1 2018-06-21 Martin Liska <mliska@suse.cz>
3 * ipa-pure-const.c (propagate_nothrow): Use
4 funct_state_summaries->get.
5 (dump_malloc_lattice): Likewise.
6 (propagate_malloc): Likewise.
8 2018-06-21 Richard Biener <rguenther@suse.de>
10 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
11 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
12 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
14 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
15 BLOCK_ABSTRACT_ORIGIN unconditionally.
17 2018-06-21 David Malcolm <dmalcolm@redhat.com>
19 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
21 * ipa-reference.c (ipa_reference_c_finalize): Delete
22 ipa_ref_opt_sum_summaries and set it to NULL.
24 2018-06-21 Tom de Vries <tdevries@suse.de>
26 PR tree-optimization/85859
27 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
28 test with comment from bb_no_side_effects_p.
30 2018-06-21 Richard Biener <rguenther@suse.de>
32 PR tree-optimization/86232
33 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
34 max for constant niter.
36 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
38 * config/aarch64/aarch64-simd.md
39 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
41 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
43 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
44 Make opernads of the unspec commutative.
46 2018-06-21 Richard Biener <rguenther@suse.de>
48 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
49 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
50 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
51 (vect_analyze_data_ref_dependence): Re-order checks to deal with
53 (vect_record_base_alignments): Do not record base alignment
54 for gathers or scatters.
55 (vect_compute_data_ref_alignment): Drop return value that is always
56 true. Bail out early for gathers or scatters.
57 (vect_enhance_data_refs_alignment): Bail out early for gathers
59 (vect_find_same_alignment_drs): Likewise.
60 (vect_analyze_data_refs_alignment): Remove dead code.
61 (vect_slp_analyze_and_verify_node_alignment): Likewise.
62 (vect_analyze_data_refs): For possible gathers or scatters do
63 not create an alternate DR, just check their possible validity
64 and mark them. Adjust DECL_NONALIASED handling to not rely
66 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
67 update inits of gathers or scatters.
68 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
69 Also copy gather/scatter flag to pattern vinfo.
71 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
73 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
74 behavior of vec_packsu (vector unsigned long long, vector unsigned
75 long long) to match behavior of vec_packs with same signature.
77 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
78 Thomas Schwinge <thomas@codesourcery.com>
79 Cesar Philippidis <cesar@codesourcery.com>
81 * gimplify.c (gimplify_scan_omp_clauses): Add support for
82 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
83 (gimplify_adjust_omp_clauses): Likewise.
84 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
85 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
86 (gimplify_omp_target_update): Update handling of acc update and
88 * omp-low.c (install_var_field): Remove unused parameter
89 base_pointers_restrict.
90 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
91 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
93 (omp_target_base_pointers_restrict_p): Delete.
94 (scan_omp_target): Update call to scan_sharing_clauses.
95 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
97 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
98 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
99 (convert_local_omp_clauses): Likewise.
100 * tree-pretty-print.c (dump_omp_clause): Likewise.
101 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
103 (omp_clause_code_name): Likewise.
105 2018-06-20 Jakub Jelinek <jakub@redhat.com>
108 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
111 PR tree-optimization/86231
112 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
113 anti-range don't overwrite *vr0min before using it to compute *vr0max.
115 2018-06-20 Tom de Vries <tdevries@suse.de>
117 PR tree-optimization/86097
118 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
119 iv type if signedness of iv type is not the same as that of *nit.
121 2018-06-20 Jakub Jelinek <jakub@redhat.com>
123 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
124 EDGE_EH edges, verify they are all EDGE_EH.
126 2018-06-20 Maya Rashish <coypu@sdf.org>
128 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
130 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
132 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
133 * config/aarch64/aarch64.c (xgene1_tunings): Add
134 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
135 (aarch64_mode_valid_for_sched_fusion_p):
137 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
138 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
140 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
142 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
143 * config/aarch64/iterators.md (VQ2): New mode iterator.
145 2018-06-20 Martin Liska <mliska@suse.cz>
147 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
148 Change default ratio from 10 to 8.
150 2018-06-20 Martin Liska <mliska@suse.cz>
152 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
154 (bit_test_cluster::find_bit_tests): Likewise.
155 (switch_decision_tree::analyze_switch_statement): Find clusters.
156 * tree-switch-conversion.h (struct jump_table_cluster): Document
159 2018-06-20 Martin Liska <mliska@suse.cz>
161 * tree-switch-conversion.c (switch_conversion::collect):
162 Record m_uniq property.
163 (switch_conversion::expand): Bail out for special conditions.
164 (group_cluster::~group_cluster): New.
165 (group_cluster::group_cluster): Likewise.
166 (group_cluster::dump): Likewise.
167 (jump_table_cluster::emit): New.
168 (switch_decision_tree::fix_phi_operands_for_edges): New.
169 (struct case_node): Remove struct.
170 (jump_table_cluster::can_be_handled): New.
171 (case_values_threshold): Moved to header.
172 (reset_out_edges_aux): Likewise.
173 (jump_table_cluster::is_beneficial): New.
174 (bit_test_cluster::can_be_handled): Likewise.
175 (add_case_node): Remove.
176 (bit_test_cluster::is_beneficial): New.
177 (case_bit_test::cmp): New.
178 (bit_test_cluster::emit): New.
179 (expand_switch_as_decision_tree_p): Remove.
180 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
181 (fix_phi_operands_for_edge): Likewise.
182 (switch_decision_tree::analyze_switch_statement): New.
183 (compute_cases_per_edge): Move ...
184 (switch_decision_tree::compute_cases_per_edge): ... here.
185 (try_switch_expansion): Likewise.
186 (switch_decision_tree::try_switch_expansion): Likewise.
187 (record_phi_operand_mapping): Likewise.
188 (switch_decision_tree::record_phi_operand_mapping): Likewise.
189 (emit_case_decision_tree): Likewise.
190 (switch_decision_tree::emit): Likewise.
191 (balance_case_nodes): Likewise.
192 (switch_decision_tree::balance_case_nodes): Likewise.
193 (dump_case_nodes): Likewise.
194 (switch_decision_tree::dump_case_nodes): Likewise.
195 (emit_jump): Likewise.
196 (switch_decision_tree::emit_jump): Likewise.
197 (emit_cmp_and_jump_insns): Likewise.
198 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
199 (emit_case_nodes): Likewise.
200 (switch_decision_tree::emit_case_nodes): Likewise.
201 (conditional_probability): Remove.
202 * tree-switch-conversion.h (enum cluster_type): New.
204 (struct cluster): Likewise.
205 (cluster::cluster): Likewise.
206 (struct simple_cluster): Likewise.
207 (simple_cluster::simple_cluster): Likewise.
208 (struct group_cluster): Likewise.
209 (struct jump_table_cluster): Likewise.
210 (struct bit_test_cluster): Likewise.
211 (struct min_cluster_item): Likewise.
212 (struct case_tree_node): Likewise.
213 (case_tree_node::case_tree_node): Likewise.
214 (jump_table_cluster::case_values_threshold): Likewise.
215 (struct case_bit_test): Likewise.
216 (struct switch_decision_tree): Likewise.
217 (struct switch_conversion): Likewise.
218 (switch_decision_tree::reset_out_edges_aux): Likewise.
220 2018-06-20 Martin Liska <mliska@suse.cz>
222 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
223 (hoist_edge_and_branch_if_true): Likewise.
224 (expand_switch_using_bit_tests_p): Likewise.
225 (struct case_bit_test): Likewise.
226 (case_bit_test_cmp): Likewise.
227 (emit_case_bit_tests): Likewise.
228 (switch_conversion::switch_conversion): New class.
229 (struct switch_conv_info): Remove old struct.
230 (collect_switch_conv_info): More to ...
231 (switch_conversion::collect): ... this.
232 (check_range): Likewise.
233 (switch_conversion::check_range): Likewise.
234 (check_all_empty_except_final): Likewise.
235 (switch_conversion::check_all_empty_except_final): Likewise.
236 (check_final_bb): Likewise.
237 (switch_conversion::check_final_bb): Likewise.
238 (create_temp_arrays): Likewise.
239 (switch_conversion::create_temp_arrays): Likewise.
240 (free_temp_arrays): Likewise.
241 (gather_default_values): Likewise.
242 (switch_conversion::gather_default_values): Likewise.
243 (build_constructors): Likewise.
244 (switch_conversion::build_constructors): Likewise.
245 (constructor_contains_same_values_p): Likewise.
246 (switch_conversion::contains_same_values_p): Likewise.
247 (array_value_type): Likewise.
248 (switch_conversion::array_value_type): Likewise.
249 (build_one_array): Likewise.
250 (switch_conversion::build_one_array): Likewise.
251 (build_arrays): Likewise.
252 (switch_conversion::build_arrays): Likewise.
253 (gen_def_assigns): Likewise.
254 (switch_conversion::gen_def_assigns): Likewise.
255 (prune_bbs): Likewise.
256 (switch_conversion::prune_bbs): Likewise.
257 (fix_phi_nodes): Likewise.
258 (switch_conversion::fix_phi_nodes): Likewise.
259 (gen_inbound_check): Likewise.
260 (switch_conversion::gen_inbound_check): Likewise.
261 (process_switch): Use the newly created class.
262 (switch_conversion::expand): New.
263 (switch_conversion::~switch_conversion): New.
264 * tree-switch-conversion.h: New file.
266 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
268 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
269 tree-vect-patterns.c.
270 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
271 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
272 (vect_recog_sad_pattern): Likewise.
273 (vect_recog_widen_sum_pattern): Likewise.
274 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
275 (vect_recog_widen_shift_pattern): Remove the type_in argument.
276 (vect_recog_rotate_pattern): Likewise.
277 (vect_recog_mult_pattern): Likewise.
278 (vect_recog_vector_vector_shift_pattern): Likewise.
279 (vect_recog_divmod_pattern): Likewise.
280 (vect_recog_mixed_size_cond_pattern): Likewise.
281 (vect_recog_bool_pattern): Likewise.
282 (vect_recog_mask_conversion_pattern): Likewise.
283 (vect_try_gather_scatter_pattern): Likewise.
284 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
285 (vect_recog_over_widening_pattern): Likewise.
286 (vect_recog_gather_scatter_pattern): Likewise.
287 (vect_recog_func_ptr): Move from tree-vectorizer.h
288 (vect_vect_recog_func_ptrs): Move further down the file.
289 (vect_recog_func): Likewise. Remove the third argument.
290 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
291 (vect_pattern_recog_1): Expect the pattern function to do any
292 necessary target tests. Also expect it to provide a vector type.
293 Remove the type_in handling.
295 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
297 * tree-vect-patterns.c (vect_pattern_detected): New function.
298 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
299 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
300 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
301 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
302 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
303 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
304 (vect_recog_mask_conversion_pattern)
305 (vect_try_gather_scatter_pattern): Likewise.
307 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
309 * tree-vect-patterns.c (vect_get_internal_def): New function.
310 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
311 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
312 (search_type_for_mask_1): Use it.
314 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
316 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
317 redundant WIDEN_SUM_EXPR handling.
318 (vect_recog_sad_pattern): Likewise.
320 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
322 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
323 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
324 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
325 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
326 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
328 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
330 * tree-vect-stmts.c (vectorizable_call): Make sure that we
331 use the stmt_vec_info of the original bb statement for the
332 new zero assignment, even if the call is part of a pattern.
334 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
336 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
337 that the sequence is attached to the original statement rather
338 than the pattern statement.
339 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
340 PATTERN_DEF_SEQ from the original statement rather than
341 the main pattern statement.
342 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
343 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
344 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
346 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
348 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
349 definition statements before the early exit for statements that aren't
351 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
353 (vect_transform_loop): ...here. Process pattern definition
354 statements without first checking whether the main pattern
355 statement is live or relevant.
357 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
359 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
360 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
362 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
364 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
365 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
366 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
367 (expand_block_compare): Change select_block_compare_mode call.
368 (expand_strncmp_align_check): Use new functions, fix comment.
369 (emit_final_str_compare_gpr): New function.
370 (expand_strn_compare): Refactor and clean up code.
371 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
373 2018-06-19 Tony Reix <tony.reix@atos.com>
374 Damien Bergamini <damien.bergamini@atos.com>
375 David Edelsohn <dje.gcc@gmail.com>
377 * collect2.c (static_obj): New variable.
378 (static_libs): New variable.
379 (is_in_list): Uncomment declaration.
380 (main): Track AIX libraries linked statically.
381 (is_in_list): Uncomment definition.
382 (scan_prog_file): Don't add AIX shared libraries initializer
383 to constructor list if linking statically.
385 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
387 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
389 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
391 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
393 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
396 2018-06-19 Martin Liska <mliska@suse.cz>
398 * config/i386/i386.c (ix86_can_inline_p): Do not use
399 ipa_fn_summaries::get_create.
400 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
402 (devirtualization_time_bonus): Likewise.
403 (ipcp_propagate_stage): Likewise.
404 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
405 (edge_set_predicate): Likewise.
406 (evaluate_conditions_for_known_args): Likewise.
407 (evaluate_properties_for_edge): Likewise.
408 (ipa_call_summary::reset): Tranform to ...
409 (ipa_call_summary::~ipa_call_summary): ... this.
410 (ipa_fn_summary::reset): Transform to ...
411 (ipa_fn_summary::~ipa_fn_summary): ... this.
412 (ipa_fn_summary_t::remove): Rename to ...
413 (ipa_fn_summary_t::remove_callees): ... this.
414 (ipa_fn_summary_t::duplicate): Use placement new
415 instead of memory copy.
416 (ipa_call_summary_t::duplicate): Likewise.
417 (ipa_call_summary_t::remove): Remove.
418 (dump_ipa_call_summary): Change get_create to get.
419 (ipa_dump_fn_summary): Dump only when summary exists.
420 (analyze_function_body): Use symbol_summary::get instead
422 (compute_fn_summary): Likewise.
423 (estimate_edge_devirt_benefit): Likewise.
424 (estimate_edge_size_and_time): Likewise.
425 (inline_update_callee_summaries): Likewise.
426 (remap_edge_change_prob): Likewise.
427 (remap_edge_summaries): Likewise.
428 (ipa_merge_fn_summary_after_inlining): Likewise.
429 (write_ipa_call_summary): Likewise.
430 (ipa_fn_summary_write): Likewise.
431 (ipa_free_fn_summary): Likewise.
432 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
433 (struct ipa_call_summary): Likewise.
434 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
436 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
437 (estimate_size_after_inlining): Likewise.
438 (estimate_growth): Likewise.
439 (growth_likely_positive): Likewise.
440 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
441 (inline_call): Likewise.
442 * ipa-inline.c (caller_growth_limits): Likewise.
443 (can_inline_edge_p): Likewise.
444 (can_inline_edge_by_limits_p): Likewise.
445 (compute_uninlined_call_time): Likewise.
446 (compute_inlined_call_time): Likewise.
447 (want_inline_small_function_p): Likewise.
448 (edge_badness): Likewise.
449 (update_caller_keys): Likewise.
450 (update_callee_keys): Likewise.
451 (inline_small_functions): Likewise.
452 (inline_to_all_callers_1): Likewise.
453 (dump_overall_stats): Likewise.
454 (early_inline_small_functions): Likewise.
455 (early_inliner): Likewise.
456 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
457 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
458 * ipa-pure-const.c (malloc_candidate_p): Likewise.
459 * ipa-split.c (execute_split_functions): Likewise.
460 * symbol-summary.h: Likewise.
461 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
463 2018-06-19 Richard Biener <rguenther@suse.de>
465 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
466 (vectorize_loops): ... here. Fix dbgcnt handling.
467 (try_vectorize_loop): Wrap try_vectorize_loop_1.
469 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
472 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
473 ieee128 argument takes up only one (vector) register, not two (floating
476 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
478 * gimplify.c (gimplify_init_constructor): Really never clear for an
479 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
481 2018-06-19 Richard Biener <rguenther@suse.de>
483 PR tree-optimization/86179
484 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
485 after failed recognition.
487 2018-06-18 Martin Sebor <msebor@redhat.com>
490 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
491 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
492 Handle integer subtraction.
493 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
494 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
496 2018-06-18 David Malcolm <dmalcolm@redhat.com>
498 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
499 param from rtx to rtx_insn *.
500 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
502 (frv_ifcvt_modify_insn): Likwise.
503 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
504 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
505 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
506 as_a <rtx_insn *> cast to local "unprotected_region" once
507 it's been established that it's not NULL or pc_rtx.
508 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
509 param "sethi" from rtx to rtx_insn *.
510 (nds32_group_float_insns): Likewise for param "insn".
511 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
513 (vax_output_int_subtract): Likewise.
514 * config/vax/vax.c (vax_output_int_add): Likewise for param
516 (vax_output_int_subtract): Likewise.
517 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
518 (emit_pattern_after): Likewise for param "after".
519 (emit_insn_after): Likewise.
520 (emit_jump_insn_after): Likewise.
521 (emit_call_insn_after): Likewise.
522 (emit_debug_insn_after): Likewise.
523 (emit_pattern_before): Likewise for param "before".
524 (emit_insn_before): Likewise.
525 (emit_jump_insn_before): Likewise.
526 * final.c (get_insn_template): Likewise for param "insn", removing
528 * output.h (get_insn_template): Likewise for 2nd param.
529 * rtl.h (emit_insn_before): Likewise.
530 (emit_jump_insn_before): Likewise.
531 (emit_debug_insn_before_noloc): Likewise.
532 (emit_insn_after): Likewise.
533 (emit_jump_insn_after): Likewise.
534 (emit_call_insn_after): Likewise.
535 (emit_debug_insn_after): Likewise.
536 (set_insn_deleted): Likewise for param.
538 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
541 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
542 floating point modes, so that IFmode is numerically greater than
543 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
544 to declare the ordering. This prevents IFmode from being
545 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
546 machine. Include rs6000-modes.h to share the fractional values
547 between genmodes* and the rest of the compiler.
551 * config/rs6000/rs6000-modes.h: New file.
552 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
553 meaning of rs6000_long_double_size so that 126..128 selects an
554 appropriate 128-bit floating point type.
555 (rs6000_option_override_internal): Likewise.
