 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
16041 (VECTOR_MODES): Use it.
16042 (make_vector_modes): Take the prefix as an argument.
16044 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16045 Alan Hayward <alan.hayward@arm.com>
16046 David Sherwood <david.sherwood@arm.com>
16048 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
16049 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
16050 for MODE_VECTOR_BOOL.
16051 * machmode.def (VECTOR_BOOL_MODE): Document.
16052 * genmodes.c (VECTOR_BOOL_MODE): New macro.
16053 (make_vector_bool_mode): New function.
16054 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
16056 * lto-streamer-in.c (lto_input_mode_table): Likewise.
16057 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
16059 * stor-layout.c (int_mode_for_mode): Likewise.
16060 * tree.c (build_vector_type_for_mode): Likewise.
16061 * varasm.c (output_constant_pool_2): Likewise.
16062 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
16063 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
16064 for MODE_VECTOR_BOOL.
16065 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
16066 of mode class checks.
16067 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
16068 instead of a list of mode class checks.
16069 (expand_vector_scalar_condition): Likewise.
16070 (type_for_widest_vector_mode): Handle BImode as an inner mode.
16072 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16073 Alan Hayward <alan.hayward@arm.com>
16074 David Sherwood <david.sherwood@arm.com>
16076 * machmode.h (mode_size): Change from unsigned short to
16078 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
16079 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16080 or if measurement_type is not polynomial.
16081 (fixed_size_mode::includes_p): Check for constant-sized modes.
16082 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
16083 return a poly_uint16 rather than an unsigned short.
16084 (emit_mode_size): Change the type of mode_size from unsigned short
16085 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
16086 (emit_mode_adjustments): Cope with polynomial vector sizes.
16087 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16089 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16091 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
16092 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
16093 * caller-save.c (setup_save_areas): Likewise.
16094 (replace_reg_with_saved_mem): Likewise.
16095 * calls.c (emit_library_call_value_1): Likewise.
16096 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
16097 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
16098 (gen_lowpart_for_combine): Likewise.
16099 * convert.c (convert_to_integer_1): Likewise.
16100 * cse.c (equiv_constant, cse_insn): Likewise.
16101 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
16102 (cselib_subst_to_values): Likewise.
16103 * dce.c (word_dce_process_block): Likewise.
16104 * df-problems.c (df_word_lr_mark_ref): Likewise.
16105 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
16106 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
16107 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
16108 (rtl_for_decl_location): Likewise.
16109 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
16110 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
16111 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
16112 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
16113 (expand_expr_real_1): Likewise.
16114 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
16115 (pad_below): Likewise.
16116 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16117 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
16118 * ira.c (get_subreg_tracking_sizes): Likewise.
16119 * ira-build.c (ira_create_allocno_objects): Likewise.
16120 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
16121 (ira_sort_regnos_for_alter_reg): Likewise.
16122 * ira-costs.c (record_operand_costs): Likewise.
16123 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
16124 (resolve_simple_move): Likewise.
16125 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
16126 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
16127 (lra_constraints): Likewise.
16128 (CONST_POOL_OK_P): Reject variable-sized modes.
16129 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
16130 (add_pseudo_to_slot, lra_spill): Likewise.
16131 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16132 * optabs-query.c (get_best_extraction_insn): Likewise.
16133 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16134 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
16135 (expand_mult_highpart, valid_multiword_target_p): Likewise.
16136 * recog.c (offsettable_address_addr_space_p): Likewise.
16137 * regcprop.c (maybe_mode_change): Likewise.
16138 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
16139 * regrename.c (build_def_use): Likewise.
16140 * regstat.c (dump_reg_info): Likewise.
16141 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
16142 (find_reloads, find_reloads_subreg_address): Likewise.
16143 * reload1.c (eliminate_regs_1): Likewise.
16144 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
16145 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
16146 (simplify_binary_operation_1, simplify_subreg): Likewise.
16147 * targhooks.c (default_function_arg_padding): Likewise.
16148 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
16149 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
16150 (verify_gimple_assign_ternary): Likewise.
16151 * tree-inline.c (estimate_move_cost): Likewise.
16152 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16153 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
16154 (get_address_cost_ainc): Likewise.
16155 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
16156 (vect_supportable_dr_alignment): Likewise.
16157 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16158 (vectorizable_reduction): Likewise.
16159 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
16160 (vectorizable_operation, vectorizable_load): Likewise.
16161 * tree.c (build_same_sized_truth_vector_type): Likewise.
16162 * valtrack.c (cleanup_auto_inc_dec): Likewise.
16163 * var-tracking.c (emit_note_insn_var_location): Likewise.
16164 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
16165 (ADDR_VEC_ALIGN): Likewise.
16167 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16168 Alan Hayward <alan.hayward@arm.com>
16169 David Sherwood <david.sherwood@arm.com>
16171 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
16173 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16174 or if measurement_type is polynomial.
16175 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
16176 * combine.c (make_extraction): Likewise.
16177 * dse.c (find_shift_sequence): Likewise.
16178 * dwarf2out.c (mem_loc_descriptor): Likewise.
16179 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
16180 (extract_bit_field, extract_low_bits): Likewise.
16181 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
16182 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
16183 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
16184 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
16185 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16186 * reload.c (find_reloads): Likewise.
16187 * reload1.c (alter_reg): Likewise.
16188 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
16189 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
16190 * tree-if-conv.c (predicate_mem_writes): Likewise.
16191 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
16192 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
16193 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
16194 * valtrack.c (dead_debug_insert_temp): Likewise.
16195 * varasm.c (mergeable_constant_section): Likewise.
16196 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
16198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16199 Alan Hayward <alan.hayward@arm.com>
16200 David Sherwood <david.sherwood@arm.com>
16202 * expr.c (expand_assignment): Cope with polynomial mode sizes
16203 when assigning to a CONCAT.
16205 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16206 Alan Hayward <alan.hayward@arm.com>
16207 David Sherwood <david.sherwood@arm.com>
16209 * machmode.h (mode_precision): Change from unsigned short to
16211 (mode_to_precision): Return a poly_uint16 rather than an unsigned
16213 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
16214 or if measurement_type is not polynomial.
16215 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
16216 in which the mode is already known to be a scalar_int_mode.
16217 * genmodes.c (emit_mode_precision): Change the type of mode_precision
16218 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
16220 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16221 for GET_MODE_PRECISION.
16222 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16223 for GET_MODE_PRECISION.
16224 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
16226 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
16227 (expand_field_assignment, make_extraction): Likewise.
16228 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
16229 (get_last_value): Likewise.
16230 * convert.c (convert_to_integer_1): Likewise.
16231 * cse.c (cse_insn): Likewise.
16232 * expr.c (expand_expr_real_1): Likewise.
16233 * lra-constraints.c (simplify_operand_subreg): Likewise.
16234 * optabs-query.c (can_atomic_load_p): Likewise.
16235 * optabs.c (expand_atomic_load): Likewise.
16236 (expand_atomic_store): Likewise.
16237 * ree.c (combine_reaching_defs): Likewise.
16238 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
16239 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
16240 * tree.h (type_has_mode_precision_p): Likewise.
16241 * ubsan.c (instrument_si_overflow): Likewise.
16243 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16244 Alan Hayward <alan.hayward@arm.com>
16245 David Sherwood <david.sherwood@arm.com>
16247 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
16248 polynomial numbers of units.
16249 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
16250 (valid_vector_subparts_p): New function.
16251 (build_vector_type): Remove temporary shim and take the number
16252 of units as a poly_uint64 rather than an int.
16253 (build_opaque_vector_type): Take the number of units as a
16254 poly_uint64 rather than an int.
16255 * tree.c (build_vector_from_ctor): Handle polynomial
16256 TYPE_VECTOR_SUBPARTS.
16257 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
16258 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
16259 (build_vector_from_val): If the number of units is variable,
16260 use build_vec_duplicate_cst for constant operands and
16261 VEC_DUPLICATE_EXPR otherwise.
16262 (make_vector_type): Remove temporary is_constant ().
16263 (build_vector_type, build_opaque_vector_type): Take the number of
16264 units as a poly_uint64 rather than an int.
16265 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
16267 * cfgexpand.c (expand_debug_expr): Likewise.
16268 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
16269 (store_constructor, expand_expr_real_1): Likewise.
16270 (const_scalar_mask_from_tree): Likewise.
16271 * fold-const-call.c (fold_const_reduction): Likewise.
16272 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
16273 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
16274 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
16275 (fold_relational_const): Likewise.
16276 (native_interpret_vector): Likewise. Change the size from an
16277 int to an unsigned int.
16278 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
16279 TYPE_VECTOR_SUBPARTS.
16280 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
16281 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
16282 duplicating a non-constant operand into a variable-length vector.
16283 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
16284 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
16285 * ipa-icf.c (sem_variable::equals): Likewise.
16286 * match.pd: Likewise.
16287 * omp-simd-clone.c (simd_clone_subparts): Likewise.
16288 * print-tree.c (print_node): Likewise.
16289 * stor-layout.c (layout_type): Likewise.
16290 * targhooks.c (default_builtin_vectorization_cost): Likewise.
16291 * tree-cfg.c (verify_gimple_comparison): Likewise.
16292 (verify_gimple_assign_binary): Likewise.
16293 (verify_gimple_assign_ternary): Likewise.
16294 (verify_gimple_assign_single): Likewise.
16295 * tree-pretty-print.c (dump_generic_node): Likewise.
16296 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16297 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
16298 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
16299 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
16300 (vect_shift_permute_load_chain): Likewise.
16301 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
16302 (expand_vector_condition, optimize_vector_constructor): Likewise.
16303 (lower_vec_perm, get_compute_type): Likewise.
16304 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16305 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
16306 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
16307 (vect_recog_mask_conversion_pattern): Likewise.
16308 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
16309 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
16310 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16311 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
16312 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
16313 (vectorizable_shift, vectorizable_operation, vectorizable_store)
16314 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
16315 (supportable_widening_operation): Likewise.
16316 (supportable_narrowing_operation): Likewise.
16317 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
16319 * varasm.c (output_constant): Likewise.
16321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16322 Alan Hayward <alan.hayward@arm.com>
16323 David Sherwood <david.sherwood@arm.com>
16325 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
16326 so that both the length == 3 and length != 3 cases set up their
16327 own permute vectors. Add comments explaining why we know the
16328 number of elements is constant.
16329 (vect_permute_load_chain): Likewise.
16331 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16332 Alan Hayward <alan.hayward@arm.com>
16333 David Sherwood <david.sherwood@arm.com>
16335 * machmode.h (mode_nunits): Change from unsigned char to
16337 (ONLY_FIXED_SIZE_MODES): New macro.
16338 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
16339 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
16340 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
16342 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
16343 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
16344 or if measurement_type is not polynomial.
16345 * genmodes.c (ZERO_COEFFS): New macro.
16346 (emit_mode_nunits_inline): Make mode_nunits_inline return a
16348 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
16349 Use ZERO_COEFFS when emitting initializers.
16350 * data-streamer.h (bp_pack_poly_value): New function.
16351 (bp_unpack_poly_value): Likewise.
16352 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16353 for GET_MODE_NUNITS.
16354 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16355 for GET_MODE_NUNITS.
16356 * tree.c (make_vector_type): Remove temporary shim and make
16357 the real function take the number of units as a poly_uint64
16358 rather than an int.
16359 (build_vector_type_for_mode): Handle polynomial nunits.
16360 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
16361 * emit-rtl.c (const_vec_series_p_1): Likewise.
16362 (gen_rtx_CONST_VECTOR): Likewise.
16363 * fold-const.c (test_vec_duplicate_folding): Likewise.
16364 * genrecog.c (validate_pattern): Likewise.
16365 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
16366 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16367 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
16368 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
16369 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
16370 * rtlanal.c (subreg_get_info): Likewise.
16371 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16372 (vect_grouped_load_supported): Likewise.
16373 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
16374 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
16375 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
16376 (simplify_const_unary_operation, simplify_binary_operation_1)
16377 (simplify_const_binary_operation, simplify_ternary_operation)
16378 (test_vector_ops_duplicate, test_vector_ops): Likewise.
16379 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
16380 instead of CONST_VECTOR_NUNITS.
16381 * varasm.c (output_constant_pool_2): Likewise.
16382 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
16383 explicit-encoded elements in the XVEC for variable-length vectors.
16385 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16387 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
16389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16390 Alan Hayward <alan.hayward@arm.com>
16391 David Sherwood <david.sherwood@arm.com>
16393 * coretypes.h (fixed_size_mode): Declare.
16394 (fixed_size_mode_pod): New typedef.