556 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
557 (TARGET_LONG_DOUBLE_128): Change the meaning of
558 rs6000_long_double_size so that 126..128 selects an appropriate
559 128-bit floating point type.
560 (LONG_DOUBLE_TYPE_SIZE): Update comment.
561 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
562 source and destination to match the standard usage.
563 (truncifkf2): Likewise.
564 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
565 ISA 2.07 to use an explicit clobber, instead of passing in a
567 (copysign<mode>3_soft): Likewise.
569 2018-06-18 David Malcolm <dmalcolm@redhat.com>
571 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
572 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
573 (vect_slp_analyze_instance_dependence): Likewise.
574 (vect_enhance_data_refs_alignment): Likewise.
575 (vect_analyze_data_refs_alignment): Likewise.
576 (vect_slp_analyze_and_verify_instance_alignment
577 (vect_analyze_data_ref_accesses): Likewise.
578 (vect_prune_runtime_alias_test_list): Likewise.
579 (vect_analyze_data_refs): Likewise.
580 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
581 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
582 (vect_analyze_scalar_cycles_1): Likewise.
583 (vect_get_loop_niters): Likewise.
584 (vect_analyze_loop_form_1): Likewise.
585 (vect_update_vf_for_slp): Likewise.
586 (vect_analyze_loop_operations): Likewise.
587 (vect_analyze_loop): Likewise.
588 (vectorizable_induction): Likewise.
589 (vect_transform_loop): Likewise.
590 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
591 * tree-vect-slp.c (vect_analyze_slp): Likewise.
592 (vect_make_slp_decision): Likewise.
593 (vect_detect_hybrid_slp): Likewise.
594 (vect_slp_analyze_operations): Likewise.
595 (vect_slp_bb): Likewise.
596 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
597 (vectorizable_bswap): Likewise.
598 (vectorizable_call): Likewise.
599 (vectorizable_simd_clone_call): Likewise.
600 (vectorizable_conversion): Likewise.
601 (vectorizable_assignment): Likewise.
602 (vectorizable_shift): Likewise.
603 (vectorizable_operation): Likewise.
604 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
606 2018-06-18 Martin Sebor <msebor@redhat.com>
608 PR tree-optimization/81384
609 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
610 * builtins.c (expand_builtin_strnlen): New function.
611 (expand_builtin): Call it.
612 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
613 * builtins.def (BUILT_IN_STRNLEN): New.
614 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
615 Warn for bounds in excess of maximum object size.
616 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
617 single-value ranges. Handle strnlen.
618 (handle_builtin_strlen): Handle strnlen.
619 (strlen_check_and_optimize_stmt): Same.
620 * doc/extend.texi (Other Builtins): Document strnlen.
622 2018-06-18 Maya Rashish <coypu@sdf.org>
624 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
625 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
626 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
628 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
630 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
632 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
634 * tree.c (escaped_string::escape): Replace cast to char * by
635 const_cast<char *> (unescaped).
637 2018-06-18 Nick Clifton <nickc@redhat.com>
640 * tree.c (escaped_string): New class. Converts an unescaped
641 string into its escaped equivalent.
642 (warn_deprecated_use): Use the new class to convert the
643 deprecation message, if present.
644 (test_escaped_strings): New self test.
645 (test_c_tests): Add test_escaped_strings.
646 * doc/extend.texi (deprecated): Add a note that the
647 deprecation message is affected by the -fmessage-length
648 option, and that control characters will be escaped.
649 (#pragma GCC error): Document this pragma.
650 (#pragma GCC warning): Likewise.
651 * doc/invoke.texi (-fmessage-length): Document this option's
652 effect on the #warning and #error preprocessor directives and
653 the deprecated attribute.
655 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
657 * tree.c (decl_value_expr_lookup): Revert latest change.
658 (decl_value_expr_insert): Likewise.
660 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
662 * gimplify.c (nonlocal_vlas): Delete.
663 (nonlocal_vla_vars): Likewise.
664 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
666 (gimplify_body): Do not create and destroy nonlocal_vlas.
667 * tree-nested.c: Include diagnostic.h.
668 (use_pointer_in_frame): Tweak.
669 (lookup_field_for_decl): Add assertion and declare the transformation.
670 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
671 internal error when the reference is in a wrong context. Do not
672 create a debug decl by default.
673 (note_nonlocal_block_vlas): Delete.
674 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
675 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
676 create a debug decl by default.
677 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
678 call is in a wrong context.
679 (fixup_vla_decls): New function.
680 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
681 debug variables were created.
682 * tree.c (decl_value_expr_lookup): Add checking assertion.
683 (decl_value_expr_insert): Likewise.
685 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
688 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
689 * tree-scalar-evolution.c (interpret_expr): Likewise.
690 (expression_expensive_p): Likewise.
691 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
692 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
693 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
694 (ssa_defined_by_minus_one_stmt_p): New.
696 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
699 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
700 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
701 * dojump.c (do_jump): Likewise.
702 * expr.c (expand_expr_real_2): Check operand type's sign.
703 * fold-const.c (const_unop): Handle ABSU_EXPR.
704 (fold_abs_const): Likewise.
705 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
706 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
707 (strip_sign_op_1): Likesise.
708 * match.pd: Add new pattern to generate ABSU_EXPR.
709 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
710 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
711 * tree-eh.c (operation_could_trap_helper_p): Likewise.
712 * tree-inline.c (estimate_operator_cost): Likewise.
713 * tree-pretty-print.c (dump_generic_node): Likewise.
714 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
715 * tree.def (ABSU_EXPR): New.
717 2018-06-16 Jakub Jelinek <jakub@redhat.com>
720 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
721 documented as preserved for backward compatibility only.
722 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
724 PR rtl-optimization/86108
725 * bb-reorder.c (create_forwarder_block): Renamed to ...
726 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
727 jump from new landing pad to the second part.
728 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
731 2018-06-15 Jakub Jelinek <jakub@redhat.com>
734 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
735 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
736 Only call store_expr for halves if the mode is the same.
739 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
740 Fix up comment formatting.
742 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
744 * typed-splay-tree.h (typed_splay_tree::remove): New function.
745 (typed_splay_tree::closure,
746 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
747 (typed_splay_tree::typed_splay_tree,
748 typed_splay_tree::operator =): Declared private.
749 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
750 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
751 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
752 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
753 typed_splay_tree::splay_tree_splay,
754 typed_splay_tree::splay_tree_foreach_helper,
755 typed_splay_tree::splay_tree_insert,
756 typed_splay_tree::splay_tree_remove,
757 typed_splay_tree::splay_tree_lookup,
758 typed_splay_tree::splay_tree_predecessor,
759 typed_splay_tree::splay_tree_successor,
760 typed_splay_tree::splay_tree_min,
761 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
762 (typed_splay_tree::root, typed_splay_tree::comp,
763 typed_splay_tree::delete_key,
764 typed_splay_tree::delete_value): New data members.
765 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
766 typed_splay_tree::remove.
768 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
770 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
771 -mginv and -mno-ginv to the assembler.
772 * config/mips/mips.opt (-mcrc): New option.
774 * doc/invoke.text (-mcrc): Document.
777 2018-06-15 Nick Clifton <nickc@redhat.com>
780 * tree.c (escaped_string): New class. Converts an unescaped
781 string into its escaped equivalent.
782 (warn_deprecated_use): Use the new class to convert the
783 deprecation message, if present.
784 (test_escaped_strings): New self test.
785 (test_c_tests): Add test_escaped_strings.
786 * doc/extend.texi (deprecated): Add a note that the
787 deprecation message is affected by the -fmessage-length
788 option, and that control characters will be escaped.
789 (#pragma GCC error): Document this pragma.
790 (#pragma GCC warning): Likewise.
791 * doc/invoke.texi (-fmessage-length): Document this option's
792 effect on the #warning and #error preprocessor directives and
793 the deprecated attribute.
795 2018-06-15 Richard Biener <rguenther@suse.de>
797 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
798 here, also noting vector size used.
799 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
800 size used in MSG_OPTIMIZED_LOCATIONS dump.
801 (pass_slp_vectorize::execute): Adjust.
803 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
806 * config/arc/arc.c (arc_return_address_register): Fix
809 2018-06-15 Richard Biener <rguenther@suse.de>
812 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
813 leave useless conversion stripping to force_gimple_operand_gsi.
814 (gimplify_build2): Likewise.
815 (gimplify_build1): Likewise.
817 2018-06-15 Richard Biener <rguenther@suse.de>
820 * tree-cfg.c (move_stmt_op): unshare invariant addresses
821 before adjusting their block.
823 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
825 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
826 multilibs for *-*-rtems*.
827 * config/riscv/t-rtems: New file.
829 2018-06-14 Jakub Jelinek <jakub@redhat.com>
832 * match.pd ((A +- CST1) +- CST2): Punt if last resort
833 unsigned_type_for returns NULL.
836 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
837 subregs of multi-word pseudos unless the float mode has word size.
839 2018-06-14 Richard Biener <rguenther@suse.de>
842 * tree-vect-generic.c (build_word_mode_vector_type): Remove
843 duplicate and harmful type_hash_canon.
844 * tree.c (type_hash_canon): Assert we didn't find ourselves.
846 2018-06-14 Richard Biener <rguenther@suse.de>
849 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
852 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
854 * config/rtems.h (STDINT_LONG32): Define.
856 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
857 Prachi Godbole <prachi.godbole@imgtec.com>
859 * config/mips/mips-cpus.def: Define P6600.
860 * config/mips/mips-tables.opt: Regenerate.
861 * config/mips/mips.c (mips_ucbranch_type): New enum.
862 (mips_rtx_cost_data): Add support for P6600.
863 (mips_issue_rate): Likewise.
864 (mips_multipass_dfa_lookahead): Likewise.
865 (mips_avoid_hazard): Likewise.
866 (mips_reorg_process_insns): Likewise.
867 (mips_classify_branch_p6600): New function.
868 * config/mips/mips.h (TUNE_P6600): New define.
869 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
870 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
871 * config/mips/mips.md: Include p6600.md.
872 (processor): Add p6600.
873 * config/mips/p6600.md: New file.
874 * doc/invoke.texi: Add p6600 to supported architectures.
876 2018-06-13 Martin Sebor <msebor@redhat.com>
878 PR tree-optimization/86114
879 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
881 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
883 2018-06-13 Richard Biener <rguenther@suse.de>
885 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
886 Properly set vector type of the intermediate stmt.
887 * tree-vect-stmts.c (vectorizable_operation): The destination
888 var always has vectype_out type.
890 2018-06-13 Jeff Law <law@redhat.com>
892 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
893 integer 0 for argument to print_rtl_with_bb.
894 (rl78_reorg): Likewise.
896 2018-06-13 David Malcolm <dmalcolm@redhat.com>
898 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
899 from rtx to rtx_insn *.
900 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
902 (add_sched_insns_for_speculation): Likewise for local "target",
903 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
904 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
905 from rtx_insn ** to rtx_code_label **.
906 (reorg_emit_nops): Likewise.
907 (c6x_reorg): Likewise for local "call_labels".
908 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
910 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
911 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
912 the loops over LABEL_REFS.
913 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
915 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
916 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
917 (split_branches): Strengthen local "olabel" from rtx to
918 rtx_insn *, adding a safe_as_a cast.
919 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
921 (add_insn_after): Likewise for first two params.
922 (add_insn_before): Likewise.
923 (remove_insn): Likewise for param.
924 (emit_pattern_before_noloc): Likewise for second and third params.
925 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
926 (emit_call_insn_before_noloc): Likewise.
927 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
929 (emit_barrier_before): Likewise.
930 (emit_label_before): Strengthen "label" param from "rtx" to
931 "rtx_code_label *". Strengthen "before" param from "rtx" to
933 (emit_insn_after_1): Strengthen "after" param from "rtx" to
935 (emit_pattern_after_noloc): Likewise.
936 (emit_insn_after_noloc): Likewise.
937 (emit_jump_insn_after_noloc): Likewise.
938 (emit_call_insn_after_noloc): Likewise.
939 (emit_debug_insn_after_noloc): Likewise.
940 (emit_barrier_after): Likewise.
941 (emit_label_after): Likewise for both params.
942 (emit_pattern_after_setloc): Likewise for "after" param. Convert
943 "loc" param from "int" to "location_t".
944 (emit_insn_after_setloc): Likewise.
945 (emit_jump_insn_after_setloc): Likewise.
946 (emit_call_insn_after_setloc): Likewise.
947 (emit_debug_insn_after_setloc): Likewise.
948 (emit_pattern_before_setloc): Likewise for "before" param. Convert
949 "loc" param from "int" to "location_t".
950 (emit_pattern_before): Convert NULL_RTX to NULL.
951 (emit_insn_before_setloc): Convert "loc" param from "int" to
953 (emit_jump_insn_before_setloc): Likewise.
954 (emit_call_insn_before_setloc): Likewise.
955 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
956 rtx_insn *. Convert "loc" param from "int" to "location_t".
957 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
958 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
959 Convert 3rd param from "int" to "location_t".
960 (emit_barrier_before, emit_barrier_after, next_real_insn):
961 Strengthen param from rtx to rtx_insn *.
962 (emit_label_before): Strengthen 1st param from "rtx" to
963 "rtx_code_label *". Strengthen 2nd param from "rtx" to
965 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
966 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
967 Strengthen 2nd param from "rtx" to "rtx_insn *".
968 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
969 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
970 Likewise. Convert 3rd param from "int" to "location_t".
971 (emit_label_after): Strengthen 1st param from "rtx" to
973 (next_real_insn, remove_insn): Strengthen param from "rtx" to
975 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
976 from "rtx" to "rtx_insn *".
978 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
980 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
981 bodies streamed in with -Q.
982 * dumpfile.c (dump_files): Add lto-stream-out dump file.
983 * dumpfile.h (tree_dump_index): Add lto_stream_out.
984 * gimple-streamer-out.c: Include gimple-pretty-print.h
985 (output_bb): Dump stmts streamed.
986 * lto-section-out.c: Include print-tree.h
987 (lto_begin_section): Dump sections created.
988 (lto_output_decl_index): Dump decl encoded.
989 * lto-streamer-out.c: Include print-tree.h
990 (create_output_block): Dump output block created.
991 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
992 (output_function): Dump function output.
993 (output_constructor): Dump constructor streamed.
994 (write_global_stream): Output indexes encoded.
995 (produce_asm_for_decls): Dump streams encoded.
996 * lto-streamer.c (streamer_dump_file): New global var.
997 * lto-streamer.h (streamer_dump_file): Declare.
998 * passes.c (ipa_write_summaries): Initialize streamer dump.
999 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
1002 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
1005 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
1006 offsets for register save directives. Emit a second batch of save
1007 directives, if need be, when the function accesses prior frames.
1009 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1011 * config/arc/fpu.md (fmasf4): Force operand to register.
1012 (fnmasf4): Likewise.
1014 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1016 * config/arc/arc-protos.h (arc_pad_return): Remove.
1017 * config/arc/arc.c (machine_function): Remove force_short_suffix
1019 (arc_print_operand): Adjust printing of '&'.
1020 (arc_verify_short): Remove conditional printing of short suffix.
1021 (arc_final_prescan_insn): Remove reference to size_reason.
1022 (pad_return): New function.
1023 (arc_reorg): Call pad_return.
1024 (arc_pad_return): Remove.
1025 (arc_init_machine_status): Remove reference to force_short_suffix.
1026 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
1027 (attr length): When attribute iscompact is true force to 2
1028 regardless; in the case of maybe check if we want to force the
1029 instruction to have 4 bytes length.
1030 (nopv): Change it to generate 4 byte long nop as well.
1031 (blockage): New pattern.
1032 (simple_return): Remove call to arc_pad_return.
1033 (p_return_i): Likewise.
1035 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1037 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
1039 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1041 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
1044 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1046 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
1047 for ARC700 and ARCv2.
1049 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
1052 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
1053 operands[2] instead of operands[1].
1056 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1058 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
1059 case, check whether the outer register overlaps an unallocatable
1060 register, not just whether it fits the required class.
1062 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1064 * poly-int.h (can_div_trunc_p): Add new overload in which all values
1066 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
1067 (memrefs_conflict_p): Likewise.
1068 (init_alias_analysis): Likewise.
1069 * cfgexpand.c (expand_debug_expr): Likewise.
1070 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
1071 * cse.c (fold_rtx): Likewise.
1072 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
1073 * expr.c (emit_block_move_hints): Likewise.
1074 (clear_storage_hints, push_block, emit_push_insn): Likewise.
1075 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
1076 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
1077 (emit_group_store): Likewise.
1078 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
1079 to read the PRE/POST_MODIFY increment.
1080 * calls.c (store_one_arg): Use strip_offset.
1081 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
1083 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
1085 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
1086 (simplify_binary_operation_1): Extend CONST_INT handling to
1088 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
1089 than a HOST_WIDE_INT.
1090 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
1092 (adjust_mems, add_stores): Update accodingly.
1093 (vt_canonicalize_addr): Track polynomial offsets.
1094 (emit_note_insn_var_location): Likewise.
1095 (vt_add_function_parameter): Likewise.
1096 (vt_initialize): Likewise.
1098 2018-06-12 Jeff Law <law@redhat.com>
1100 * config.gcc (alpha*-*-freebsd*): Remove.
1101 * config/alpha/freebsd.h: Remove.
1103 2018-06-12 David Malcolm <dmalcolm@redhat.com>
1106 * spellcheck-tree.c (levenshtein_distance): Rename to...
1107 (get_edit_distance): ...this, and update for underlying renaming.
1108 * spellcheck-tree.h (levenshtein_distance): Rename to...
1109 (get_edit_distance): ...this.
1110 * spellcheck.c (levenshtein_distance): Rename to...
1111 (get_edit_distance): ...this. Convert from Levenshtein distance
1112 to Damerau-Levenshtein distance by supporting transpositions of
1113 adjacent characters. Rename "v1" to "v_next" and "v0" to
1115 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
1116 (selftest::test_edit_distance_unit_test_oneway): ...this, and
1117 update for underlying renaming.
1118 (selftest::levenshtein_distance_unit_test): Rename to...
1119 (selftest::test_get_edit_distance_unit): ...this, and update for
1120 underlying renaming.
1121 (selftest::test_find_closest_string): Add example from PR 69968
1122 where transposition helps
1123 (selftest::test_metric_conditions): Update for renaming.
1124 (selftest::test_metric_conditions): Likewise.
1125 (selftest::spellcheck_c_tests): Likewise.
1126 * spellcheck.h (levenshtein_distance): Rename both overloads to...
1127 (get_edit_distance): ...this.
1128 (best_match::consider): Update for renaming.
1130 2018-06-12 Martin Sebor <msebor@redhat.com>
1132 PR tree-optimization/85259
1133 * builtins.c (compute_objsize): Handle constant offsets.
1134 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
1135 true iff a warning has been issued.
1136 * gimple.h (gimple_nonartificial_location): New function.
1137 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
1138 gimple_nonartificial_location and handle -Wno-system-headers.
1139 (handle_builtin_stxncpy): Same.
1141 2018-06-12 Martin Sebor <msebor@redhat.com>
1144 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
1146 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
1148 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
1149 BUILTIN_VEC_XST entries for pointer to double and long long.
1151 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
1154 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1156 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1159 2018-06-12 Martin Liska <mliska@suse.cz>
1161 * doc/options.texi: Document IntegerRange.
1163 2018-06-12 Martin Liska <mliska@suse.cz>
1165 * config/i386/i386.opt: Make MPX-related options as Deprecated.
1166 * opt-functions.awk: Handle Deprecated flag.
1167 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
1169 (read_cmdline_option): Report warning for a deprecated option.
1170 * opts.h (struct cl_option): Add new field cl_deprecated.
1171 (CL_ERR_DEPRECATED): New.
1173 2018-06-12 Martin Liska <mliska@suse.cz>
1175 * doc/options.texi: Document Deprecated option flag.
1177 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1179 * config/arc/arc-arch.h (arc_extras): New enum.
1180 (arc_cpu_t):Add field extra.
1181 (arc_cpu_types): Consider the extras.
1182 * config/arc/arc-cpus.def: Add extras info.
1183 * config/arc/arc-opts.h (processor_type): Consider extra field.
1184 * config/arc/arc.c (arc_override_options): Handle extra field.
1186 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1188 * config/arc/arc-arch.h: Update ARC_OPTX macro.
1189 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
1191 * config/arc/arc.c (arc_init): Update pic warning.
1192 (irq_range): Update irq range parsing warnings.
1193 (arc_override_options): Update various warning messages.
1194 (arc_handle_aux_attribute): Likewise.
1196 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
1198 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
1200 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1202 * doc/sourcebuild.texi: Document usage of line number 0 in verify
1203 compiler messages directives.
1205 2018-06-12 Matthew Fortune <mfortune@gmail.com>
1207 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
1208 * config/mips/mips-tables.opt: Regenerate.
1209 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
1211 * doc/invoke.texi: Document -march=i6500.
1213 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
1215 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
1216 (i6400_gpmul): Add cpu_unit.
1217 (i6400_gpdiv): Likewise.
1218 (i6400_msa_add_d): Update reservations.
1219 (i6400_msa_int_add) Likewise.
1220 (i6400_msa_short_logic3) Likewise.
1221 (i6400_msa_short_logic2) Likewise.
1222 (i6400_msa_short_logic) Likewise.
1223 (i6400_msa_move) Likewise.
1224 (i6400_msa_cmp) Likewise.
1225 (i6400_msa_short_float2) Likewise.
1226 (i6400_msa_div_d) Likewise.
1227 (i6400_msa_long_logic1) Likewise.
1228 (i6400_msa_long_logic2) Likewise.
1229 (i6400_msa_mult) Likewise.
1230 (i6400_msa_long_float2) Likewise.
1231 (i6400_msa_long_float4) Likewise.
1232 (i6400_msa_long_float5) Likewise.
1233 (i6400_msa_long_float8) Likewise.
1234 (i6400_fpu_fadd): Include frint type.
1235 (i6400_fpu_store): New define_insn_reservation.
1236 (i6400_fpu_load): Likewise.
1237 (i6400_fpu_move): Likewise.
1238 (i6400_fpu_fcmp): Likewise.
1239 (i6400_fpu_fmadd): Likewise.
1240 (i6400_int_mult): Include imul3nc type and update reservation.
1241 (i6400_int_div): Include idiv3 type and update reservation.
1242 (i6400_int_load): Update to check type not move_type.
1243 (i6400_int_store): Likewise.
1244 (i6400_int_prefetch): Set zero latency.
1246 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
1248 * gcc.c: Document new %@{...} sequence.
1249 (LINK_COMMAND_SPEC): Use it for the -L switches.
1250 (cpp_unique_options): Use it for the -I switches.
1251 (at_file_argbuf): New global variable.
1252 (in_at_file): Likewise.
1253 (alloc_args): Create at_file_argbuf.
1254 (clear_args): Truncate at_file_argbuf.
1255 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
1256 (open_at_file): New function.
1257 (close_at_file): Likewise.
1258 (create_at_file): Delete.
1259 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
1262 (validate_switches_from_spec): Deal with %@{...} sequence.
1263 (validate_switches): Likewise.
1264 (driver::finalize): Call clear_args.
1266 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
1268 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
1270 2018-06-11 Martin Sebor <msebor@redhat.com>
1272 * doc/invoke.texi (-Wall): List -Wc++17-compat.
1273 (Wno-class-memaccess): Add @opindex.
1274 (Wno-templates, Wno-multiple-inheritance): Same.
1275 (Wno-virtual-inheritance, Wno-namespaces): Same.
1276 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
1277 (Wno-format-overflow, Wno-format-truncation): Same.
1278 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
1279 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
1280 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
1281 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
1282 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
1283 (Wno-misspelled-isr): Same.
1285 2018-06-11 Martin Sebor <msebor@redhat.com>
1287 * PR tree-optimization/86083
1288 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
1290 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
1292 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
1294 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
1297 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
1298 on the correct operand.
1299 (*movdi_internal64): Ditto.
1301 2018-06-11 Martin Liska <mliska@suse.cz>
1303 PR tree-optimization/86089
1304 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
1306 2018-06-11 Julia Koval <julia.koval@intel.com>
1308 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
1309 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
1310 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
1312 2018-06-11 Olivier Hainque <hainque@adacore.com>
1314 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
1315 for Ada with strict dwarf2.
1317 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
1320 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
1323 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
1325 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
1327 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
1329 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
1332 2018-06-08 Martin Liska <mliska@suse.cz>
1334 * tree-cfg.h (debug_function): Fix argument type to match
1337 2018-06-08 Martin Liska <mliska@suse.cz>
1339 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
1340 Remove usage of MPX-related (and removed) fields.
1341 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
1343 2018-06-08 David Malcolm <dmalcolm@redhat.com>
1345 * cfg.c (debug): Use TDF_NONE rather than 0.
1346 * cfghooks.c (debug): Likewise.
1347 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
1348 (struct dump_option_value_info): Convert to...
1349 (struct kv_pair): ...this template type.
1350 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
1352 (optinfo_verbosity_options): Likewise.
1353 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
1355 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
1356 than int for "optgroup_flags" param.
1357 (dump_generic_expr_loc): Use dump_flags_t rather than int for
1359 (dump_dec): Likewise.
1360 (dump_finish): Use TDF_NONE rather than 0.
1361 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
1362 rather than int for "optgroup_flags" param. Use TDF_NONE rather
1363 than 0. Update for change to option_ptr.
1364 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
1365 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
1366 0. Update for changes to optinfo_verbosity_options and
1368 (opt_info_switch_p): Convert optgroup_flags from int to
1370 (dump_basic_block): Use dump_flags_t rather than int
1371 for "dump_kind" param.
1372 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
1373 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
1374 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
1375 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
1376 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
1377 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
1378 TDF_NONE): Convert from macros to...
1379 (enum dump_flag): ...this new enum.
1380 (dump_flags_t): Update to use enum.
1381 (operator|, operator&, operator~, operator|=, operator&=):
1382 Implement for dump_flags_t.
1383 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
1384 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
1385 Convert from macros to...
1386 (enum optgroup_flag): ...this new enum.
1387 (optgroup_flags_t): New typedef.
1388 (operator|, operator|=): Implement for optgroup_flags_t.
1389 (struct dump_file_info): Convert field "alt_flags" to
1390 dump_flags_t. Convert field "optgroup_flags" to
1392 (dump_basic_block): Use dump_flags_t rather than int for param.
1393 (dump_generic_expr_loc): Likewise.
1394 (dump_dec): Likewise.
1395 (dump_register): Convert param "optgroup_flags" to
1397 (opt_info_enable_passes): Likewise.
1398 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
1400 * gimple-pretty-print.c (debug): Likewise.
1401 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
1402 (merged_store_group::apply_stores): Likewise.
1403 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
1404 * gimple.c (verify_gimple_pp): Likewise.
1405 * graphite-poly.c (print_pbb_body): Likewise.
1406 * passes.c (pass_manager::register_one_dump_file): Convert
1407 local "optgroup_flags" to optgroup_flags_t.
1408 * print-tree.c (print_node): Use TDF_NONE rather than 0.
1410 (debug_body): Likewise.
1411 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
1412 to optgroup_flags_t.
1413 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
1415 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
1416 (convert_mult_to_fma): Likewise.
1417 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
1418 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
1419 * tree-vect-data-refs.c (dump_lower_bound): Convert param
1420 "dump_kind" to dump_flags_t.
1422 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1424 * config/rs6000/rs6000.c (min, max): Delete.
1426 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1428 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
1429 -mabi=spe and -mabi=no-spe.
1431 2018-06-08 Martin Liska <mliska@suse.cz>
1433 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
1434 where we expect an existing summary.
1436 2018-06-08 Martin Liska <mliska@suse.cz>
1438 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
1439 * ipa-inline.h (estimate_edge_growth): Likewise.
1441 2018-06-08 Martin Liska <mliska@suse.cz>
1443 * cgraph.c (function_version_hasher::hash): Use
1444 cgraph_node::get_uid ().
1445 (function_version_hasher::equal):
1446 * cgraph.h (cgraph_node::get_uid): New method.
1447 * ipa-inline.c (update_caller_keys): Use
1448 cgraph_node::get_uid ().
1449 (update_callee_keys): Likewise.
1450 * ipa-utils.c (searchc): Likewise.
1451 (ipa_reduced_postorder): Likewise.
1452 * lto-cgraph.c (input_node): Likewise.
1453 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
1454 * symbol-summary.h (symtab_insertion): Likewise.
1455 (symtab_removal): Likewise.
1456 (symtab_duplication): Likewise.
1457 * tree-pretty-print.c (dump_function_header): Likewise.
1458 * tree-sra.c (convert_callers_for_node): Likewise.
1460 2018-06-08 Martin Liska <mliska@suse.cz>
1462 * cgraph.c (symbol_table::create_edge): Always assign a new
1464 (symbol_table::free_edge): Do not recycle numbers.
1465 * cgraph.h (cgraph_edge::get): New method.
1466 * symbol-summary.h (symtab_removal): Use it.
1467 (symtab_duplication): Likewise.
1468 (call_summary::hashable_uid): Remove.
1470 2018-06-08 Martin Liska <mliska@suse.cz>
1472 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
1473 (initialize_growth_caches): Remove.
1474 (free_growth_caches): Likewise.
1475 (do_estimate_edge_time): Use edge_growth_cache.
1476 (do_estimate_edge_size): Likewise.
1477 (do_estimate_edge_hints): Likewise.
1478 * ipa-inline.c (reset_edge_caches): Likewise.
1479 (recursive_inlining): Likewise.
1480 (inline_small_functions): Likewise.
1481 * ipa-inline.h (initialize_growth_caches): Remove.
1482 (estimate_edge_size): Likewise.
1483 (estimate_edge_time): Likewise.
1484 (estimate_edge_hints): Likewise.
1485 (reset_edge_growth_cache): Likewise.
1486 * symbol-summary.h (call_summary::remove): New method.
1488 2018-06-08 Martin Liska <mliska@suse.cz>
1490 * ipa-cp.c (class edge_clone_summary): New summary.
1491 (grow_edge_clone_vectors): Remove.
1492 (ipcp_edge_duplication_hook): Remove.
1493 (class edge_clone_summary_t): New call_summary class.
1494 (ipcp_edge_removal_hook): Remove.
1495 (edge_clone_summary_t::duplicate): New function.
1496 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
1497 (create_specialized_node): Likewise.
1498 (ipcp_driver): Initialize edge_clone_summaries and do not
1501 2018-06-08 Martin Liska <mliska@suse.cz>
1503 * symbol-summary.h (get): New function.
1504 (call_summary::m_initialize_when_cloning): New class member.
1506 2018-06-08 Martin Liska <mliska@suse.cz>
1508 * cgraph.c (cgraph_node::remove): Do not recycle uid.
1509 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
1510 (symbol_table::allocate_cgraph_symbol): Do not set uid.
1511 * passes.c (uid_hash_t): Record removed_nodes by their uids.
1512 (remove_cgraph_node_from_order): Use the removed_nodes set.
1513 (do_per_function_toporder): Likwise.
1514 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
1515 instead of summary_uid.
1516 (symtab_removal): Likewise.
1517 (symtab_duplication): Likewise.
1519 2018-06-08 Martin Liska <mliska@suse.cz>
1521 * ipa-cp.c (ipcp_store_bits_results): Use
1522 ipcp_transformation_sum.
1523 (ipcp_store_vr_results): Likewise.
1524 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
1526 (ipcp_transformation_initialize): ... this.
1527 (ipa_set_node_agg_value_chain):
1528 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
1529 (write_ipcp_transformation_info): Likewise.
1530 (read_ipcp_transformation_info): Likewise.
1531 (ipcp_update_bits): Likewise.
1532 (ipcp_update_vr): Likewise.
1533 (ipcp_transform_function): Likewise.
1534 * ipa-prop.h: Rename ipcp_transformation_summary to
1535 ipcp_transformation.
1536 (class ipcp_transformation_t): New function summary.
1537 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
1538 (ipa_get_agg_replacements_for_node): Likewise.
1540 2018-06-08 Martin Liska <mliska@suse.cz>
1542 * ipa-pure-const.c (struct funct_state_d): Do it class instead
1544 (class funct_state_summary_t): New function_summary class.
1545 (has_function_state): Remove.
1546 (get_function_state): Likewise.
1547 (set_function_state): Likewise.
1548 (add_new_function): Likewise.
1549 (funct_state_summary_t::insert): New function.
1550 (duplicate_node_data): Remove.
1551 (remove_node_data): Remove.
1552 (funct_state_summary_t::duplicate): New function.
1553 (register_hooks): Create new funct_state_summaries.
1554 (pure_const_generate_summary): Use it.
1555 (pure_const_write_summary): Likewise.
1556 (pure_const_read_summary): Likewise.
1557 (propagate_pure_const): Likewise.
1558 (propagate_nothrow): Likewise.
1559 (dump_malloc_lattice): Likewise.
1560 (propagate_malloc): Likewise.
1561 (execute): Do not register hooks, just remove summary
1563 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
1566 2018-06-08 Martin Liska <mliska@suse.cz>
1568 * ipa-reference.c (remove_node_data): Remove.
1569 (duplicate_node_data): Likewise.
1570 (class ipa_ref_var_info_summary_t): New class.
1571 (class ipa_ref_opt_summary_t): Likewise.
1572 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
1573 (get_reference_optimization_summary): Use
1574 ipa_ref_opt_sum_summaries.
1575 (set_reference_vars_info): Remove.
1576 (set_reference_optimization_summary): Likewise.
1577 (ipa_init): Create summaries.
1578 (init_function_info): Use function summary.
1579 (ipa_ref_opt_summary_t::duplicate): New function.
1580 (ipa_ref_opt_summary_t::remove): New function.
1581 (get_read_write_all_from_node): Fix GNU coding style.
1582 (propagate): Use function summary.
1583 (write_node_summary_p): Fix GNU coding style.
1584 (stream_out_bitmap): Likewise.
1585 (ipa_reference_read_optimization_summary): Use function summary.
1586 (ipa_reference_c_finalize): Do not release hooks.
1588 2018-06-08 Martin Liska <mliska@suse.cz>
1590 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
1591 (analyze_function_body): Extract multiple calls of get_create.
1592 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1593 * ipa-inline.c (recursive_inlining): Use ::get method.
1594 * ipa-inline.h (estimate_edge_growth): Likewise.
1596 2018-06-08 Martin Liska <mliska@suse.cz>
1598 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
1600 (hsa_function_summary::hsa_function_summary): Use the new enum
1602 (hsa_gpu_implementation_p): Use hsa_summaries::get.
1603 * hsa-gen.c (hsa_get_host_function): Likewise.
1604 (get_brig_function_name): Likewise.
1605 * ipa-hsa.c (process_hsa_functions): Likewise.
1606 (ipa_hsa_write_summary): Likewise.
1607 * symbol-summary.h (symtab_duplication): Use ::get function/
1608 (get): New function.
1610 2018-06-08 Martin Liska <mliska@suse.cz>
1612 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
1614 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
1615 (hsa_register_kernel): Likewise.
1616 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
1617 * hsa-gen.c (hsa_get_host_function): Likewise.
1618 (get_brig_function_name): Likewise.
1619 (generate_hsa): Likewise.
1620 (pass_gen_hsail::execute): Likewise.
1621 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
1622 (devirtualization_time_bonus): Likewise.
1623 (ipcp_propagate_stage): Likewise.
1624 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1625 (edge_set_predicate): Likewise.
1626 (evaluate_conditions_for_known_args): Likewise.
1627 (evaluate_properties_for_edge): Likewise.
1628 (ipa_fn_summary::reset): Likewise.
1629 (ipa_fn_summary_t::duplicate): Likewise.
1630 (dump_ipa_call_summary): Likewise.
1631 (ipa_dump_fn_summary): Likewise.
1632 (analyze_function_body): Likewise.
1633 (compute_fn_summary): Likewise.
1634 (estimate_edge_devirt_benefit): Likewise.
1635 (estimate_edge_size_and_time): Likewise.
1636 (estimate_calls_size_and_time): Likewise.
1637 (estimate_node_size_and_time): Likewise.
1638 (inline_update_callee_summaries): Likewise.
1639 (remap_edge_change_prob): Likewise.
1640 (remap_edge_summaries): Likewise.
1641 (ipa_merge_fn_summary_after_inlining): Likewise.