16395 * builtins.h (target_builtins::x_apply_args_mode)
16396 (target_builtins::x_apply_result_mode): Change type to
16397 fixed_size_mode_pod.
16398 * builtins.c (apply_args_size, apply_result_size, result_vector)
16399 (expand_builtin_apply_args_1, expand_builtin_apply)
16400 (expand_builtin_return): Update accordingly.
16402 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16404 * cse.c (hash_rtx_cb): Hash only the encoded elements.
16405 * cselib.c (cselib_hash_rtx): Likewise.
16406 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
16407 CONST_VECTOR encoding.
16409 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16410 Jeff Law <law@redhat.com>
16413 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
16414 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
16415 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
16416 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
16419 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
16420 explicitly probe *sp in a noreturn function if there were any callee
16421 register saves or frame pointer is needed.
16423 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16426 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
16427 BLKmode for ternary, binary or unary expressions.
16430 * var-tracking.c (delete_vta_debug_insn): New inline function.
16431 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
16432 insns from get_insns () to NULL instead of each bb separately.
16433 Use delete_vta_debug_insn. No longer static.
16434 (vt_debug_insns_local, variable_tracking_main_1): Adjust
16435 delete_vta_debug_insns callers.
16436 * rtl.h (delete_vta_debug_insns): Declare.
16437 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
16438 instead of variable_tracking_main.
16440 2018-01-03 Martin Sebor <msebor@redhat.com>
16442 PR tree-optimization/83603
16443 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
16444 arguments past the endof the argument list in functions declared
16445 without a prototype.
16446 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
16447 Avoid checking when arguments are null.
16449 2018-01-03 Martin Sebor <msebor@redhat.com>
16452 * doc/extend.texi (attribute const): Fix a typo.
16453 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
16454 issuing -Wsuggest-attribute for void functions.
16456 2018-01-03 Martin Sebor <msebor@redhat.com>
16458 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
16459 offset_int::from instead of wide_int::to_shwi.
16460 (maybe_diag_overlap): Remove assertion.
16461 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
16462 * gimple-ssa-sprintf.c (format_directive): Same.
16463 (parse_directive): Same.
16464 (sprintf_dom_walker::compute_format_length): Same.
16465 (try_substitute_return_value): Same.
16467 2018-01-03 Jeff Law <law@redhat.com>
16469 PR middle-end/83654
16470 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
16471 non-constant residual for zero at runtime and avoid probing in
16472 that case. Reorganize code for trailing problem to mirror handling
16475 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16477 PR tree-optimization/83501
16478 * tree-ssa-strlen.c (get_string_cst): New.
16479 (handle_char_store): Call get_string_cst.
16481 2018-01-03 Martin Liska <mliska@suse.cz>
16483 PR tree-optimization/83593
16484 * tree-ssa-strlen.c: Include tree-cfg.h.
16485 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
16486 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
16487 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
16489 (strlen_dom_walker::before_dom_children): Call
16490 gimple_purge_dead_eh_edges. Dump tranformation with details
16492 (strlen_dom_walker::before_dom_children): Update call by adding
16493 new argument cleanup_eh.
16494 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
16496 2018-01-03 Martin Liska <mliska@suse.cz>
16499 * cif-code.def (VARIADIC_THUNK): New enum value.
16500 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
16503 2018-01-03 Jan Beulich <jbeulich@suse.com>
16505 * sse.md (mov<mode>_internal): Tighten condition for when to use
16506 vmovdqu<ssescalarsize> for TI and OI modes.
16508 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16510 Update copyright years.
16512 2018-01-03 Martin Liska <mliska@suse.cz>
16515 * ipa-visibility.c (function_and_variable_visibility): Skip
16516 functions with noipa attribure.
16518 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16520 * gcc.c (process_command): Update copyright notice dates.
16521 * gcov-dump.c (print_version): Ditto.
16522 * gcov.c (print_version): Ditto.
16523 * gcov-tool.c (print_version): Ditto.
16524 * gengtype.c (create_file): Ditto.
16525 * doc/cpp.texi: Bump @copying's copyright year.
16526 * doc/cppinternals.texi: Ditto.
16527 * doc/gcc.texi: Ditto.
16528 * doc/gccint.texi: Ditto.
16529 * doc/gcov.texi: Ditto.
16530 * doc/install.texi: Ditto.
16531 * doc/invoke.texi: Ditto.
16533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16535 * vector-builder.h (vector_builder::m_full_nelts): Change from
16536 unsigned int to poly_uint64.
16537 (vector_builder::full_nelts): Update prototype accordingly.
16538 (vector_builder::new_vector): Likewise.
16539 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
16540 (vector_builder::operator ==): Likewise.
16541 (vector_builder::finalize): Likewise.
16542 * int-vector-builder.h (int_vector_builder::int_vector_builder):
16543 Take the number of elements as a poly_uint64 rather than an
16545 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
16546 from unsigned int to poly_uint64.
16547 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
16548 (vec_perm_indices::new_vector): Likewise.
16549 (vec_perm_indices::length): Likewise.
16550 (vec_perm_indices::nelts_per_input): Likewise.
16551 (vec_perm_indices::input_nelts): Likewise.
16552 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
16553 number of elements per input as a poly_uint64 rather than an
16554 unsigned int. Use the original encoding for variable-length
16555 vectors, rather than clamping each individual element.
16556 For the second and subsequent elements in each pattern,
16557 clamp the step and base before clamping their sum.
16558 (vec_perm_indices::series_p): Handle polynomial element counts.
16559 (vec_perm_indices::all_in_range_p): Likewise.
16560 (vec_perm_indices_to_tree): Likewise.
16561 (vec_perm_indices_to_rtx): Likewise.
16562 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
16563 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
16564 (tree_vector_builder::new_binary_operation): Handle polynomial
16565 element counts. Return false if we need to know the number
16566 of elements at compile time.
16567 * fold-const.c (fold_vec_perm): Punt if the number of elements
16568 isn't known at compile time.
16570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16572 * vec-perm-indices.h (vec_perm_builder): Change element type
16573 from HOST_WIDE_INT to poly_int64.
16574 (vec_perm_indices::element_type): Update accordingly.
16575 (vec_perm_indices::clamp): Handle polynomial element_types.
16576 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16577 (vec_perm_indices::all_in_range_p): Likewise.
16578 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
16580 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
16581 polynomial vec_perm_indices element types.