1642 (ipa_update_overall_fn_summary): Likewise.
1643 (read_ipa_call_summary): Likewise.
1644 (inline_read_section): Likewise.
1645 (write_ipa_call_summary): Likewise.
1646 (ipa_fn_summary_write): Likewise.
1647 (ipa_free_fn_summary): Likewise.
1648 * ipa-hsa.c (process_hsa_functions): Likewise.
1649 (ipa_hsa_write_summary): Likewise.
1650 (ipa_hsa_read_section): Likewise.
1651 * ipa-icf.c (sem_function::merge): Likewise.
1652 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1653 (do_estimate_edge_time): Likewise.
1654 (estimate_size_after_inlining): Likewise.
1655 (estimate_growth): Likewise.
1656 (growth_likely_positive): Likewise.
1657 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1658 (inline_call): Likewise.
1659 * ipa-inline.c (caller_growth_limits): Likewise.
1660 (can_inline_edge_p): Likewise.
1661 (can_inline_edge_by_limits_p): Likewise.
1662 (compute_uninlined_call_time): Likewise.
1663 (compute_inlined_call_time): Likewise.
1664 (want_inline_small_function_p): Likewise.
1665 (edge_badness): Likewise.
1666 (update_caller_keys): Likewise.
1667 (update_callee_keys): Likewise.
1668 (recursive_inlining): Likewise.
1669 (inline_small_functions): Likewise.
1670 (inline_to_all_callers_1): Likewise.
1671 (dump_overall_stats): Likewise.
1672 (early_inline_small_functions): Likewise.
1673 (early_inliner): Likewise.
1674 * ipa-inline.h (estimate_edge_growth): Likewise.
1675 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1676 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1677 * ipa-prop.h (IPA_NODE_REF): Likewise.
1678 (IPA_EDGE_REF): Likewise.
1679 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1680 (propagate_malloc): Likewise.
1681 * ipa-split.c (execute_split_functions): Likewise.
1682 * symbol-summary.h: Rename get to get_create.
1684 (get_create): Likewise.
1685 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1687 2018-06-08 Martin Liska <mliska@suse.cz>
1689 * symbol-summary.h (release): Move definition out of class
1691 (symtab_removal): Likewise.
1692 (symtab_duplication): Likewise.
1694 2018-06-08 Martin Liska <mliska@suse.cz>
1696 * symbol-summary.h (function_summary): Move constructor
1697 implementation out of class declaration.
1698 (release): Likewise.
1699 (symtab_insertion): Likewise.
1700 (symtab_removal): Likewise.
1701 (symtab_duplication): Likewise.
1704 2018-06-08 Martin Liska <mliska@suse.cz>
1706 * Makefile.in: Remove support for MPX (macros, related functions,
1707 fields in cgraph_node, ...).
1708 * builtin-types.def (BT_BND): Likewise.
1709 (BT_FN_BND_CONST_PTR): Likewise.
1710 (BT_FN_CONST_PTR_BND): Likewise.
1711 (BT_FN_VOID_PTR_BND): Likewise.
1712 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
1713 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
1714 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
1715 (expand_builtin_mempcpy_with_bounds): Likewise.
1716 (expand_builtin_memset_with_bounds): Likewise.
1717 (expand_builtin_memset_args): Likewise.
1718 (std_expand_builtin_va_start): Likewise.
1719 (expand_builtin): Likewise.
1720 (expand_builtin_with_bounds): Likewise.
1721 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
1722 (DEF_LIB_BUILTIN_CHKP): Likewise.
1723 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
1724 (DEF_CHKP_BUILTIN): Likewise.
1725 (BUILT_IN_MEMCPY): Likewise.
1726 (BUILT_IN_MEMMOVE): Likewise.
1727 (BUILT_IN_MEMPCPY): Likewise.
1728 (BUILT_IN_MEMSET): Likewise.
1729 (BUILT_IN_STPCPY): Likewise.
1730 (BUILT_IN_STRCAT): Likewise.
1731 (BUILT_IN_STRCHR): Likewise.
1732 (BUILT_IN_STRCPY): Likewise.
1733 (BUILT_IN_STRLEN): Likewise.
1734 (BUILT_IN_MEMCPY_CHK): Likewise.
1735 (BUILT_IN_MEMMOVE_CHK): Likewise.
1736 (BUILT_IN_MEMPCPY_CHK): Likewise.
1737 (BUILT_IN_MEMSET_CHK): Likewise.
1738 (BUILT_IN_STPCPY_CHK): Likewise.
1739 (BUILT_IN_STRCAT_CHK): Likewise.
1740 (BUILT_IN_STRCPY_CHK): Likewise.
1741 * calls.c (store_bounds): Likewise.
1742 (emit_call_1): Likewise.
1743 (special_function_p): Likewise.
1744 (maybe_warn_nonstring_arg): Likewise.
1745 (initialize_argument_information): Likewise.
1746 (finalize_must_preallocate): Likewise.
1747 (compute_argument_addresses): Likewise.
1748 (expand_call): Likewise.
1749 * cfgexpand.c (expand_call_stmt): Likewise.
1750 (expand_return): Likewise.
1751 (expand_gimple_stmt_1): Likewise.
1752 (pass_expand::execute): Likewise.
1753 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
1754 (cgraph_node::remove): Likewise.
1755 (cgraph_node::dump): Likewise.
1756 (cgraph_node::verify_node): Likewise.
1757 * cgraph.h (chkp_function_instrumented_p): Likewise.
1758 (symtab_node::get_alias_target): Likewise.
1759 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
1760 (cgraph_local_p): Likewise.
1761 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
1762 (cgraph_edge::rebuild_references): Likewise.
1763 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
1764 (walk_polymorphic_call_targets): Likewise.
1765 (cgraph_node::expand_thunk): Likewise.
1766 (symbol_table::output_weakrefs): Likewise.
1767 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
1768 (ix86_handle_option): Likewise.
1769 * config/i386/constraints.md: Likewise.
1770 * config/i386/i386-builtin-types.def (BND): Likewise.
1774 * config/i386/i386-builtin.def (BDESC_END): Likewise.
1775 (BDESC_FIRST): Likewise.
1777 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
1778 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
1779 * config/i386/i386.c (enum reg_class): Likewise.
1780 (ix86_target_string): Likewise.
1781 (ix86_option_override_internal): Likewise.
1782 (ix86_conditional_register_usage): Likewise.
1783 (ix86_valid_target_attribute_inner_p): Likewise.
1784 (ix86_set_indirect_branch_type): Likewise.
1785 (ix86_set_current_function): Likewise.
1786 (ix86_function_arg_regno_p): Likewise.
1787 (init_cumulative_args): Likewise.
1788 (ix86_function_arg_advance): Likewise.
1789 (ix86_function_arg): Likewise.
1790 (ix86_pass_by_reference): Likewise.
1791 (ix86_function_value_regno_p): Likewise.
1792 (ix86_function_value_1): Likewise.
1793 (ix86_function_value_bounds): Likewise.
1794 (ix86_return_in_memory): Likewise.
1795 (ix86_setup_incoming_vararg_bounds): Likewise.
1796 (ix86_va_start): Likewise.
1797 (indirect_thunk_need_prefix): Likewise.
1798 (print_reg): Likewise.
1799 (ix86_print_operand): Likewise.
1800 (ix86_expand_call): Likewise.
1801 (ix86_output_function_return): Likewise.
1802 (reg_encoded_number): Likewise.
1803 (BDESC_VERIFYS): Likewise.
1804 (ix86_init_mpx_builtins): Likewise.
1805 (ix86_init_builtins): Likewise.
1806 (ix86_emit_cmove): Likewise.
1807 (ix86_emit_move_max): Likewise.
1808 (ix86_expand_builtin): Likewise.
1809 (ix86_builtin_mpx_function): Likewise.
1810 (ix86_get_arg_address_for_bt): Likewise.
1811 (ix86_load_bounds): Likewise.
1812 (ix86_store_bounds): Likewise.
1813 (ix86_load_returned_bounds): Likewise.
1814 (ix86_store_returned_bounds): Likewise.
1815 (ix86_class_likely_spilled_p): Likewise.
1816 (ix86_hard_regno_mode_ok): Likewise.
1817 (x86_order_regs_for_local_alloc): Likewise.
1818 (ix86_mitigate_rop): Likewise.
1819 (ix86_bnd_prefixed_insn_p): Likewise.
1820 (ix86_mpx_bound_mode): Likewise.
1821 (ix86_make_bounds_constant): Likewise.
1822 (ix86_initialize_bounds): Likewise.
1823 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
1824 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
1825 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
1826 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
1827 (TARGET_CHKP_BOUND_MODE): Likewise.
1828 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
1829 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
1830 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
1831 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
1832 * config/i386/i386.h (TARGET_MPX): Likewise.
1833 (TARGET_MPX_P): Likewise.
1834 (VALID_BND_REG_MODE): Likewise.
1835 (FIRST_BND_REG): Likewise.
1836 (LAST_BND_REG): Likewise.
1837 (enum reg_class): Likewise.
1838 (BND_REG_P): Likewise.
1839 (BND_REGNO_P): Likewise.
1840 (BNDmode): Likewise.
1841 (ADJUST_INSN_LENGTH): Likewise.
1842 * config/i386/i386.md: Likewise.
1843 * config/i386/i386.opt: Likewise.
1844 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
1845 (defined): Likewise.
1846 (LINK_MPX): Likewise.
1847 (MPX_SPEC): Likewise.
1848 (LIBMPX_SPEC): Likewise.
1849 (LIBMPXWRAPPERS_SPEC): Likewise.
1850 (CHKP_SPEC): Likewise.
1851 * config/i386/predicates.md: Likewise.
1852 * dbxout.c (dbxout_type): Likewise.
1853 * doc/extend.texi: Likewise.
1854 * doc/invoke.texi: Likewise.
1855 * doc/md.texi: Likewise.
1856 * doc/tm.texi: Likewise.
1857 * doc/tm.texi.in: Likewise.
1858 * dwarf2out.c (is_base_type): Likewise.
1859 (gen_formal_types_die): Likewise.
1860 (gen_subprogram_die): Likewise.
1861 (gen_type_die_with_usage): Likewise.
1862 (gen_decl_die): Likewise.
1863 (dwarf2out_late_global_decl): Likewise.
1864 * expr.c (expand_assignment): Likewise.
1865 (emit_storent_insn): Likewise.
1866 (store_expr_with_bounds): Likewise.
1867 (store_expr): Likewise.
1868 (expand_expr_real_1): Likewise.
1869 * expr.h (store_expr_with_bounds): Likewise.
1870 * function.c (use_register_for_decl): Likewise.
1871 (struct bounds_parm_data): Likewise.
1872 (assign_parms_augmented_arg_list): Likewise.
1873 (assign_parm_find_entry_rtl): Likewise.
1874 (assign_parm_is_stack_parm): Likewise.
1875 (assign_parm_load_bounds): Likewise.
1876 (assign_bounds): Likewise.
1877 (assign_parms): Likewise.
1878 (expand_function_start): Likewise.
1879 * gcc.c (CHKP_SPEC): Likewise.
1880 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1881 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
1882 (wrestrict_dom_walker::check_call): Likewise.
1883 * gimple.c (gimple_build_call_from_tree): Likewise.
1884 * gimple.h (enum gf_mask): Likewise.
1885 (gimple_call_with_bounds_p): Likewise.
1886 (gimple_call_set_with_bounds): Likewise.
1887 * gimplify.c (gimplify_init_constructor): Likewise.
1888 * ipa-cp.c (initialize_node_lattices): Likewise.
1889 (propagate_constants_across_call): Likewise.
1890 (find_more_scalar_values_for_callers_subset): Likewise.
1891 * ipa-hsa.c (process_hsa_functions): Likewise.
1892 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
1893 * ipa-icf.c (sem_function::merge): Likewise.
1894 * ipa-inline.c (early_inliner): Likewise.
1895 * ipa-pure-const.c (warn_function_noreturn): Likewise.
1896 (warn_function_cold): Likewise.
1897 (propagate_pure_const): Likewise.
1898 * ipa-ref.h (enum GTY): Likewise.
1899 * ipa-split.c (find_retbnd): Likewise.
1900 (consider_split): Likewise.
1901 (split_function): Likewise.
1902 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1903 * ipa.c (walk_polymorphic_call_targets): Likewise.
1904 (symbol_table::remove_unreachable_nodes): Likewise.
1905 (process_references): Likewise.
1906 (cgraph_build_static_cdtor_1): Likewise.
1907 * lto-cgraph.c (lto_output_node): Likewise.
1908 (output_refs): Likewise.
1909 (compute_ltrans_boundary): Likewise.
1910 (input_overwrite_node): Likewise.
1911 (input_node): Likewise.
1912 (input_cgraph_1): Likewise.
1913 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
1914 * passes.c (pass_manager::execute_early_local_passes): Likewise.
1915 (class pass_chkp_instrumentation_passes): Likewise.
1916 (make_pass_chkp_instrumentation_passes): Likewise.
1917 * passes.def: Likewise.
1918 * rtl.h (struct GTY): Likewise.
1919 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
1920 * stor-layout.c (layout_type): Likewise.
1921 * symtab.c: Likewise.
1922 * target.def: Likewise.
1923 * targhooks.c (default_chkp_bound_type): Likewise.
1924 (default_chkp_bound_mode): Likewise.
1925 (default_builtin_chkp_function): Likewise.
1926 (default_chkp_function_value_bounds): Likewise.
1927 (default_chkp_make_bounds_constant): Likewise.
1928 (default_chkp_initialize_bounds): Likewise.
1929 * targhooks.h (default_chkp_bound_type): Likewise.
1930 (default_chkp_bound_mode): Likewise.
1931 (default_builtin_chkp_function): Likewise.
1932 (default_chkp_function_value_bounds): Likewise.
1933 (default_chkp_make_bounds_constant): Likewise.
1934 (default_chkp_initialize_bounds): Likewise.
1935 * toplev.c (compile_file): Likewise.
1936 (process_options): Likewise.
1937 * tree-core.h (DEF_BUILTIN): Likewise.
1938 (DEF_BUILTIN_CHKP): Likewise.
1939 * tree-inline.c (declare_return_variable): Likewise.
1940 (remap_gimple_stmt): Likewise.
1941 (copy_bb): Likewise.
1942 (initialize_inlined_parameters): Likewise.
1943 (expand_call_inline): Likewise.
1944 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
1945 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
1946 (make_pass_ipa_chkp_produce_thunks): Likewise.
1947 (make_pass_chkp): Likewise.
1948 (make_pass_chkp_opt): Likewise.
1949 (make_pass_chkp_instrumentation_passes): Likewise.
1950 * tree-pretty-print.c (dump_generic_node): Likewise.
1951 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
1952 * tree-ssa-dce.c (propagate_necessity): Likewise.
1953 (eliminate_unnecessary_stmts): Likewise.
1954 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
1955 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1956 * tree-ssa-sccvn.h: Likewise.
1957 * tree-ssa-strlen.c (get_string_length): Likewise.
1958 (valid_builtin_call): Likewise.
1959 (adjust_last_stmt): Likewise.
1960 (handle_builtin_strchr): Likewise.
1961 (handle_builtin_strcpy): Likewise.
1962 (handle_builtin_stxncpy): Likewise.
1963 (handle_builtin_memcpy): Likewise.
1964 (handle_builtin_strcat): Likewise.
1965 (strlen_check_and_optimize_stmt): Likewise.
1966 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
1967 * tree-streamer-in.c: Likewise.
1968 * tree-streamer.c (record_common_node): Likewise.
1969 * tree.c (tree_code_size): Likewise.
1970 (wide_int_to_tree_1): Likewise.
1971 (type_contains_placeholder_1): Likewise.
1972 (build_common_tree_nodes): Likewise.
1973 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
1974 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
1975 (POINTER_BOUNDS_P): Likewise.
1976 (BOUNDED_TYPE_P): Likewise.
1977 (BOUNDED_P): Likewise.
1978 (CALL_WITH_BOUNDS_P): Likewise.
1979 (pointer_bounds_type_node): Likewise.
1980 * value-prof.c (gimple_ic): Likewise.
1981 * var-tracking.c (vt_add_function_parameters): Likewise.
1982 * varasm.c (make_decl_rtl): Likewise.
1983 (assemble_start_function): Likewise.
1984 (output_constant): Likewise.
1985 (maybe_assemble_visibility): Likewise.
1986 * varpool.c (ctor_for_folding): Likewise.
1987 * chkp-builtins.def: Remove.
1988 * ipa-chkp.c: Remove.
1989 * ipa-chkp.h: Remove.
1990 * rtl-chkp.c: Remove.
1991 * rtl-chkp.h: Remove.
1992 * tree-chkp-opt.c: Remove.
1993 * tree-chkp.c: Remove.
1994 * tree-chkp.h: Remove.
1996 2018-06-07 Carl Love <cel@us.ibm.com>
1998 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
1999 vextract_fp_from_shortl): Add BE support.
2001 2018-06-07 Paul Koning <ni1d@arrl.net>
2003 * compare-elim.c (try_merge_compare): Don't merge compare if
2004 address contains a side effect.
2005 (try_eliminate_compare): Likewise.
2007 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
2009 * config.gcc: Support "tremont".
2010 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
2011 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2013 * config/i386/i386.c (m_TREMONT): Define.
2014 (processor_target_table): Add "tremont".
2015 (PTA_TREMONT): Define.
2016 (ix86_lea_outperforms): Add TARGET_TREMONT.
2017 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
2018 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
2019 and M_INTEL_GOLDMONT_PLUS.
2020 (fold_builtin_cpu): Add "tremont".
2021 (ix86_add_stmt_cost): Add TARGET_TREMONT.
2022 (ix86_option_override_internal): Add "tremont".
2023 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
2024 (processor_type): Add PROCESSOR_TREMONT.
2025 * config/i386/x86-tune.def: Add m_TREMONT.
2026 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
2028 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2030 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
2031 symbol defined for msp430i* devices to be lower case.
2033 gcc/testsuite/gcc.target/msp430/
2034 * msp430i-device-symbol.c: New test.
2035 * msp430f-device-symbol.c: New test.
2036 * msp430.h: New test header file.
2038 2018-06-07 Richard Biener <rguenther@suse.de>
2040 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
2041 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
2042 Properly wrap signed arithmetic if overflow wraps.
2044 2018-06-07 Jakub Jelinek <jakub@redhat.com>
2046 PR tree-optimization/69615
2047 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
2048 of a cast from a same precision integral SSA_NAME in a bb dominated
2049 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
2050 cast to utype if rhs2 has already a compatible type.
2052 2018-06-07 Richard Biener <rguenther@suse.de>
2054 PR tree-optimization/85935
2055 * graphite-scop-detection.c (find_params_in_bb): Analyze
2056 condition operands with respect to the correct loop. Assert
2057 the analysis doesn't fail.
2059 2018-06-04 Carl Love <cel@us.ibm.com>
2061 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
2062 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
2063 as it is slightly cheaper.
2064 (first_match_or_eos_index_<mode>):
2065 Calculate index using natural element order.
2066 (first_match_index_<mode>):
2067 Calculate index using natural element order.
2068 (first_match_or_eos_index_<mode>):
2069 Calculate index using natural order.
2070 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
2071 for BE and LE modes.
2072 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
2073 P9V_BUILTIN_VCLZLSBB_V16QI.
2074 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
2077 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2079 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
2080 indentation and line wrap for many prototypes. Add missing
2081 @smallexample directives around block of prototypes for vec_xl and
2084 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
2086 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
2087 track if we pass or return IEEE 128-bit floating point.
2088 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
2089 C++ mangling that is compatible with GCC 8.1.
2090 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
2091 (init_cumulative_args): Note if we pass or return IEEE 128-bit
2092 floating point types.
2093 (rs6000_function_arg_advance_1): Likewise.
2094 (rs6000_mangle_type): Optionally generate mangled names that match
2095 what GCC 8.1 generated for IEEE 128-bit floating point types.
2096 (rs6000_globalize_decl_name): If we have an external function that
2097 passes or returns IEEE 128-bit types, generate a weak reference
2098 from the mangled name used in GCC 8.1 to the current mangled
2100 (rs6000_init_builtins): Make __ibm128 use the long double type if
2101 long double is IBM extended double. Make __float128 use the long
2102 double type if long double is IEEE 128-bit.
2105 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
2106 macro for __ibm128 built-in functions.
2107 (PACK_IF): Add __ibm128 pack/unpack functions.
2108 (UNPACK_IF): Likewise.
2109 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
2110 enable long double built-in functions if long double is IEEE
2111 128-bit floating point.
2112 (rs6000_invalid_builtin): Update long double built-in function
2114 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
2115 functions, adjust the built-in function to use the long double
2116 built-in function if __ibm128 and long double are the same type.
2117 * doc/extend.texi (PowerPC builtins): Update documention for
2118 __builtin_{,un}pack_longdouble. Add documentation for
2119 __builtin_{,un}pack_ibm128.
2121 2018-06-06 Jim Wilson <jimw@sifive.com>
2123 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
2124 (struct machine_function): New field interrupt_mode.
2125 (riscv_handle_type_attribute): New function. Add forward declaration.
2126 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
2127 (riscv_expand_epilogue): Check interrupt_mode field.
2128 (riscv_set_current_function): Check interrupt attribute args and
2129 set interrupt_mode field.
2130 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
2131 (riscv_sret, riscv_uret): New.
2132 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
2133 new arguments to interrupt attribute.
2135 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
2138 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
2139 Don't handle -mcpu=power8 if -mpower9-vector is also used.
2141 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2143 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
2144 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
2145 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2146 several redundant entries.
2148 2018-06-06 David Malcolm <dmalcolm@redhat.com>
2150 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
2151 type from "rtx" to "rtx_insn *".
2152 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
2153 for local "call_insn", removing cast.
2154 (ix86_expand_call): Likewise, introducing a "call_insn" local.
2156 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
2158 PR tree-optimization/86066
2159 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
2160 for BIT_INSERT_EXPR stores.
2162 2018-06-06 Richard Biener <rguenther@suse.de>
2164 PR tree-optimization/86062
2165 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2166 component refs ontop
2167 of to be offsetted base.
2169 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2171 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
2172 to be static and remove check on interrupt attribute name.
2174 gcc/testsuite/gcc.target/msp430/
2175 * function-attributes-4.c: New test.
2176 * static-interrupts.c: New test.
2178 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
2180 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2181 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
2183 2018-06-05 Steve Ellcey <sellcey@cavium.com>
2186 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
2188 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
2189 Remove second argument, change how error is called.
2190 (aarch64_layout_arg): Remove second argument from
2191 aarch64_err_no_fpadvsimd call.
2192 (aarch64_init_cumulative_args): Ditto.
2193 (aarch64_gimplify_va_arg_expr): Ditto.
2194 * config/aarch64/aarch64.md (mov<mode>): Ditto.
2196 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
2198 * config/i386/i386.md (simple_return_indirect_internal): New expander.
2199 (*simple_return_indirect_internal<mode>): Rename from
2200 simple_return_indirect_internal. Use W mode iterator.
2201 (rstorssp): New expander.
2202 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
2203 (clrssbsy): New expander.
2204 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
2206 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
2208 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
2210 (cmse_check_pointed_object): Likewise.
2212 2018-06-05 Martin Liska <mliska@suse.cz>
2214 PR gcov-profile/47618
2215 * doc/invoke.texi: Document how -fprofile-dir format
2218 2018-06-05 Richard Biener <rguenther@suse.de>
2220 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
2221 removal pretend DOM info isn't available so we do not update
2222 it and only remove edges, not dominated blocks. Actually free
2223 DOM info in case we removed something. Remove unreachable blocks.
2224 (mfb_keep_latches): Work with either DOM info or marked backedges.
2225 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
2226 first. Mark backedges if DOM info isn't available.
2227 (Re-)compute DOM info after cleanup_control_flow_pre.
2229 2018-06-05 Richard Biener <rguenther@suse.de>
2231 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
2232 (locus_discrim_hasher::hash): Adjust.
2233 (locus_discrim_hasher::equal): Likewise.
2234 (next_discriminator_for_locus): Work on line directly.
2235 (same_line_p): Pass in expanded locus1 as well.
2236 (assign_discriminators): Avoid redundant location expansions.
2238 2018-06-05 Richard Biener <rguenther@suse.de>
2240 PR tree-optimization/86046
2241 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
2242 if required after clearing TREE_ADDRESSABLE.
2244 2018-06-05 Richard Biener <rguenther@suse.de>
2246 PR tree-optimization/86047
2247 * tree-ssa-loop.c (for_each_index): Glob handling of all
2248 decls and constants and really handle all of them.
2250 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2253 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
2254 qualifier_void_pointer and qualifier_const_void_pointer.
2255 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
2256 (arm_init_builtins): Handle the above.
2257 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
2258 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
2259 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
2262 2018-06-05 Martin Liska <mliska@suse.cz>
2264 * auto-profile.c (read_autofdo_file): Do not use
2265 gcov_ctr_summary struct.
2266 (afdo_callsite_hot_enough_for_early_inline): Likewise.
2267 * coverage.c (struct counts_entry): Likewise.
2268 (read_counts_file): Read just single summary entry.
2269 (get_coverage_counts): Use gcov_summary struct.
2270 * coverage.h (get_coverage_counts): Likewise.
2271 * gcov-dump.c (dump_working_sets): Likewise.
2272 (tag_summary): Dump just single summary.
2273 * gcov-io.c (gcov_write_summary): Write just histogram
2275 (gcov_read_summary): Read just single summary.
2276 (compute_working_sets): Use gcov_summary struct.
2277 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
2278 of GCOV_COUNTERS_SUMMABLE.
2279 (GCOV_COUNTERS_SUMMABLE): Remove.
2280 (GCOV_FIRST_VALUE_COUNTER): Replace with
2281 GCOV_COUNTER_V_INTERVAL.
2282 (struct gcov_ctr_summary): Remove.
2283 (struct gcov_summary): Directly use fields of former
2285 (compute_working_sets): Use gcov_summary struct.
2286 * gcov.c (read_count_file): Do not use ctrs fields.
2287 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
2289 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
2291 * profile.c: Likewise.
2292 * profile.h: Likewise.
2294 2018-06-05 Martin Liska <mliska@suse.cz>
2296 PR gcov-profile/84846
2297 * gcov.c (output_lines): Print working directory only
2298 in intermediate format.
2300 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
2302 * config/s390/s390-builtin-types.def: Add void function type.
2303 * config/s390/s390-builtins.def: Use the function type for the
2306 2018-06-04 Jim Wilson <jimw@sifive.com>
2308 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
2310 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
2311 and maybe_eh_return. Change regno to unsigned int. Use new args to
2312 handle EH_RETURN_DATA_REGNO registers properly.
2313 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
2314 (riscv_expand_epilogue): Update comment. Change argument name and
2315 type. Update code to use new name and type. Pass new args to
2316 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
2318 * config/riscv/riscv.md (NORMAL_RETURN): New.
2319 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
2320 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
2321 (eh_return): Call gen_eh_return_internal and emit barrier.
2322 (eh_return_internal): Call riscv_expand_epilogue.
2324 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
2326 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
2327 bit_insertion field and declare can_be_merged_into method.
2328 (merged_store_group::can_be_merged_into): New method.
2329 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
2330 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
2331 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
2333 2018-06-04 Richard Biener <rguenther@suse.de>
2335 PR tree-optimization/85955
2336 * builtins.c (fold_builtin_sincos): Convert pointers to
2337 destination to appropriate type before dereferencing.
2339 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
2341 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
2343 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
2345 * expr.c (expand_expr_real_1): Force the operand into memory if
2346 its TYPE_MODE is BLKmode and if there is no integer mode for
2347 the number of bits being extracted.
2349 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2353 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2354 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
2356 2018-06-04 Richard Biener <rguenther@suse.de>
2358 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
2359 (cleanup_tree_cfg_noloop): ... single caller. Do
2360 start_recording_case_labels later.
2362 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
2364 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
2365 to _IMMINTRIN_H_INCLUDED.
2366 * config/i386/pconfigintrin.h: Ditto.
2367 * config/i386/waitpkgintrin.h: Ditto.
2368 * config/i386/immintrin.h: Add includes for sgxintrin.h,
2369 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
2370 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
2371 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
2372 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
2373 waitpkgintrin.h and cldemoteintrin.h.
2375 2018-06-04 Richard Biener <rguenther@suse.de>
2377 PR tree-optimization/86038
2378 * tracer.c (find_best_successor): Check probability for
2379 being initialized, bail out if not.
2381 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
2384 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
2385 of bits to ignore when comparing architectures.
2387 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2389 PR tree-optimization/69615
2390 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
2391 maximum or minimum of the type, try to merge it also as if
2392 range1 is + [-, x - 1] or + [x + 1, -].
2395 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
2397 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
2399 PR tree-optimization/86034
2400 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
2401 the unsigned bitfield type in a bit insertion sequence if it does not
2402 have a larger precision than the bitfield size.
2403 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
2405 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
2407 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
2409 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2411 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
2412 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
2413 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
2414 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
2416 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2418 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
2419 Disable -fdelete-null-pointer-checks for ELF toolchain.
2421 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2422 Kito Cheng <kito.cheng@gmail.com>
2424 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
2425 (nds32le-*-*, nds32be-*-*): Integrate checking process.
2426 (nds32*-*-*): Add glibc and uclibc conditions.
2427 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
2428 (TARGET_EXCEPT_UNWIND_INFO): Define.
2429 * config/nds32/elf.h: New file.
2430 * config/nds32/linux.h: New file.
2431 * config/nds32/nds32-elf.opt: New file.
2432 * config/nds32/nds32-linux.opt: New file.
2433 * config/nds32/nds32-fp-as-gp.c
2434 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
2435 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
2437 (nds32_asm_file_end): Ditto.
2438 (nds32_print_operand): Ditto.
2439 (nds32_insert_attributes): Ditto.
2440 (nds32_init_libfuncs): New function.
2441 (TARGET_HAVE_TLS): Define.
2442 (TARGET_INIT_LIBFUNCS): Define.
2443 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
2445 (TARGET_ELF): Apply different mcmodel setting.
2446 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
2447 been migrated into elf.h and linux.h files.
2448 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
2449 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
2450 (mcmodel): The content has been migrated into nds32-elf.opt and
2451 nds32-linux.opt files.
2452 * config/nds32/t-elf: New file.
2453 * config/nds32/t-linux: New file.
2455 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2456 Shiva Chen <shiva0217@gmail.com>
2458 * config/nds32/constants.md (unspec_volatile_element): Add
2459 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
2460 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
2462 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
2463 (make_pass_nds32_fp_as_gp): Declare.
2464 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
2466 (nds32_asm_function_end_prologue): Remove unused asm output.
2467 (nds32_asm_function_begin_epilogue): Remove unused asm output.
2468 (nds32_asm_file_start): Output necessary fp_as_gp information.
2469 (nds32_option_override): Adjust register usage.
2470 (nds32_expand_prologue): Consider fp_as_gp situation.
2471 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
2472 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
2475 (simple_return): Ditto.
2476 (omit_fp_begin): Output special directive for fp_as_gp.
2477 (omit_fp_end): Output special directive for fp_as_gp.
2478 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
2479 mforbid-fp-as-gp): New options.
2481 2018-06-01 Mark Wielaard <mark@klomp.org>
2483 * dwarf2out.c (dwarf2out_finish): Remove generation of
2484 DW_AT_loclists_base.
2486 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
2488 * gimple-ssa-store-merging.c: Include gimple-fold.h.
2489 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
2490 (struct merged_store_group): Add bit_insertion field.
2491 (dump_char_array): Use standard hexadecimal format.
2492 (merged_store_group::merged_store_group): Set bit_insertion to false.
2493 (merged_store_group::apply_stores): Use optimal buffer size. Deal
2494 with BIT_INSERT_EXPR stores. Move up code updating the mask and
2495 also print the mask in the dump file.
2496 (pass_store_merging::gate): Minor tweak.
2497 (imm_store_chain_info::coalesce_immediate): Fix wrong association
2498 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
2499 stores with INTEGER_CST stores.
2500 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
2501 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
2502 and use it throughout. Generate bit insertion sequences if need be.
2503 (pass_store_merging::process_store): Remove redundant condition.
2504 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
2506 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
2508 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
2509 the 128-bit floating point types. Fix function comment.
2511 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2513 * config/aarch64/aarch64-simd.md
2514 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
2516 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
2519 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2521 PR tree-optimization/85989
2522 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
2524 (backprop::intersect_uses): Check it when deciding whether this
2525 is a backedge reference.
2526 (backprop::process_block): Add each phi to m_visited_phis
2527 after visiting it, then clear it at the end.
2529 2018-06-01 Richard Biener <rguenther@suse.de>
2531 * tree-vectorizer.h (vect_dr_stmt): New function.
2532 (vect_get_load_cost): Adjust.
2533 (vect_get_store_cost): Likewise.
2534 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
2535 Use vect_dr_stmt instead of DR_SMTT.
2536 (vect_record_base_alignments): Likewise.
2537 (vect_calculate_target_alignment): Likewise.
2538 (vect_compute_data_ref_alignment): Likewise and make static.
2539 (vect_update_misalignment_for_peel): Likewise.
2540 (vect_verify_datarefs_alignment): Likewise.
2541 (vector_alignment_reachable_p): Likewise.
2542 (vect_get_data_access_cost): Likewise. Pass down
2543 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
2544 (vect_get_peeling_costs_all_drs): Likewise.
2545 (vect_peeling_hash_get_lowest_cost): Likewise.
2546 (vect_enhance_data_refs_alignment): Likewise.
2547 (vect_find_same_alignment_drs): Likewise.
2548 (vect_analyze_data_refs_alignment): Likewise.
2549 (vect_analyze_group_access_1): Likewise.
2550 (vect_analyze_group_access): Likewise.
2551 (vect_analyze_data_ref_access): Likewise.
2552 (vect_analyze_data_ref_accesses): Likewise.
2553 (vect_vfa_segment_size): Likewise.
2554 (vect_small_gap_p): Likewise.
2555 (vectorizable_with_step_bound_p): Likewise.
2556 (vect_prune_runtime_alias_test_list): Likewise.
2557 (vect_analyze_data_refs): Likewise.
2558 (vect_supportable_dr_alignment): Likewise.
2559 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2560 (vect_gen_prolog_loop_niters): Likewise.
2561 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2562 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
2564 (vect_recog_mask_conversion_pattern): Likewise.
2565 (vect_try_gather_scatter_pattern): Likewise.
2566 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
2567 to vect_get_store_cost.
2568 (vect_get_store_cost): Get stmt_info instead of DR.
2569 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
2570 (vect_get_load_cost): Get stmt_info instead of DR.
2572 2018-06-01 Richard Biener <rguenther@suse.de>
2575 * gimple-fold.c (var_decl_component_p): Also allow offsetted
2576 vars wrapped in MEM_REFs.
2578 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2580 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2581 Fix subreg tests so that we only return a choice between
2582 GENERAL_REGS and FP_REGS if the original classes included both.
2584 2018-06-01 Richard Biener <rguenther@suse.de>
2587 * tree-ssa-structalias.c (get_function_part_constraint):
2588 Handle NULL fi->decl.
2589 (find_func_aliases_for_call): Properly handle indirect
2590 fi from direct call.
2591 (find_func_clobbers): Likewise.
2592 (ipa_pta_execute): Likewise.
2593 (create_variable_info_for): For functions that are ifunc_resolver
2594 resolve to a varinfo that contains the result of the resolver call.
2595 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
2598 2018-05-31 Michael Collison <michael.collison@arm.com>
2600 * config/aarch64/aarch64.md:
2601 (*fix_to_zero_extenddfdi2): New pattern.
2602 * gcc.target/aarch64/fix_extend1.c: New testcase.
2604 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
2608 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
2609 and BUILT_IN_STRNCMP_EQ.
2610 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
2611 BUILT_IN_STRNCMP_EQ.
2612 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
2613 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2614 (gimple_fold_builtin): Likewise.
2615 * tree-ssa-strlen.c (compute_string_length): New function.
2616 (determine_min_obsize): New function.
2617 (handle_builtin_string_cmp): New function to handle calls to
2618 string compare functions.