16582 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
16583 * fold-const.c (fold_vec_perm): Likewise.
16584 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
16585 * tree-vect-generic.c (lower_vec_perm): Likewise.
16586 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16587 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
16588 element type to HOST_WIDE_INT.
16590 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16591 Alan Hayward <alan.hayward@arm.com>
16592 David Sherwood <david.sherwood@arm.com>
16594 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
16595 rather than an int. Use plus_constant.
16596 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
16597 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
16599 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16600 Alan Hayward <alan.hayward@arm.com>
16601 David Sherwood <david.sherwood@arm.com>
16603 * calls.c (emit_call_1, expand_call): Change struct_value_size from
16604 a HOST_WIDE_INT to a poly_int64.
16606 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16607 Alan Hayward <alan.hayward@arm.com>
16608 David Sherwood <david.sherwood@arm.com>
16610 * calls.c (load_register_parameters): Cope with polynomial
16611 mode sizes. Require a constant size for BLKmode parameters
16612 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
16613 forces a parameter to be padded at the lsb end in order to
16614 fill a complete number of words, require the parameter size
16615 to be ordered wrt UNITS_PER_WORD.
16617 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16618 Alan Hayward <alan.hayward@arm.com>
16619 David Sherwood <david.sherwood@arm.com>
16621 * reload1.c (spill_stack_slot_width): Change element type
16622 from unsigned int to poly_uint64_pod.
16623 (alter_reg): Treat mode sizes as polynomial.
16625 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16626 Alan Hayward <alan.hayward@arm.com>
16627 David Sherwood <david.sherwood@arm.com>
16629 * reload.c (complex_word_subreg_p): New function.
16630 (reload_inner_reg_of_subreg, push_reload): Use it.
16632 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16633 Alan Hayward <alan.hayward@arm.com>
16634 David Sherwood <david.sherwood@arm.com>
16636 * lra-constraints.c (process_alt_operands): Reject matched
16637 operands whose sizes aren't ordered.
16638 (match_reload): Refer to this check here.
16640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16641 Alan Hayward <alan.hayward@arm.com>
16642 David Sherwood <david.sherwood@arm.com>
16644 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
16645 that the mode size is in the set {1, 2, 4, 8, 16}.
16647 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16648 Alan Hayward <alan.hayward@arm.com>
16649 David Sherwood <david.sherwood@arm.com>
16651 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
16652 Use plus_constant instead of gen_rtx_PLUS.
16654 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16655 Alan Hayward <alan.hayward@arm.com>
16656 David Sherwood <david.sherwood@arm.com>
16658 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
16659 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
16660 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
16661 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
16662 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
16663 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
16664 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
16665 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
16666 * config/i386/i386.c (ix86_push_rounding): ...this new function.
16667 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
16669 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
16670 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
16671 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
16672 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
16673 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
16674 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
16675 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
16676 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
16677 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
16678 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
16680 * expr.c (emit_move_resolve_push): Treat the input and result
16681 of PUSH_ROUNDING as a poly_int64.
16682 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
16683 (emit_push_insn): Likewise.
16684 * lra-eliminations.c (mark_not_eliminable): Likewise.
16685 * recog.c (push_operand): Likewise.
16686 * reload1.c (elimination_effects): Likewise.
16687 * rtlanal.c (nonzero_bits1): Likewise.
16688 * calls.c (store_one_arg): Likewise. Require the padding to be
16689 known at compile time.
16691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16692 Alan Hayward <alan.hayward@arm.com>
16693 David Sherwood <david.sherwood@arm.com>
16695 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
16696 Use plus_constant instead of gen_rtx_PLUS.
16698 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16699 Alan Hayward <alan.hayward@arm.com>
16700 David Sherwood <david.sherwood@arm.com>
16702 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
16703 rather than an int.
16705 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16706 Alan Hayward <alan.hayward@arm.com>
16707 David Sherwood <david.sherwood@arm.com>
16709 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
16710 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
16711 via stack temporaries. Treat the mode size as polynomial too.
16713 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16714 Alan Hayward <alan.hayward@arm.com>
16715 David Sherwood <david.sherwood@arm.com>
16717 * expr.c (expand_expr_real_2): When handling conversions involving
16718 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
16719 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
16720 as a poly_uint64 too.
16722 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16723 Alan Hayward <alan.hayward@arm.com>
16724 David Sherwood <david.sherwood@arm.com>
16726 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
16728 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16729 Alan Hayward <alan.hayward@arm.com>
16730 David Sherwood <david.sherwood@arm.com>
16732 * combine.c (can_change_dest_mode): Handle polynomial
16733 REGMODE_NATURAL_SIZE.
16734 * expmed.c (store_bit_field_1): Likewise.
16735 * expr.c (store_constructor): Likewise.
16736 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
16737 and polynomial REGMODE_NATURAL_SIZE.
16738 (gen_lowpart_common): Likewise.
16739 * reginfo.c (record_subregs_of_mode): Likewise.
16740 * rtlanal.c (read_modify_subreg_p): Likewise.
16742 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16743 Alan Hayward <alan.hayward@arm.com>
16744 David Sherwood <david.sherwood@arm.com>
16746 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
16747 numbers of elements.
16749 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16750 Alan Hayward <alan.hayward@arm.com>
16751 David Sherwood <david.sherwood@arm.com>
16753 * match.pd: Cope with polynomial numbers of vector elements.
16755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16756 Alan Hayward <alan.hayward@arm.com>
16757 David Sherwood <david.sherwood@arm.com>
16759 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
16760 in a POINTER_PLUS_EXPR.
16762 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16763 Alan Hayward <alan.hayward@arm.com>
16764 David Sherwood <david.sherwood@arm.com>
16766 * omp-simd-clone.c (simd_clone_subparts): New function.
16767 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
16768 (ipa_simd_modify_function_body): Likewise.
16770 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16771 Alan Hayward <alan.hayward@arm.com>
16772 David Sherwood <david.sherwood@arm.com>
16774 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
16775 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
16776 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
16777 (expand_vector_condition, vector_element): Likewise.
16778 (subparts_gt): New function.
16779 (get_compute_type): Use subparts_gt.
16780 (count_type_subparts): Delete.
16781 (expand_vector_operations_1): Use subparts_gt instead of
16782 count_type_subparts.
16784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16785 Alan Hayward <alan.hayward@arm.com>
16786 David Sherwood <david.sherwood@arm.com>
16788 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
16789 (vect_compile_time_alias): ...this new function. Do the calculation
16790 on poly_ints rather than trees.