2619 (strlen_optimize_stmt): Add handling to builtin string compare
2621 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2622 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2623 * tree.c (build_common_builtin_nodes): Add new defines of
2624 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
2626 2018-05-31 Jakub Jelinek <jakub@redhat.com>
2629 * bb-reorder.c (pass_partition_blocks::gate): Return false for
2630 functions with naked attribute.
2632 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
2634 * config/i386/sse.md (avx_vec_concat<mode>):
2635 Substitute concat_tg_mode mode attribute with xtg_mode.
2636 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
2637 (concat_tg_mode): Remove mode attribute.
2639 2018-05-31 Martin Sebor <msebor@redhat.com>
2642 * calls.c (alloc_max_size): Correct a logic error/typo.
2643 Treat excessive arguments as infinite. Warn for invalid arguments.
2644 * doc/invoke.texi (-Walloc-size-larger-than): Update.
2646 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
2649 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
2650 and movx for Haswell.
2652 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
2653 Cesar Philippidis <cesar@codesourcery.com>
2656 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
2657 when emitting error on private/firstprivate reductions.
2658 * omp-low.c (lower_omp_target): Avoid reference-type processing
2659 on pointers for firstprivate clause.
2661 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2663 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
2666 * config/aarch64/aarch64-simd.md
2667 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
2668 (aarch64_ld1_x3_<mode>): Likewise
2669 (aarch64_st1x2<VALLDIF:mode>): Likewise
2670 (aarch64_st1_x2_<mode>): Likewise
2671 (aarch64_st1x3<VALLDIF:mode>): Likewise
2672 (aarch64_st1_x3_<mode>): Likewise
2673 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
2674 (vld1_s8_x3): Likewise.
2675 (vld1_u16_x3): Likewise.
2676 (vld1_s16_x3): Likewise.
2677 (vld1_u32_x3): Likewise.
2678 (vld1_s32_x3): Likewise.
2679 (vld1_u64_x3): Likewise.
2680 (vld1_s64_x3): Likewise.
2681 (vld1_f16_x3): Likewise.
2682 (vld1_f32_x3): Likewise.
2683 (vld1_f64_x3): Likewise.
2684 (vld1_p8_x3): Likewise.
2685 (vld1_p16_x3): Likewise.
2686 (vld1_p64_x3): Likewise.
2687 (vld1q_u8_x3): Likewise.
2688 (vld1q_s8_x3): Likewise.
2689 (vld1q_u16_x3): Likewise.
2690 (vld1q_s16_x3): Likewise.
2691 (vld1q_u32_x3): Likewise.
2692 (vld1q_s32_x3): Likewise.
2693 (vld1q_u64_x3): Likewise.
2694 (vld1q_s64_x3): Likewise.
2695 (vld1q_f16_x3): Likewise.
2696 (vld1q_f32_x3): Likewise.
2697 (vld1q_f64_x3): Likewise.
2698 (vld1q_p8_x3): Likewise.
2699 (vld1q_p16_x3): Likewise.
2700 (vld1q_p64_x3): Likewise.
2701 (vst1_s64_x2): Likewise.
2702 (vst1_u64_x2): Likewise.
2703 (vst1_f64_x2): Likewise.
2704 (vst1_s8_x2): Likewise.
2705 (vst1_p8_x2): Likewise.
2706 (vst1_s16_x2): Likewise.
2707 (vst1_p16_x2): Likewise.
2708 (vst1_s32_x2): Likewise.
2709 (vst1_u8_x2): Likewise.
2710 (vst1_u16_x2): Likewise.
2711 (vst1_u32_x2): Likewise.
2712 (vst1_f16_x2): Likewise.
2713 (vst1_f32_x2): Likewise.
2714 (vst1_p64_x2): Likewise.
2715 (vst1q_s8_x2): Likewise.
2716 (vst1q_p8_x2): Likewise.
2717 (vst1q_s16_x2): Likewise.
2718 (vst1q_p16_x2): Likewise.
2719 (vst1q_s32_x2): Likewise.
2720 (vst1q_s64_x2): Likewise.
2721 (vst1q_u8_x2): Likewise.
2722 (vst1q_u16_x2): Likewise.
2723 (vst1q_u32_x2): Likewise.
2724 (vst1q_u64_x2): Likewise.
2725 (vst1q_f16_x2): Likewise.
2726 (vst1q_f32_x2): Likewise.
2727 (vst1q_f64_x2): Likewise.
2728 (vst1q_p64_x2): Likewise.
2729 (vst1_s64_x3): Likewise.
2730 (vst1_u64_x3): Likewise.
2731 (vst1_f64_x3): Likewise.
2732 (vst1_s8_x3): Likewise.
2733 (vst1_p8_x3): Likewise.
2734 (vst1_s16_x3): Likewise.
2735 (vst1_p16_x3): Likewise.
2736 (vst1_s32_x3): Likewise.
2737 (vst1_u8_x3): Likewise.
2738 (vst1_u16_x3): Likewise.
2739 (vst1_u32_x3): Likewise.
2740 (vst1_f16_x3): Likewise.
2741 (vst1_f32_x3): Likewise.
2742 (vst1_p64_x3): Likewise.
2743 (vst1q_s8_x3): Likewise.
2744 (vst1q_p8_x3): Likewise.
2745 (vst1q_s16_x3): Likewise.
2746 (vst1q_p16_x3): Likewise.
2747 (vst1q_s32_x3): Likewise.
2748 (vst1q_s64_x3): Likewise.
2749 (vst1q_u8_x3): Likewise.
2750 (vst1q_u16_x3): Likewise.
2751 (vst1q_u32_x3): Likewise.
2752 (vst1q_u64_x3): Likewise.
2753 (vst1q_f16_x3): Likewise.
2754 (vst1q_f32_x3): Likewise.
2755 (vst1q_f64_x3): Likewise.
2756 (vst1q_p64_x3): Likewise.
2758 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2760 * config/msp430/msp430.c (msp430_output_labelref): Prepend
2761 user_label_prefix to name.
2763 * tree-core.h: Update comment about the format of NAME string
2764 passed to handler in attribute_spec.
2766 * config/msp430/msp430.md: Remove erroneous subreg expression from
2767 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
2768 zero_extend{q,h}isi2.
2770 2018-05-30 Borislav Petkov <bp@suse.de>
2772 * doc/extend.texi: Document some architecture specific
2773 constraints and sort entries.
2775 2018-05-30 Martin Sebor <msebor@redhat.com>
2778 * builtins.c (expand_builtin_stpcpy_1): New function.
2779 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
2780 only if the former succeeds.
2782 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2784 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
2787 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2789 * doc/invoke.texi (-flinker-output): Document
2791 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2793 * passes.c (ipa_write_summaries): Only modify statements if body
2795 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
2796 incrementally linking.
2797 (ipa_passes): Likewise.
2798 * lto-cgraph.c (lto_output_node): When incrementally linking do not
2799 pass down resolution info.
2800 * common.opt (flag_incremental_link): Update info.
2801 * gcc.c (plugin specs): Turn flinker-output=* to
2802 -plugin-opt=-linker-output-known
2803 * toplev.c (compile_file): Also cut compilation when doing incremental
2805 * flag-types. (enum lto_partition_model): Add
2806 LTO_LINKER_OUTPUT_NOLTOREL.
2807 (invoke.texi): Add -flinker-output docs.
2808 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
2809 link same way as WPA; do not stream in dead initializers.
2811 * dwarf2out.c (dwarf2out_die_ref_for_decl,
2812 darf2out_register_external_decl): Support incremental link.
2814 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2816 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
2818 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2820 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
2821 it down to simple_object_copy_lto_debug_sections.
2822 (run_gcc): Determine incremental LTO link time and configure
2823 lto1 into non-wpa mode, disable renaming of debug sections.
2825 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
2827 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2828 descriptions of various incorrectly documented functions.
2830 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
2833 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2834 address check not strict.
2836 2018-05-30 Richard Biener <rguenther@suse.de>
2838 PR tree-optimization/85964
2839 * tracer.c (better_p): Drop initialized count check, we only
2840 call the function with initialized counts now.
2841 (find_best_successor): Do find a best edge if one
2842 has uninitialized count.
2843 (find_best_predecessor): Likewise. Do BB frequency check only
2844 if count is initialized.
2846 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
2848 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
2849 (aarch64_ldrstr_offset_compare): New.
2850 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
2851 load/store orderings.
2852 (aarch64_gen_adjusted_ldpstp): Likewise.
2854 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
2856 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2857 Check for subset of GENERAL_REGS and FP_REGS.
2858 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
2861 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
2863 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
2864 and wi::to_poly_offset. Add the current offset and then check
2865 whether the sum fits, rather than using an unchecked addition of
2866 a checked term. Check for a shwi rather than a uhwi.
2867 * expr.c (get_bit_range): Use tree_to_poly_uint64.
2868 (store_constructor): Use poly_int_tree_p.
2869 (expand_expr_real_1): Likewise.
2870 * function.c (assign_temp): Likewise.
2871 * fold-const.c (const_binop): Use poly_int_tree_p and
2873 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
2875 * ipa-icf-gimple.c (func_checker::compare_operand): Use
2876 to_poly_offset for MEM offsets.
2877 * ipa-icf.c (sem_variable::equals): Likewise.
2878 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
2879 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
2880 wi::to_poly_offset for BIT_FIELD_REF offsets.
2881 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
2883 * var-tracking.c (emit_note_insn_var_location): Use
2884 tree_to_poly_uint64.
2886 2018-05-29 Jim Wilson <jimw@sifive.com>
2888 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
2890 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
2893 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
2894 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
2896 (sse4_1_round<mode>2): Use nonimmediate_operand
2897 for operand 1 predicate.
2899 2018-05-29 Martin Sebor <msebor@redhat.com>
2900 Richard Biener <rguenther@suse.de>
2903 * calls.c (get_size_range): Call determine_value_range instead
2904 of get_value_range..
2905 * tree-vrp.h (determine_value_range): Declared new function.
2906 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
2908 2018-05-29 Richard Biener <rguenther@suse.de>
2910 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
2911 sure to use non-pattern stmts for get_earlier_stmt arguments.
2912 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
2913 called on pattern stmts.
2914 (get_later_stmt): Likewise.
2916 2018-05-29 Martin Liska <mliska@suse.cz>
2918 PR gcov-profile/85759
2919 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
2922 2018-05-29 Jakub Jelinek <jakub@redhat.com>
2924 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
2926 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
2930 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2931 VEC_PACK_FLOAT_EXPR): New tree codes.
2932 * tree-pretty-print.c (op_code_prio): Handle
2933 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2934 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2935 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2936 * tree-inline.c (estimate_operator_cost): Likewise.
2937 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
2938 * fold-const.c (const_binop): Likewise.
2939 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
2940 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2941 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2942 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
2943 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2944 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2945 * expr.c (expand_expr_real_2): Likewise.
2946 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
2947 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
2948 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
2950 * optabs.c (expand_widen_pattern_expr): For
2951 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
2952 sign from result type rather than operand's type.
2953 (expand_binop_directly): For vec_packu_float_optab and
2954 vec_packs_float_optab allow result type to be different from operand's
2956 * optabs-tree.c (optab_for_tree_code): Handle
2957 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2958 VEC_PACK_FLOAT_EXPR. Formatting fixes.
2959 * tree-vect-generic.c (expand_vector_operations_1): Handle
2960 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2961 VEC_PACK_FLOAT_EXPR.
2962 * tree-vect-stmts.c (supportable_widening_operation): Handle
2964 (supportable_narrowing_operation): Handle FLOAT_EXPR.
2965 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
2966 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
2967 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
2968 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
2970 (vec_pack<floatprefix>_float_<mode>): New expander.
2971 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
2973 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
2974 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
2975 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
2976 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
2977 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
2979 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
2980 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
2981 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2982 VEC_PACK_FLOAT_EXPR): Document.
2984 2018-05-29 Richard Biener <rguenther@suse.de>
2986 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
2988 (stmt_vec_info_vec): Make pointer.
2989 (init_stmt_vec_info_vec): Remove.
2990 (free_stmt_vec_info_vec): Likewise.
2991 (set_stmt_vec_info_vec): New function.
2992 (free_stmt_vec_infos): Likewise.
2993 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
2994 (set_vinfo_for_stmt): Likewise.
2995 (get_earlier_stmt): Likewise.
2996 (get_later_stmt): Likewise.
2997 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
2998 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
2999 (vec_info::~vec_info): Free stmt_vec_infos.
3000 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
3001 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
3002 (pass_slp_vectorize::execute): Likewise.
3003 * tree-vect-stmts.c (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 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
3008 the global stmt_vec_info_vec.
3009 * tree-parloops.c (gather_scalar_reductions): Use
3010 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
3013 2018-05-29 Richard Biener <rguenther@suse.de>
3015 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
3017 2018-05-29 Martin Liska <mliska@suse.cz>
3018 David Malcolm <dmalcolm@redhat.com>
3020 * vec.c (test_reverse): New.
3021 (vec_c_tests): Add new test.
3022 * vec.h (vl_ptr>::reverse): New function.
3024 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
3026 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
3028 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
3031 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
3033 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
3035 2018-05-28 Richard Biener <rguenther@suse.de>
3037 PR tree-optimization/85933
3038 * tree-vect-data-refs.c (vect_record_base_alignments): Only
3039 look at stmts marked as vectorizable.
3041 2018-05-28 Richard Biener <rguenther@suse.de>
3043 PR tree-optimization/85934
3044 * tree-vect-generic.c (expand_vector_operations_1): Hoist
3045 vector boolean check before scalar optimization.
3047 2018-05-28 Jakub Jelinek <jakub@redhat.com>
3049 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
3052 2018-05-28 Mark Wielaard <mark@klomp.org>
3054 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
3055 if it is an expression containing a minus sign.
3057 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
3059 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
3061 2018-05-27 Paul Koning <ni1d@arrl.net>
3063 * config/pdp11/pdp11.md (truncsihi2): Remove.
3065 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
3066 Chung-Ju Wu <jasonwucj@gmail.com>
3068 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
3070 (unaligned_store_dw): Ditto.
3071 * config/nds32/nds32-memory-manipulation.c
3072 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
3073 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
3074 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
3075 (emit_setmem_word_loop): Rename to ...
3076 (emit_setmem_doubleword_loop): ... this.
3077 (nds32_gen_dup_4_byte_to_word_value): New function.
3078 (nds32_gen_dup_8_byte_to_double_word_value): New function.
3079 (nds32_expand_setmem_loop): Refine implementation.
3080 (nds32_expand_setmem_loop_v3m): Ditto.
3081 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
3084 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3086 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
3088 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3090 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
3091 (nds32_init_machine_status): Initialize machine->attr_naked_p and
3092 machine->attr_no_prologue_p.
3093 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
3094 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
3095 (nds32_expand_epilogue): Consider attr_naked_p.
3096 (nds32_expand_epilogue_v3pop): Likewise.
3097 (nds32_can_use_return_insn): Likewise.
3098 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
3099 attr_no_prologue_p fields.
3100 * config/nds32/nds32.opt (mret-in-naked-func): New option.
3102 2018-05-27 Jakub Jelinek <jakub@redhat.com>
3105 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
3107 * config/i386/sse.md
3108 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
3110 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
3112 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
3114 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
3116 (*<floatsuffix>floatv2div2sf2): Rename to ...
3117 (*float<floatunssuffix>v2div2sf2): ... this.
3118 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
3119 (float<floatunssuffix>v2div2sf2_mask): ... this.
3120 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
3121 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
3122 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
3124 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
3126 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3128 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3130 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3132 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3134 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
3135 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
3136 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
3137 gen_ufix_truncv8dfv8si2.
3138 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
3139 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
3140 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
3141 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
3142 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
3143 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
3144 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
3145 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
3147 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3151 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
3153 2018-05-25 Jim Wilson <jimw@sifive.com>
3155 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
3156 * config/riscv/riscv.c (struct machine_function): Add
3157 interrupt_handler_p and attribute_checked_p fields.
3158 (riscv_attribute_table): Add interrupt.
3159 (riscv_interrupt_type_p): New.
3160 (riscv_save_reg_p): Save extra regs for interrupt handler.
3161 (riscv_use_save_libcall): Return false for interrupt handler.
3162 (riscv_first_stack_step): Add forward declaration.
3163 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
3164 for interrupt handler with large frame. Use it for saved reg list.
3165 (riscv_expand_prologue): Move flag_stack_usage_info support to
3166 eliminate duplication.
3167 (riscv_expand_epilogue): Generate mret for interrupt handler.
3168 (riscv_epilogue_uses): New.
3169 (riscv_can_use_return_insn): Return false for interrupt handler.
3170 (riscv_function_ok_for_sibcall): Likewise.
3171 (riscv_set_current_function): Add interrupt handler support.
3172 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
3173 * config/riscv/riscv.md (UNSPECV_MRET): New.
3175 (riscv_frflags, riscv_fsflags): Use tab after opcode.
3177 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
3179 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
3181 PR tree-optimization/85712
3182 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
3183 this candidate has already been replaced in-situ by a copy.
3185 2018-05-25 Jason Merrill <jason@redhat.com>
3187 PR c++/80485 - inline function non-zero address.
3188 * symtab.c (nonzero_address): Check DECL_COMDAT.
3190 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
3193 * config/alpha/alpha.md (ashlsi3): New insn pattern.
3194 (*ashlsi_se): Rename from *ashldi_se. Define as sign
3195 extension of SImode operation. Use const123_operand predicate.
3196 (*saddsi_1): Remove.
3197 (*saddl_se_1): Ditto.
3199 (*ssubl_se_1): Ditto.
3200 * config/alpha/predicates.md (const123_operand): New predicate.
3201 * config/alpha/constraints.md (P): Use IN_RANGE.
3203 2018-05-25 Richard Biener <rguenther@suse.de>
3205 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
3207 (ref_maybe_used_by_stmt_p): Likewise.
3208 (stmt_may_clobber_ref_p): Likewise.
3209 (stmt_may_clobber_ref_p_1): Likewise.
3210 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
3212 (ref_maybe_used_by_stmt_p): Likewise.
3213 (stmt_may_clobber_ref_p): Likewise.
3214 (stmt_may_clobber_ref_p_1): Likewise.
3215 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
3216 the alias oracle to disambiguate DRs with stmts DR analysis
3218 (vect_analyze_data_refs): Do not give up on not analyzable
3219 DRs for BB vectorization. Remove code truncating the dataref
3222 2018-05-25 Jakub Jelinek <jakub@redhat.com>
3225 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3226 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
3227 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
3229 2018-05-25 Richard Biener <rguenther@suse.de>
3231 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
3232 function, combining stmt data ref gathering and fatal analysis
3234 (vect_analyze_data_refs): Remove now redudnant code and simplify.
3235 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
3236 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
3237 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
3238 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
3240 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3242 PR tree-optimization/85720
3243 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
3244 SCC if all partitions are builtins.
3245 (version_loop_by_alias_check): New parameter. Generate cancelable
3246 runtime alias check if all partitions are builtins.
3247 (distribute_loop): Update call to above function.
3249 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3251 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
3252 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
3253 (parm_default_def_partition_arg): Ditto.
3254 (set_parm_default_def_partition): Ditto.
3255 (get_parm_default_def_partitions): Ditto and make it static.
3256 (get_undefined_value_partitions): Ditto and make it static.
3257 (remove_ssa_form): Refactor call to init_var_map here.
3258 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
3259 computation for loop region.
3260 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
3261 (register_default_def): Delete.
3262 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
3263 (parm_default_def_partition_arg): Ditto.
3264 (set_parm_default_def_partition): Ditto.
3265 (get_parm_default_def_partitions): Ditto and make it static.
3266 (get_undefined_value_partitions): Ditto and make it static.
3267 (coalesce_with_default, coalesce_with_default): Update comment.
3268 (create_coalesce_list_for_region): New func factored out from
3269 create_outofssa_var_map.
3270 (populate_coalesce_list_for_outofssa): New func factored out from
3271 create_outofssa_var_map and coalesce_ssa_name.
3272 (create_outofssa_var_map): Delete.
3273 (coalesce_ssa_name): Refactor to support live range computation.
3274 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
3275 (get_parm_default_def_partitions): Delete.
3276 (get_undefined_value_partitions): Ditto.
3277 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
3278 computation for loop region.
3279 (new_tree_live_info, loe_visit_block): Ditto.
3280 (live_worklist, set_var_live_on_entry): Ditto.
3281 (calculate_live_on_exit, verify_live_on_entry): Ditto.
3282 * tree-ssa-live.h (struct _var_map): New fields.
3283 (init_var_map): Change decl.
3284 (region_contains_p): New.
3286 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3288 * tree-ssa-live.h (live_merge_and_clear): Delete.
3290 2018-05-25 Richard Biener <rguenther@suse.de>
3293 * tree-dump.c (dequeue_and_dump): Remove access to removed
3294 operand 2 of a SWITCH_EXPR.
3296 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3298 * doc/sourcebuild.texi (vect_double_cond_arith): Include
3299 multiplication and division.
3300 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
3301 (cond_udiv@var{m}, cond_umod@var{m}): Document.
3302 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
3303 (cond_udiv_optab, cond_umod_optab): New optabs.
3304 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
3305 (IFN_COND_RDIV): New internal functions.
3306 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
3307 TRUNC_MOD_EXPR and RDIV_EXPR.
3308 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
3309 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
3311 (SVE_INT_BINARY): Include mult.
3312 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
3313 (optab, sve_int_op): Handle mult.
3314 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
3316 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
3317 for SVE_INT_BINARY_SD.
3319 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3321 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
3322 (optab, sve_int_op): Handle div and udiv.
3323 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
3324 for SVE_INT_BINARY_SD.
3325 (*<optab><mode>3): New insn for the same.
3327 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3329 * tree-vect-patterns.c: Include predict.h.
3330 (vect_recog_divmod_pattern): Restrict check for division support
3331 to when optimizing for size.
3333 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3335 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
3336 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
3337 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
3338 (gimple_match_op::set_op): Likewise.
3339 (gimple_resimplify4): Declare.
3340 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
3341 (expr::gen_transform): Likewise.
3342 (decision_tree::gen): Generate a simplification routine for 4 operands.
3343 * gimple-match-head.c (gimple_simplify): Add an overload for
3344 4 operands. In the top-level function, handle up to 4 call
3345 arguments and call gimple_resimplify4.
3346 (gimple_resimplify4): New function.
3347 (build_call_internal): Pass a fourth operand.
3348 (maybe_push_to_seq): Likewise.
3349 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
3350 Fold VEC_COND_EXPRs of an operation and a default value into
3351 an IFN_COND_* function if possible.
3352 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
3354 (SVE_COND_FP_BINARY): Include them.
3355 (optab, sve_fp_op): Handle them.
3356 (SVE_INT_BINARY_REV): New code iterator.
3357 (SVE_COND_FP_BINARY_REV): New int iterator.
3358 (commutative): New int attribute.
3359 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
3361 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
3363 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
3364 (*cond_<optab><mode>): New patterns for reversed operands.
3366 2018-05-25 Richard Biener <rguenther@suse.de>
3368 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
3369 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
3370 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
3371 (STMT_VINFO_GROUPED_ACCESS): Adjust.
3372 * tree-vect-data-refs.c (everywhere): Adjust users.
3373 * tree-vect-loop.c (everywhere): Likewise.
3374 * tree-vect-slp.c (everywhere): Likewise.
3375 * tree-vect-stmts.c (everywhere): Likewise.
3376 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
3378 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3380 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
3382 (gcc_cv_as_section_exclude): ... this.
3383 Try Solaris as #exclude syntax.
3384 * configure: Regenerate.
3385 * config.in: Regenerate.
3386 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
3388 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
3389 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
3391 * varasm.c (default_elf_asm_named_section): Don't check if
3392 HAVE_GAS_SECTION_EXCLUDE is defined.
3394 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3396 * doc/md.texi: Update the documentation of the cond_* optabs
3397 to mention the new final operand. Fix GET_MODE_NUNITS call.
3398 Describe the scalar case too.
3399 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
3400 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
3402 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
3403 (get_conditional_internal_fn): Update comment.
3404 * tree-vect-loop.c (vectorizable_reduction): Pass the original
3405 accumulator value as a final argument to conditional functions.
3406 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
3407 a define_expand and add an "else" operand. Assert for now that
3408 the else operand is equal to operand 2. Use SVE_INT_BINARY and
3409 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
3410 (*cond_<optab><mode>): New patterns.
3411 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
3412 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
3413 (UNSPEC_COND_EOR): Delete.
3414 (optab): Remove associated mappings.
3415 (SVE_INT_BINARY): New code iterator.
3416 (sve_int_op): Remove int attribute and add "minus" to the code
3418 (SVE_COND_INT_OP): Delete.
3419 (SVE_COND_FP_OP): Rename to...
3420 (SVE_COND_FP_BINARY): ...this.
3422 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3424 * optabs.c (can_reuse_operands_p): New function.
3425 (maybe_legitimize_operands): Try to reuse the results for
3428 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3430 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
3431 Add {q} suffix to insn mnemonic.
3433 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3435 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
3436 (msp430_warn_func_return): New.
3438 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
3440 * fold-const.c (tree_nonzero_bits): New function.
3441 * fold-const.h (tree_nonzero_bits): Likewise.
3442 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
3443 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
3445 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3449 * varasm.c (assemble_alias): Check ifunc_resolver only on
3452 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3455 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
3456 when memory input operand is handled.
3458 2018-05-24 Luis Machado <luis.machado@linaro.org>
3460 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
3462 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
3464 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3466 * match.pd: Delay FMA folds until after vectorization.
3468 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
3471 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3472 address check not strict.
3474 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3476 * gimple-match.h (gimple_match_op): New class.
3477 (mprts_hook): Replace parameters with a gimple_match_op *.
3478 (maybe_build_generic_op): Likewise.
3479 (gimple_simplified_result_is_gimple_val): Replace parameters with
3480 a const gimple_match_op *.
3481 (gimple_simplify): Replace code_helper * and tree * parameters with
3482 a gimple_match_op * parameter.
3483 (gimple_resimplify1): Replace code_helper *, tree and tree *
3484 parameters with a gimple_match_op * parameter.
3485 (gimple_resimplify2): Likewise.
3486 (gimple_resimplify3): Likewise.
3487 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
3488 parameters with a gimple_match_op * parameter.
3489 * gimple-match-head.c (gimple_simplify): Change prototypes of
3490 auto-generated functions to take a gimple_match_op * instead of
3491 separate code_helper * and tree * parameters. Make the same
3492 change in the top-level overload and update calls to the
3493 gimple_resimplify routines. Update calls to the auto-generated
3494 functions and to maybe_push_res_to_seq in the publicly-facing
3495 operation-specific gimple_simplify overloads.
3496 (gimple_match_op::MAX_NUM_OPS): Define.
3497 (gimple_resimplify1): Replace rcode and ops with a single res_op
3498 parameter. Update call to gimple_simplify.
3499 (gimple_resimplify2): Likewise.
3500 (gimple_resimplify3): Likewise.
3501 (mprts_hook): Replace parameters with a gimple_match_op *.
3502 (maybe_build_generic_op): Likewise.
3503 (build_call_internal): Replace type, nargs and ops with
3504 a gimple_match_op *.
3505 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
3506 with a single gimple_match_op *. Update calls to mprts_hook,
3507 build_call_internal and gimple_simplified_result_is_gimple_val.
3508 Factor out code that is common to the tree_code and combined_fn cases.
3509 * genmatch.c (expr::gen_transform): Replace tem_code and
3510 tem_ops with a gimple_match_op called tem_op. Update calls
3511 to the gimple_resimplify functions and maybe_push_res_to_seq.
3512 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
3513 res_ops. Update call to the gimple_resimplify functions.
3514 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
3515 (decision_tree::gen): Make the functions take a gimple_match_op *
3516 called res_op instead of separate res_code and res_ops parameters.
3517 Update call accordingly.
3518 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
3519 and ops with a single res_op parameter. Update calls to
3520 maybe_build_generic_op and maybe_push_res_to_seq.
3521 (fold_stmt_1): Update calls to gimple_simplify and
3522 replace_stmt_with_simplification.
3523 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
3524 and gimple_simplified_result_is_gimple_val.
3525 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
3527 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
3528 with a gimple_match_op *.
3529 (vn_nary_build_or_lookup): Likewise. Update call to
3530 vn_nary_build_or_lookup_1.
3531 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
3532 gimple_match_op *. Update calls to the gimple_resimplify routines
3533 and to gimple_simplified_result_is_gimple_val.
3534 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
3535 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
3536 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
3537 (visit_nary_op): Likewise.
3538 (visit_reference_op_load): Likewise.
3540 2018-05-23 Luis Machado <luis.machado@linaro.org>
3542 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
3543 modifier for printing the step amount.
3545 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
3548 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
3551 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
3553 * doc/sourcebuild.texi (Endianness): New subsubsection.
3555 2018-05-23 Luis Machado <luis.machado@linaro.org>
3557 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3558 <prefetch_dynamic_strides>: New const bool field.
3559 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3560 prefetch_dynamic_strides.
3561 (exynosm1_prefetch_tune): Likewise.
3562 (thunderxt88_prefetch_tune): Likewise.
3563 (thunderx_prefetch_tune): Likewise.
3564 (thunderx2t99_prefetch_tune): Likewise.
3565 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
3567 (aarch64_override_options_internal): Update to set
3568 PARAM_PREFETCH_DYNAMIC_STRIDES.
3569 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3570 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3571 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3572 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3573 prefetch-dynamic-strides setting.
3575 2018-05-23 Luis Machado <luis.machado@linaro.org>
3577 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3578 <minimum_stride>: New const int field.
3579 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3580 minimum_stride field defaulting to -1.
3581 (exynosm1_prefetch_tune): Likewise.
3582 (thunderxt88_prefetch_tune): Likewise.
3583 (thunderx_prefetch_tune): Likewise.
3584 (thunderx2t99_prefetch_tune): Likewise.
3585 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
3586 <default_opt_level>: Set to 3.
3587 (aarch64_override_options_internal): Update to set
3588 PARAM_PREFETCH_MINIMUM_STRIDE.
3589 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3590 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3591 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3592 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3593 stride is constant and is below the minimum stride threshold.
3595 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3597 * config/arm/arm-cpus.in (mode26): Delete.
3598 (armv4): Delete mode26 reference.
3599 * config/arm/arm.c (arm_configure_build_target): Delete use of
3602 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3604 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
3606 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
3607 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
3609 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
3610 (floatunsdidf2): Ditto.
3612 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3614 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
3615 (fixuns_trunc<mode>si2_avx512f): Ditto.
3616 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
3617 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
3618 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
3620 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3622 PR rtl-optimization/79985
3623 * df-scan.c (df_insn_refs_collect): Remove special case for
3624 global registers and asm statements.
3626 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3628 * extend.texi (Global Register Variables): Rewrite the bullet list.
3629 Note that the register is available for allocation. Note that access
3630 via inline asm must use constraints. Add note about async-signal
3631 handlers. Remove paragraph about automagic register selection.
3633 2018-05-23 Richard Biener <rguenther@suse.de>
3635 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
3636 of fixed offset from memset VN.
3638 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
3640 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
3642 (alloc_cand_and_find_basis): Initialize first_interp field.
3643 (slsr_process_mul): Modify first_interp field.
3644 (slsr_process_add): Likewise.
3645 (slsr_process_cast): Modify first_interp field for each new
3647 (slsr_process_copy): Likewise.
3648 (dump_candidate): Dump first_interp field.
3649 (replace_mult_candidate): Process all interpretations, not just
3651 (replace_rhs_if_not_dup): Likewise.
3652 (replace_one_candidate): Likewise.
3654 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
3656 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
3658 (aarch64_needs_frame_chain): New function.
3659 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
3661 2018-05-23 Sudakshina Das <sudi.das@arm.com>
3664 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
3665 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
3666 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
3667 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
3668 as true for strict-align.
3669 (aarch64_can_inline_p): Perform checks even when callee has no
3670 attributes to check for strict alignment.
3671 * doc/extend.texi (AArch64 Function Attributes): Document
3673 * doc/invoke.texi: (AArch64 Options): Likewise.
3675 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
3677 PR tree-optimization/85853
3678 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
3679 the handling of the root of the node to...
3680 (vect_slp_analyze_node_operations_1): ...this new function,
3681 and run the whole thing with the child nodes' def types
3682 set according to their SLP node's def type.
3684 2018-05-23 Richard Biener <rguenther@suse.de>
3687 * tree-data-ref.c (create_runtime_alias_checks): Defer
3688 and ignore overflow warnings.
3690 2018-05-23 Yury Gribov <tetra2005@gmail.com>
3692 PR tree-optimization/85822
3693 * tree-vrp.c (is_masked_range_test): Fix handling of negative
3696 2018-05-23 Richard Biener <rguenther@suse.de>
3698 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3699 memset constants via native_interpret_expr.
3701 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
3704 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
3706 (cgraph_node::create_alias): Likewise.
3707 (cgraph_node::get_availability): Check ifunc_resolver instead
3708 of looking up ifunc attribute.
3709 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3710 * varasm.c (do_assemble_alias): Likewise.
3711 (assemble_alias): Likewise.
3712 (default_binds_local_p_3): Likewise.
3713 * cgraph.h (cgraph_node): Add ifunc_resolver.
3714 (cgraph_node::only_called_directly_or_aliased_p): Return false
3716 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
3718 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
3719 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
3720 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
3721 instead of looking up ifunc attribute.
3723 2018-05-22 Luis Machado <luis.machado@linaro.org>
3725 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
3727 2018-05-22 Martin Sebor <msebor@redhat.com>
3730 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
3731 only when expasion succeeds.
3732 (expand_builtin_strcmp): Same.
3733 (expand_builtin_strncmp): Same.
3735 2018-05-22 Martin Sebor <msebor@redhat.com>
3737 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
3739 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3740 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3742 * config/aarch64/aarch64-ldpstp.md: Replace uses of
3743 aarch64_mem_pair_operand with memory_operand and delete operand swapping
3745 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3746 Add check for legitimate_address.
3747 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
3748 (aarch64_swap_ldrstr_operands): New.
3749 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
3752 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3753 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3755 * config/aarch64/aarch64.md: New patterns to generate stp
3757 (store_pair_sw, store_pair_dw): New patterns to generate stp for
3758 single words and double words.
3759 (load_pair_sw, load_pair_dw): Likewise.
3760 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
3762 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
3764 * config/aarch64/aarch64-ldpstp.md: Modify peephole
3765 for different mode ldpstp and add peephole for merged zero stores.
3767 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3769 (aarch64_gen_store_pair): Rename calls to match new patterns.
3770 (aarch64_gen_load_pair): Rename calls to match new patterns.
3771 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
3772 (load_pair<DREG:mode><DREG2:mode>): ... This.
3773 (store_pair<mode>): Rename to...
3774 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
3775 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
3777 (V_INT_EQUIV): Handle SImode.
3778 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
3781 2018-05-22 Martin Sebor <msebor@redhat.com>
3784 * calls.c (maybe_warn_nonstring_arg): Use string length to set
3785 or ajust the presumed bound on an operation to avoid unnecessary
3788 2018-05-22 Martin Sebor <msebor@redhat.com>
3790 PR tree-optimization/85826
3791 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
3792 assuming that a DECL necesarily has a constant size.
3794 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
3797 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
3799 2018-05-22 Richard Biener <rguenther@suse.de>
3801 PR tree-optimization/85834
3802 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
3803 non-constant and non-zero memset arguments.
3805 2018-05-22 Martin Liska <mliska@suse.cz>
3808 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
3810 2018-05-22 Richard Biener <rguenther@suse.de>
3812 PR tree-optimization/85863
3813 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
3814 comparisons when vectype is specified.
3815 (vectorizable_condition): Do not specify vectype for
3816 vect_is_simple_cond when SLP vectorizing.
3818 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
3821 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
3822 define __ibm128 as long double.
3823 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
3824 as a distinct type when IEEE 128-bit support is enabled.
3825 (init_float128_ieee): Fix up conversions between IFmode and IEEE
3826 128-bit types to use the correct functions.
3827 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
3828 convert between 128-bit floating point types that have different
3829 modes but the same representation, instead of using gen_lowpart to
3831 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
3833 (IFKF_reg): New attributes to give the register constraints for
3835 (extend<mode>tf2_internal): New insns to mark an explicit
3836 conversion between 128-bit floating point types that have a
3837 different mode but share the same representation.