16791 (vect_prune_runtime_alias_test_list): Update call accordingly.
16793 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16794 Alan Hayward <alan.hayward@arm.com>
16795 David Sherwood <david.sherwood@arm.com>
16797 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
16799 (vect_schedule_slp_instance): Likewise.
16801 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16802 Alan Hayward <alan.hayward@arm.com>
16803 David Sherwood <david.sherwood@arm.com>
16805 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
16806 constant and extern definitions for variable-length vectors.
16807 (vect_get_constant_vectors): Note that the number of units
16808 is known to be constant.
16810 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16811 Alan Hayward <alan.hayward@arm.com>
16812 David Sherwood <david.sherwood@arm.com>
16814 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
16815 of units as polynomial. Choose between WIDE and NARROW based
16818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16819 Alan Hayward <alan.hayward@arm.com>
16820 David Sherwood <david.sherwood@arm.com>
16822 * tree-vect-stmts.c (simd_clone_subparts): New function.
16823 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
16825 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16826 Alan Hayward <alan.hayward@arm.com>
16827 David Sherwood <david.sherwood@arm.com>
16829 * tree-vect-stmts.c (vectorizable_call): Treat the number of
16830 vectors as polynomial. Use build_index_vector for
16831 IFN_GOMP_SIMD_LANE.
16833 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16834 Alan Hayward <alan.hayward@arm.com>
16835 David Sherwood <david.sherwood@arm.com>
16837 * tree-vect-stmts.c (get_load_store_type): Treat the number of
16838 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
16839 for variable-length vectors.
16840 (vectorizable_mask_load_store): Treat the number of units as
16841 polynomial, asserting that it is constant if the condition has
16842 already been enforced.
16843 (vectorizable_store, vectorizable_load): Likewise.
16845 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16846 Alan Hayward <alan.hayward@arm.com>
16847 David Sherwood <david.sherwood@arm.com>
16849 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
16850 of units as polynomial. Punt if we can't tell at compile time
16851 which vector contains the final result.
16853 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16854 Alan Hayward <alan.hayward@arm.com>
16855 David Sherwood <david.sherwood@arm.com>
16857 * tree-vect-loop.c (vectorizable_induction): Treat the number
16858 of units as polynomial. Punt on SLP inductions. Use an integer
16859 VEC_SERIES_EXPR for variable-length integer reductions. Use a
16860 cast of such a series for variable-length floating-point
16863 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16864 Alan Hayward <alan.hayward@arm.com>
16865 David Sherwood <david.sherwood@arm.com>
16867 * tree.h (build_index_vector): Declare.
16868 * tree.c (build_index_vector): New function.
16869 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
16870 of units as polynomial, forcibly converting it to a constant if
16871 vectorizable_reduction has already enforced the condition.
16872 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
16873 to create a {1,2,3,...} vector.
16874 (vectorizable_reduction): Treat the number of units as polynomial.
16875 Choose vectype_in based on the largest scalar element size rather
16876 than the smallest number of units. Enforce the restrictions
16879 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16880 Alan Hayward <alan.hayward@arm.com>
16881 David Sherwood <david.sherwood@arm.com>
16883 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
16884 number of units as polynomial.
16886 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16887 Alan Hayward <alan.hayward@arm.com>
16888 David Sherwood <david.sherwood@arm.com>
16890 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
16891 * target.def (autovectorize_vector_sizes): Return the vector sizes
16892 by pointer, using vector_sizes rather than a bitmask.
16893 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
16894 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
16895 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
16897 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
16898 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
16899 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
16900 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
16901 * omp-general.c (omp_max_vf): Likewise.
16902 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16903 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
16904 * tree-vect-loop.c (vect_analyze_loop): Likewise.
16905 * tree-vect-slp.c (vect_slp_bb): Likewise.
16906 * doc/tm.texi: Regenerate.
16907 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
16909 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
16910 the vector size as a poly_uint64 rather than an unsigned int.
16911 (current_vector_size): Change from an unsigned int to a poly_uint64.
16912 (get_vectype_for_scalar_type): Update accordingly.
16913 * tree.h (build_truth_vector_type): Take the size and number of
16914 units as a poly_uint64 rather than an unsigned int.
16915 (build_vector_type): Add a temporary overload that takes
16916 the number of units as a poly_uint64 rather than an unsigned int.
16917 * tree.c (make_vector_type): Likewise.
16918 (build_truth_vector_type): Take the number of units as a poly_uint64
16919 rather than an unsigned int.
16921 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16922 Alan Hayward <alan.hayward@arm.com>
16923 David Sherwood <david.sherwood@arm.com>
16925 * target.def (get_mask_mode): Take the number of units and length
16926 as poly_uint64s rather than unsigned ints.
16927 * targhooks.h (default_get_mask_mode): Update accordingly.
16928 * targhooks.c (default_get_mask_mode): Likewise.
16929 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
16930 * doc/tm.texi: Regenerate.
16932 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16933 Alan Hayward <alan.hayward@arm.com>
16934 David Sherwood <david.sherwood@arm.com>
16936 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
16937 * omp-general.c (omp_max_vf): Likewise.
16938 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
16939 (expand_omp_simd): Handle polynomial safelen.
16940 * omp-low.c (omplow_simd_context): Add a default constructor.
16941 (omplow_simd_context::max_vf): Change from int to poly_uint64.
16942 (lower_rec_simd_input_clauses): Update accordingly.
16943 (lower_rec_input_clauses): Likewise.
16945 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16946 Alan Hayward <alan.hayward@arm.com>
16947 David Sherwood <david.sherwood@arm.com>
16949 * tree-vectorizer.h (vect_nunits_for_cost): New function.
16950 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
16951 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
16952 (vect_analyze_slp_cost): Likewise.
16953 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
16954 (vect_model_load_cost): Likewise.
16956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16957 Alan Hayward <alan.hayward@arm.com>
16958 David Sherwood <david.sherwood@arm.com>
16960 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
16961 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
16962 from an unsigned int * to a poly_uint64_pod *.
16963 (calculate_unrolling_factor): New function.
16964 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
16966 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16967 Alan Hayward <alan.hayward@arm.com>
16968 David Sherwood <david.sherwood@arm.com>
16970 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
16971 from an unsigned int to a poly_uint64.
16972 (_loop_vec_info::slp_unrolling_factor): Likewise.
16973 (_loop_vec_info::vectorization_factor): Change from an int
16975 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
16976 (vect_get_num_vectors): New function.
16977 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
16978 (vect_get_num_copies): Use vect_get_num_vectors.
16979 (vect_analyze_data_ref_dependences): Change max_vf from an int *
16980 to an unsigned int *.
16981 (vect_analyze_data_refs): Change min_vf from an int * to a
16983 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16984 than an unsigned HOST_WIDE_INT.
16985 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
16986 (vect_analyze_data_ref_dependence): Change max_vf from an int *
16987 to an unsigned int *.
16988 (vect_analyze_data_ref_dependences): Likewise.
16989 (vect_compute_data_ref_alignment): Handle polynomial vf.
16990 (vect_enhance_data_refs_alignment): Likewise.
16991 (vect_prune_runtime_alias_test_list): Likewise.
16992 (vect_shift_permute_load_chain): Likewise.
16993 (vect_supportable_dr_alignment): Likewise.
16994 (dependence_distance_ge_vf): Take the vectorization factor as a
16995 poly_uint64 rather than an unsigned HOST_WIDE_INT.
16996 (vect_analyze_data_refs): Change min_vf from an int * to a
16998 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
16999 vfm1 as a poly_uint64 rather than an int. Make the same change
17000 for the returned bound_scalar.
17001 (vect_gen_vector_loop_niters): Handle polynomial vf.
17002 (vect_do_peeling): Likewise. Update call to
17003 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
17004 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
17006 * tree-vect-loop.c (vect_determine_vectorization_factor)
17007 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
17008 (vect_get_known_peeling_cost): Likewise.
17009 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
17010 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
17011 (vect_transform_loop): Likewise. Use the lowest possible VF when
17012 updating the upper bounds of the loop.
17013 (vect_min_worthwhile_factor): Make static. Return an unsigned int
17014 rather than an int.
17015 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
17016 polynomial unroll factors.
17017 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
17018 (vect_make_slp_decision): Likewise.
17019 (vect_supported_load_permutation_p): Likewise, and polynomial
17021 (vect_analyze_slp_cost): Handle polynomial vf.
17022 (vect_slp_analyze_node_operations): Likewise.
17023 (vect_slp_analyze_bb_1): Likewise.
17024 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17025 than an unsigned HOST_WIDE_INT.
17026 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
17027 (vectorizable_load): Handle polynomial vf.
17028 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
17030 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
17032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17033 Alan Hayward <alan.hayward@arm.com>
17034 David Sherwood <david.sherwood@arm.com>
17036 * match.pd: Handle bit operations involving three constants
17037 and try to fold one pair.
17039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17041 * tree-vect-loop-manip.c: Include gimple-fold.h.
17042 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
17043 niters_maybe_zero parameters. Handle other cases besides a step of 1.
17044 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
17045 Add a path that uses a step of VF instead of 1, but disable it
17047 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
17048 and niters_no_overflow parameters. Update calls to
17049 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
17050 Create a new SSA name if the latter choses to use a ste other
17051 than zero, and return it via niters_vector_mult_vf_var.
17052 * tree-vect-loop.c (vect_transform_loop): Update calls to
17053 vect_do_peeling, vect_gen_vector_loop_niters and
17054 slpeel_make_loop_iterate_ntimes.
17055 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
17056 (vect_gen_vector_loop_niters): Update declarations after above changes.
17058 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
17060 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
17061 128-bit round to integer instructions.
17062 (ceil<mode>2): Likewise.
17063 (btrunc<mode>2): Likewise.
17064 (round<mode>2): Likewise.
17066 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17068 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
17069 unaligned VSX load/store on P8/P9.
17070 (expand_block_clear): Allow the use of unaligned VSX
17071 load/store on P8/P9.
17073 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17075 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
17077 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
17078 swap associated with both a load and a store.
17080 2018-01-02 Andrew Waterman <andrew@sifive.com>
17082 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
17083 * config/riscv/riscv.md (clear_cache): Use it.
17085 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
17087 * web.c: Remove out-of-date comment.
17089 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17091 * expr.c (fixup_args_size_notes): Check that any existing
17092 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
17093 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
17094 (emit_single_push_insn): ...here.
17096 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17098 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
17099 (const_vector_encoded_nelts): New function.
17100 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
17101 (const_vector_int_elt, const_vector_elt): Declare.
17102 * emit-rtl.c (const_vector_int_elt_1): New function.
17103 (const_vector_elt): Likewise.
17104 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
17105 of CONST_VECTOR_ELT.
17107 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17109 * expr.c: Include rtx-vector-builder.h.
17110 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
17111 directly on the tree encoding.
17112 (const_vector_from_tree): Likewise.
17113 * optabs.c: Include rtx-vector-builder.h.
17114 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
17115 sequence of "u" values.
17116 * vec-perm-indices.c: Include rtx-vector-builder.h.
17117 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
17118 directly on the vec_perm_indices encoding.
17120 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17122 * doc/rtl.texi (const_vector): Describe new encoding scheme.
17123 * Makefile.in (OBJS): Add rtx-vector-builder.o.
17124 * rtx-vector-builder.h: New file.
17125 * rtx-vector-builder.c: Likewise.
17126 * rtl.h (rtx_def::u2): Add a const_vector field.
17127 (CONST_VECTOR_NPATTERNS): New macro.
17128 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
17129 (CONST_VECTOR_DUPLICATE_P): Likewise.
17130 (CONST_VECTOR_STEPPED_P): Likewise.
17131 (CONST_VECTOR_ENCODED_ELT): Likewise.
17132 (const_vec_duplicate_p): Check for a duplicated vector encoding.
17133 (unwrap_const_vec_duplicate): Likewise.
17134 (const_vec_series_p): Check for a non-duplicated vector encoding.
17135 Say that the function only returns true for integer vectors.
17136 * emit-rtl.c: Include rtx-vector-builder.h.
17137 (gen_const_vec_duplicate_1): Delete.
17138 (gen_const_vector): Call gen_const_vec_duplicate instead of
17139 gen_const_vec_duplicate_1.
17140 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
17141 (gen_const_vec_duplicate): Use rtx_vector_builder.
17142 (gen_const_vec_series): Likewise.
17143 (gen_rtx_CONST_VECTOR): Likewise.
17144 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
17145 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17146 Build a new vector rather than modifying a CONST_VECTOR in-place.
17147 (handle_special_swappables): Update call accordingly.
17148 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
17149 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17150 Build a new vector rather than modifying a CONST_VECTOR in-place.
17151 (handle_special_swappables): Update call accordingly.
17153 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17155 * simplify-rtx.c (simplify_const_binary_operation): Use
17156 CONST_VECTOR_ELT instead of XVECEXP.
17158 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17160 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
17161 the selector elements to be different from the data elements
17162 if the selector is a VECTOR_CST.
17163 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
17164 ssizetype for the selector.
17166 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17168 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
17169 before testing each element individually.
17170 * tree-vect-generic.c (lower_vec_perm): Likewise.
17172 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17174 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
17175 * selftest-run-tests.c (selftest::run_tests): Call it.
17176 * vector-builder.h (vector_builder::operator ==): New function.
17177 (vector_builder::operator !=): Likewise.
17178 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
17179 (vec_perm_indices::all_from_input_p): New function.
17180 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17181 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
17182 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
17183 instead of reading the VECTOR_CST directly. Detect whether both
17184 vector inputs are the same before constructing the vec_perm_indices,
17185 and update the number of inputs argument accordingly. Use the
17186 utility functions added above. Only construct sel2 if we need to.
17188 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17190 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
17191 the broadcast of the low byte.
17192 (expand_mult_highpart): Use an explicit encoding for the permutes.
17193 * optabs-query.c (can_mult_highpart_p): Likewise.
17194 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
17195 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17196 (vectorizable_bswap): Likewise.
17197 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
17198 explicit encoding for the power-of-2 permutes.
17199 (vect_permute_store_chain): Likewise.
17200 (vect_grouped_load_supported): Likewise.
17201 (vect_permute_load_chain): Likewise.
17203 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17205 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
17206 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
17207 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
17208 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17209 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
17210 (vect_gen_perm_mask_any): Likewise.
17212 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17214 * int-vector-builder.h: New file.
17215 * vec-perm-indices.h: Include int-vector-builder.h.
17216 (vec_perm_indices): Redefine as an int_vector_builder.
17217 (auto_vec_perm_indices): Delete.
17218 (vec_perm_builder): Redefine as a stand-alone class.
17219 (vec_perm_indices::vec_perm_indices): New function.
17220 (vec_perm_indices::clamp): Likewise.
17221 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
17222 (vec_perm_indices::new_vector): New function.
17223 (vec_perm_indices::new_expanded_vector): Update for new
17224 vec_perm_indices class.
17225 (vec_perm_indices::rotate_inputs): New function.
17226 (vec_perm_indices::all_in_range_p): Operate directly on the
17227 encoded form, without computing elided elements.
17228 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
17229 encoding. Update for new vec_perm_indices class.
17230 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
17231 the given vec_perm_builder.
17232 (expand_vec_perm_var): Update vec_perm_builder constructor.
17233 (expand_mult_highpart): Use vec_perm_builder instead of
17234 auto_vec_perm_indices.
17235 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
17236 vec_perm_indices instead of auto_vec_perm_indices. Use a single
17237 or double series encoding as appropriate.
17238 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
17239 vec_perm_indices instead of auto_vec_perm_indices.
17240 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17241 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17242 (vect_permute_store_chain): Likewise.
17243 (vect_grouped_load_supported): Likewise.
17244 (vect_permute_load_chain): Likewise.
17245 (vect_shift_permute_load_chain): Likewise.
17246 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17247 (vect_transform_slp_perm_load): Likewise.
17248 (vect_schedule_slp_instance): Likewise.
17249 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17250 (vectorizable_mask_load_store): Likewise.
17251 (vectorizable_bswap): Likewise.
17252 (vectorizable_store): Likewise.
17253 (vectorizable_load): Likewise.
17254 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
17255 vec_perm_indices instead of auto_vec_perm_indices. Use
17256 tree_to_vec_perm_builder to read the vector from a tree.
17257 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
17258 vec_perm_builder instead of a vec_perm_indices.
17259 (have_whole_vector_shift): Use vec_perm_builder and
17260 vec_perm_indices instead of auto_vec_perm_indices. Leave the
17261 truncation to calc_vec_perm_mask_for_shift.
17262 (vect_create_epilog_for_reduction): Likewise.
17263 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
17264 from auto_vec_perm_indices to vec_perm_indices.
17265 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17266 instead of changing individual elements.
17267 (aarch64_vectorize_vec_perm_const): Use new_vector to install
17268 the vector in d.perm.
17269 * config/arm/arm.c (expand_vec_perm_d::perm): Change
17270 from auto_vec_perm_indices to vec_perm_indices.
17271 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17272 instead of changing individual elements.
17273 (arm_vectorize_vec_perm_const): Use new_vector to install
17274 the vector in d.perm.
17275 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
17276 Update vec_perm_builder constructor.
17277 (rs6000_expand_interleave): Likewise.
17278 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
17279 (rs6000_expand_interleave): Likewise.
17281 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17283 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
17284 to qimode could truncate the indices.
17285 * optabs.c (expand_vec_perm_var): Likewise.
17287 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17289 * Makefile.in (OBJS): Add vec-perm-indices.o.
17290 * vec-perm-indices.h: New file.
17291 * vec-perm-indices.c: Likewise.
17292 * target.h (vec_perm_indices): Replace with a forward class
17294 (auto_vec_perm_indices): Move to vec-perm-indices.h.
17295 * optabs.h: Include vec-perm-indices.h.
17296 (expand_vec_perm): Delete.
17297 (selector_fits_mode_p, expand_vec_perm_var): Declare.
17298 (expand_vec_perm_const): Declare.
17299 * target.def (vec_perm_const_ok): Replace with...
17300 (vec_perm_const): ...this new hook.
17301 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
17302 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
17303 * doc/tm.texi: Regenerate.
17304 * optabs.def (vec_perm_const): Delete.
17305 * doc/md.texi (vec_perm_const): Likewise.
17306 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
17307 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
17308 expand_vec_perm for constant permutation vectors. Assert that
17309 the mode of variable permutation vectors is the integer equivalent
17310 of the mode that is being permuted.
17311 * optabs-query.h (selector_fits_mode_p): Declare.
17312 * optabs-query.c: Include vec-perm-indices.h.
17313 (selector_fits_mode_p): New function.
17314 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
17315 is defined, instead of checking whether the vec_perm_const_optab
17316 exists. Use targetm.vectorize.vec_perm_const instead of
17317 targetm.vectorize.vec_perm_const_ok. Check whether the indices
17318 fit in the vector mode before using a variable permute.
17319 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
17320 vec_perm_indices instead of an rtx.
17321 (expand_vec_perm): Replace with...
17322 (expand_vec_perm_const): ...this new function. Take the selector
17323 as a vec_perm_indices rather than an rtx. Also take the mode of
17324 the selector. Update call to shift_amt_for_vec_perm_mask.
17325 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
17326 Use vec_perm_indices::new_expanded_vector to expand the original
17327 selector into bytes. Check whether the indices fit in the vector
17328 mode before using a variable permute.
17329 (expand_vec_perm_var): Make global.
17330 (expand_mult_highpart): Use expand_vec_perm_const.
17331 * fold-const.c: Includes vec-perm-indices.h.
17332 * tree-ssa-forwprop.c: Likewise.
17333 * tree-vect-data-refs.c: Likewise.
17334 * tree-vect-generic.c: Likewise.
17335 * tree-vect-loop.c: Likewise.
17336 * tree-vect-slp.c: Likewise.
17337 * tree-vect-stmts.c: Likewise.
17338 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
17340 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
17341 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
17342 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
17343 (aarch64_vectorize_vec_perm_const): ...this new function.
17344 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17345 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17346 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
17347 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
17348 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17349 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17350 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
17352 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
17353 check for NEON modes.
17354 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
17355 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
17356 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
17357 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
17359 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
17360 the old VEC_PERM_CONST conditions.
17361 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
17362 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
17363 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
17364 (ia64_vectorize_vec_perm_const_ok): Merge into...
17365 (ia64_vectorize_vec_perm_const): ...this new function.
17366 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
17367 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
17368 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
17369 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
17370 * config/mips/mips.c (mips_expand_vec_perm_const)
17371 (mips_vectorize_vec_perm_const_ok): Merge into...
17372 (mips_vectorize_vec_perm_const): ...this new function.
17373 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
17374 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
17375 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
17376 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
17377 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
17378 (rs6000_expand_vec_perm_const): Delete.
17379 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
17381 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17382 (altivec_expand_vec_perm_const_le): Take each operand individually.
17383 Operate on constant selectors rather than rtxes.
17384 (altivec_expand_vec_perm_const): Likewise. Update call to
17385 altivec_expand_vec_perm_const_le.
17386 (rs6000_expand_vec_perm_const): Delete.
17387 (rs6000_vectorize_vec_perm_const_ok): Delete.
17388 (rs6000_vectorize_vec_perm_const): New function.
17389 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17390 an element count and rtx array.
17391 (rs6000_expand_extract_even): Update call accordingly.
17392 (rs6000_expand_interleave): Likewise.
17393 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
17394 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
17395 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
17396 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
17397 (rs6000_expand_vec_perm_const): Delete.
17398 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17399 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17400 (altivec_expand_vec_perm_const_le): Take each operand individually.
17401 Operate on constant selectors rather than rtxes.
17402 (altivec_expand_vec_perm_const): Likewise. Update call to
17403 altivec_expand_vec_perm_const_le.
17404 (rs6000_expand_vec_perm_const): Delete.
17405 (rs6000_vectorize_vec_perm_const_ok): Delete.
17406 (rs6000_vectorize_vec_perm_const): New function. Remove stray
17407 reference to the SPE evmerge intructions.
17408 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17409 an element count and rtx array.
17410 (rs6000_expand_extract_even): Update call accordingly.
17411 (rs6000_expand_interleave): Likewise.
17412 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
17413 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
17415 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17417 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17419 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
17420 vector mode and that that mode matches the mode of the data
17422 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
17423 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
17424 directly using expand_vec_perm_1 when forcing selectors into
17426 (expand_vec_perm_var): New function, split out from expand_vec_perm.
17428 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17430 * optabs-query.h (can_vec_perm_p): Delete.
17431 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
17432 * optabs-query.c (can_vec_perm_p): Split into...
17433 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
17434 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
17435 particular selector is valid.
17436 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17437 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17438 (vect_grouped_load_supported): Likewise.
17439 (vect_shift_permute_load_chain): Likewise.
17440 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17441 (vect_transform_slp_perm_load): Likewise.
17442 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17443 (vectorizable_bswap): Likewise.
17444 (vect_gen_perm_mask_checked): Likewise.
17445 * fold-const.c (fold_ternary_loc): Likewise. Don't take
17446 implementations of variable permutation vectors into account
17447 when deciding which selector to use.
17448 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
17449 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
17450 with a false third argument.
17451 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
17452 to test whether the constant selector is valid and can_vec_perm_var_p
17453 to test whether a variable selector is valid.
17455 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17457 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
17458 * optabs-query.c (can_vec_perm_p): Likewise.
17459 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
17460 instead of vec_perm_indices.
17461 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
17462 (vect_gen_perm_mask_checked): Likewise,
17463 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
17464 (vect_gen_perm_mask_checked): Likewise,
17466 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17468 * optabs-query.h (qimode_for_vec_perm): Declare.
17469 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
17470 (qimode_for_vec_perm): ...this new function.
17471 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
17473 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17475 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
17476 does not have a conditional at the top.
17478 2018-01-02 Richard Biener <rguenther@suse.de>
17480 * ipa-inline.c (big_speedup_p): Fix expression.
17482 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17485 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
17488 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17492 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
17493 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
17494 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
17495 cond_taken_branch_cost 3->4.
17497 2018-01-01 Jakub Jelinek <jakub@redhat.com>
17499 PR tree-optimization/83581
17500 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
17501 TODO_cleanup_cfg if any changes have been made.
17503 PR middle-end/83608
17504 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
17505 convert_modes if target mode has the right side, but different mode
17508 PR middle-end/83609
17509 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
17510 last argument when extracting from CONCAT. If either from_real or
17511 from_imag is NULL, use expansion through memory. If result is not
17512 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
17513 the parts directly to inner mode, if even that fails, use expansion
17516 PR middle-end/83623
17517 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
17518 check for bswap in mode rather than HImode and use that in expand_unop
17521 Copyright (C) 2018 Free Software Foundation, Inc.
17523 Copying and distribution of this file, with or without modification,
17524 are permitted in any medium without royalty provided the copyright
17525 notice and this notice are preserved.