3839 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
3841 PR tree-optimization/85814
3842 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
3843 a null return from get_strinfo when unsharing the next
3844 strinfo in the chain.
3846 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3849 * varasm.c (weak_finish): Clean up weak_decls.
3851 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3853 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
3854 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
3855 UNSPEC_UADALP values.
3856 * config/aarch64/iterators.md (ABAL): New int iterator.
3859 (sur): Add mappings for the above.
3860 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
3862 (aarch64_<sur>abal<mode>_4): Likewise.
3863 (aarch64_<sur>adalp<mode>_3): Likewise.
3864 (<sur>sadv16qi): New define_expand.
3866 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
3868 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
3869 (*movdf_internal): Ditto.
3870 (*rcpsf2_sse): Ditto.
3871 (*rsqrtsf2_sse): Ditto.
3872 (*sqrt<mode>2_sse): Ditto.
3874 2018-05-21 Tamar Christina <tamar.christina@arm.com>
3876 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
3878 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
3879 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
3880 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
3881 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
3883 * config/aarch64/arm_neon.h: Likewise.
3884 * config/aarch64/iterators.md (VQ_I): New.
3886 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
3888 * config.gcc: Add arc/t-multilib-linux to tmake_file for
3890 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
3893 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3895 * config/nds32/constraints.md (S): New constraint.
3896 * config/nds32/nds32.md (call_internal): Use constraint S.
3897 (call_value_internal): Likewise.
3898 (sibcall_internal): Likewise.
3899 (sibcall_value_internal): Likewise.
3901 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3902 Chung-Ju Wu <jasonwucj@gmail.com>
3904 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
3907 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3908 Chung-Ju Wu <jasonwucj@gmail.com>
3910 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
3911 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
3912 (nds32_rtx_costs_impl): Simplify.
3913 (nds32_address_cost_impl): Simplify.
3914 (nds32_init_rtx_costs): New function.
3915 (nds32_rtx_costs_speed_prefer): Likewise.
3916 (nds32_rtx_costs_size_prefer): Likewise.
3917 (nds32_address_cost_speed_prefer): Likewise.
3918 (nds32_address_cost_speed_fwprop): Likewise.
3919 (nds32_address_cost_size_prefer): Likewise.
3920 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
3921 * config/nds32/nds32.c (nds32_option_override): Use
3922 nds32_init_rtx_costs function.
3924 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3926 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
3927 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
3928 (TARGET_PIPELINE_N8): Likewise.
3929 (TARGET_PIPELINE_N10): Likewise.
3930 (TARGET_PIPELINE_N13): Likewise.
3931 (TARGET_PIPELINE_GRAYWOLF): Likewise.
3933 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
3935 * config/nds32/nds32-fpu.md: Update copyright year.
3937 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3939 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
3941 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3943 * config/nds32/nds32.c
3944 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
3945 * config/nds32/nds32.opt (minline-asm-r15): New option.
3947 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3949 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
3951 * config/nds32/nds32.md (abssi2): New pattern.
3953 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
3955 * config/i386/i386.md (rex64namesuffix): New mode attribute.
3956 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
3957 Merge insn pattern from sse_cvtsi2ss<round_name> and
3958 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
3959 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
3960 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
3961 using SWI48 mode iterator.
3962 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
3963 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
3964 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
3965 pattern from sse_cvttss2si<round_saeonly_name>
3966 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
3967 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
3968 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
3969 using SWI48 mode iterator.
3970 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
3971 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
3972 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
3973 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
3974 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
3975 using SWI48 mode iterator.
3976 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
3977 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
3978 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
3979 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
3980 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
3981 SWI48 mode iterator.
3982 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
3983 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
3984 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
3985 pattern from sse_cvttsd2si<round_saeonly_name>
3986 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
3988 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3990 * config/nds32/nds32-md-auxiliary.c
3991 (nds32_valid_smw_lwm_base_p): Refine.
3992 (nds32_output_smw_single_word): Refine.
3993 (nds32_output_smw_double_word): New.
3994 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
3996 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3998 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
3999 (nds32_output_stack_pop): Refine.
4000 (nds32_expand_unaligned_load): Refine.
4001 (nds32_expand_unaligned_store): Refine.
4003 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4004 Chung-Ju Wu <jasonwucj@gmail.com>
4006 * config/nds32/constants.md: Add TP_REGNUM constant.
4007 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
4008 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
4010 * config/nds32/nds32-doubleword.md: Consider flag_pic.
4011 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
4012 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
4013 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
4014 and PIC code generation.
4015 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
4017 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
4019 * config/nds32/nds32.md: Support TLS and PIC.
4020 * config/nds32/nds32.c: Support TLS and PIC.
4021 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
4022 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
4025 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4027 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
4028 mode with E_ prefix.
4030 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4031 Chung-Ju Wu <jasonwucj@gmail.com>
4033 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
4034 * config/nds32/nds32-md-auxiliary.c
4035 (symbolic_reference_mentioned_p): New.
4036 (nds32_legitimize_ict_address): New.
4037 (nds32_expand_ict_move): New.
4038 (nds32_indirect_call_referenced_p): New.
4039 (nds32_symbol_binds_local_p): Delete.
4040 (nds32_long_call_p): Modify.
4041 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
4042 * config/nds32/nds32-protos.h
4043 (symbolic_reference_mentioned_p): Declare.
4044 (nds32_legitimize_ict_address): Declare.
4045 (nds32_expand_ict_move): Declare.
4046 (nds32_indirect_call_referenced_p): Declare.
4047 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
4048 (nds32_relax_group): Use nds32_ict_const_p as condition.
4049 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
4050 (nds32_asm_file_start): Output ict_model directive in asm code.
4051 (nds32_legitimate_address_p): Consider indirect call.
4052 (nds32_print_operand): Consider indirect call.
4053 (nds32_print_operand_address): Consider indirect call.
4054 (nds32_insert_attributes): Handle "indirect_call" attribute.
4055 (TARGET_LEGITIMATE_ADDRESS_P): Define.
4056 (TARGET_LEGITIMATE_CONSTANT_P): Define.
4057 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
4058 (TARGET_DELEGITIMIZE_ADDRESS): Define.
4059 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
4060 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
4061 (TARGET_ICT_MODEL_SMALL): Define.
4062 (TARGET_ICT_MODEL_LARGE): Define.
4063 * config/nds32/nds32.md (movsi): Consider ict model.
4064 (call, call_value): Consider ict model.
4065 (sibcall, sibcall_value): Consider ict model.
4066 * config/nds32/nds32.opt (mict-model): New option.
4067 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
4070 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
4071 Monk Chiang <sh.chiang04@gmail.com>
4072 Jim Wilson <jimw@sifive.com>
4074 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
4075 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
4076 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
4077 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
4078 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
4079 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
4080 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
4081 compute save_libcall_adjustment properly.
4082 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
4083 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
4084 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
4085 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
4086 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
4087 (ABI_SPEC): Handle mabi=ilp32e.
4088 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
4089 (RVE): Add RVE mask.
4090 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
4091 <-march>: Add rv32e as an example.
4093 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
4096 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
4097 (intra_create_variable_infos): Handle C++ constructors.
4099 2018-05-18 Martin Liska <mliska@suse.cz>
4101 * passes.def: Remove a redundant pass.
4103 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
4106 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
4108 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4110 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
4112 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
4114 (armv2, armv2a, armv3, armv3m): Delete architectures.
4115 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
4116 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
4117 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
4119 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
4120 (*mulsidi3adddi): Likewise.
4121 (mulsidi3): Likewise.
4122 (*mulsidi3_nov6): Likewise.
4123 (umulsidi3): Likewise.
4124 (umulsidi3_nov6): Likewise.
4125 (umaddsidi4): Likewise.
4126 (*umulsidi3adddi): Likewise.
4127 (smulsi3_highpart): Likewise.
4128 (*smulsi3_highpart_nov6): Likewise.
4129 (umulsi3_highpart): Likewise.
4130 (*umulsi3_highpart_nov6): Likewise.
4131 * config/arm/arm.h (arm_arch3m): Delete.
4132 * config/arm/arm.c (arm_arch3m): Delete.
4133 (arm_option_override_internal): Update armv3-related comment.
4134 (arm_configure_build_target): Delete use of isa_bit_mode32.
4135 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
4136 (arm_rtx_costs_internal): Delete check of arm_arch3m.
4137 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
4139 (mulusa3): Likewise.
4140 * config/arm/arm-protos.h (arm_arch3m): Delete.
4141 * config/arm/arm-tables.opt: Regenerate.
4142 * config/arm/arm-tune.md: Likewise.
4143 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
4144 deleted architectures.
4146 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4148 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
4149 (armv5t, armv5te): New features.
4150 (ARMv5, ARMv5e): Delete fgroups.
4151 (ARMv5t, ARMv5te): Adjust for above changes.
4153 (armv5, armv5e): Delete arches.
4154 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
4156 (*call_reg_arm): Likewise.
4157 (*call_value_reg_armv5): Likewise.
4158 (*call_value_reg_arm): Likewise.
4159 (*call_symbol): Likewise.
4160 (*call_value_symbol): Likewise.
4161 (*sibcall_insn): Likewise.
4162 (*sibcall_value_insn): Likewise.
4164 (prefetch): Likewise.
4165 (define_split and define_peephole2 dependent on arm_arch5):
4167 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
4169 (TARGET_ARM_QBIT): Likewise.
4170 (TARGET_DSP_MULTIPLY): Likewise.
4171 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
4172 (arm_arch5, arm_arch5e): Delete.
4173 (arm_arch5t, arm_arch5te): Declare.
4174 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
4175 (arm_arch5t): Declare.
4176 (arm_option_reconfigure_globals): Update for the above.
4177 (arm_options_perform_arch_sanity_checks): Update comment, replace
4178 use of arm_arch5 with arm_arch5t.
4179 (use_return_insn): Likewise.
4180 (arm_emit_call_insn): Likewise.
4181 (output_return_instruction): Likewise.
4182 (arm_final_prescan_insn): Likewise.
4183 (arm_coproc_builtin_available): Likewise.
4184 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
4185 arm_arch5e with arm_arch5t and arm_arch5te.
4186 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
4187 (arm_arch5t, arm_arch5te): Declare.
4188 * config/arm/arm-tables.opt: Regenerate.
4189 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
4190 * config/arm/t-multilib: Likewise.
4191 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
4192 instead of arm_arch5.
4193 (*call_reg_thumb1): Likewise.
4194 (*call_value_reg_thumb1_v5): Likewise.
4195 (*call_value_reg_thumb1): Likewise.
4196 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
4198 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
4200 2018-05-18 Martin Liska <mliska@suse.cz>
4202 PR gcov-profile/84846
4203 * doc/gcov.texi: Document -t option of gcov tool.
4205 2018-05-18 Martin Liska <mliska@suse.cz>
4207 PR gcov-profile/84846
4208 * gcov.c (print_usage): Add new -t option.
4209 (process_args): Handle the option.
4210 (generate_results): Use stdout as output when requested by
4213 2018-05-18 Martin Liska <mliska@suse.cz>
4215 PR gcov-profile/84846
4216 * coverage.c (coverage_init): Write PWD to .gcno file.
4217 * doc/gcov.texi: Document how working directory is printed.
4218 * gcov-dump.c (dump_gcov_file): Print PWD.
4219 * gcov.c (output_intermediate_file): Likewise.
4220 (read_graph_file): Read PWD string.
4221 (output_lines): Print PWD.
4223 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4226 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
4227 for retval and return false if all args to phi are zero.
4229 2018-05-18 Richard Biener <rguenther@suse.de>
4231 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
4233 (evrp_dom_walker::before_dom_children): Call it.
4235 2018-05-18 Richard Biener <rguenther@suse.de>
4237 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
4238 results when processing array refs with variable index.
4240 2018-05-18 Toon Moene <toon@moene.org>
4242 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
4243 directly after that of -floop-interchange. Indicate that both
4244 options are enabled by default when specifying -O3.
4246 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4248 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
4249 iterator. Delete separate integer-mode vec_set<mode> expander.
4250 (aarch64_simd_vec_setv2di): Delete.
4251 (vec_setv2di): Delete.
4252 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
4253 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
4254 the "w, r" alternative.
4256 2018-05-18 Martin Liska <mliska@suse.cz>
4258 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
4259 * tree-pass.h (make_pass_lower_switch_O0): New function.
4260 * tree-switch-conversion.c (node_has_low_bound): Remove.
4261 (node_has_high_bound): Likewise.
4262 (node_is_bounded): Likewise.
4263 (class pass_lower_switch): Make it a template type and create
4265 (pass_lower_switch::execute): Add template argument.
4266 (make_pass_lower_switch): New function.
4267 (make_pass_lower_switch_O0): New function.
4268 (do_jump_if_equal): Remove.
4269 (emit_case_nodes): Simplify to just handle all 3 cases and leave
4270 all the hard work to tree optimization passes.
4272 2018-05-18 Martin Liska <mliska@suse.cz>
4274 * dbgcnt.c (limit_low): Renamed from limit.
4275 (limit_high): New variable.
4276 (dbg_cnt_is_enabled): Check for upper limit.
4277 (dbg_cnt): Adjust dumping.
4278 (dbg_cnt_set_limit_by_index): Add new argument for high
4280 (dbg_cnt_set_limit_by_name): Likewise.
4281 (dbg_cnt_process_single_pair): Parse new format.
4282 (dbg_cnt_process_opt): Use strtok.
4283 (dbg_cnt_list_all_counters): Remove 'value' and add
4285 * doc/invoke.texi: Document changes.
4287 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
4289 * doc/sourcebuild.texi (scalar_all_fma): Document.
4290 * tree.def (FMA_EXPR): Delete.
4291 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
4292 * internal-fn.c (ternary_direct): New macro.
4293 (expand_ternary_optab_fn): Likewise.
4294 (direct_ternary_optab_supported_p): Likewise.
4295 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
4296 * builtins.c (fold_builtin_fma): Delete.
4297 (fold_builtin_3): Don't call it.
4298 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
4299 * expr.c (expand_expr_real_2): Likewise.
4300 * fold-const.c (operand_equal_p): Likewise.
4301 (fold_ternary_loc): Likewise.
4302 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
4303 * gimple.c (DEFTREECODE): Likewise.
4304 * gimplify.c (gimplify_expr): Likewise.
4305 * optabs-tree.c (optab_for_tree_code): Likewise.
4306 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
4307 * tree-eh.c (operation_could_trap_p): Likewise.
4308 (stmt_could_throw_1_p): Likewise.
4309 * tree-inline.c (estimate_operator_cost): Likewise.
4310 * tree-pretty-print.c (dump_generic_node): Likewise.
4311 (op_code_prio): Likewise.
4312 * tree-ssa-loop-im.c (stmt_cost): Likewise.
4313 * tree-ssa-operands.c (get_expr_operands): Likewise.
4314 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
4315 * fold-const-call.h (fold_fma): Delete.
4316 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
4317 CFN_FNMA and CFN_FNMS.
4319 * genmatch.c (combined_fn): New enum.
4320 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
4321 (commutative_op): New function.
4322 (commutate): Use it. Handle more than 2 operands.
4323 (dt_operand::gen_gimple_expr): Use commutative_op.
4324 (parser::parse_expr): Allow :c to be used with non-binary
4325 operators if the commutative operand is known.
4326 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
4327 CFN_FMS, CFN_FNMA and CFN_FNMS.
4328 (backprop::process_assign_use): Remove FMA_EXPR handling.
4329 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
4330 (gen_hsa_fma): New function.
4331 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
4332 IFN_FNMA and IFN_FNMS.
4333 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
4334 * gimple-fold.h (follow_all_ssa_edges): Declare.
4335 * gimple-fold.c (follow_all_ssa_edges): New function.
4336 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
4337 gimple_build interface and use follow_all_ssa_edges to fold the result.
4338 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
4339 instead of checking for optabs directly.
4340 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
4341 rather than FMA_EXPRs.
4342 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
4343 call to IFN_FMA instead of an FMA_EXPR.
4345 2018-05-17 Jim Wilson <jimw@sifive.com>
4347 * expr.c (do_tablejump): When converting index to Pmode, if we have a
4348 sign extended promoted subreg, and the range does not have the sign bit
4349 set, then do a sign extend.
4351 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
4352 test, check for sign extended subreg and/or constant operands, and
4353 do a sign extend in that case.
4355 2018-05-17 Steve Ellcey <sellcey@cavium.com>
4357 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
4358 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
4360 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
4361 Change logics_shift_reg to logics_shift_imm.
4362 (thunderx2t99_fp_loadpair_basic): Delete.
4363 (thunderx2t99_fp_storepair_basic): Delete.
4364 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
4365 (thunderx2t99_asimd_polynomial): Delete.
4366 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
4367 and neon_fp_mul_d_scalar_q.
4368 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
4369 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
4370 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
4371 (thunderx2t99_asimd_lut): Add missing tbl types.
4372 (thunderx2t99_asimd_ext): Delete.
4373 (thunderx2t99_asimd_load1_1_mult): Delete.
4374 (thunderx2t99_asimd_load1_2_mult): Delete.
4375 (thunderx2t99_asimd_load1_ldp): New.
4376 (thunderx2t99_asimd_load1): New.
4377 (thunderx2t99_asimd_load2): Add missing *load2* types.
4378 (thunderx2t99_asimd_load3): New.
4379 (thunderx2t99_asimd_load4): New.
4380 (thunderx2t99_asimd_store1_1_mult): Delete.
4381 (thunderx2t99_asimd_store1_2_mult): Delete.
4382 (thunderx2t99_asimd_store2_mult): Delete.
4383 (thunderx2t99_asimd_store2_onelane): Delete.
4384 (thunderx2t99_asimd_store_stp): New.
4385 (thunderx2t99_asimd_store1): New.
4386 (thunderx2t99_asimd_store2): New.
4387 (thunderx2t99_asimd_store3): New.
4388 (thunderx2t99_asimd_store4): New.
4390 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
4392 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
4393 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
4395 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
4396 Segher Boessenkool <segher@kernel.crashing.org>
4399 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
4402 2018-05-17 Richard Biener <rguenther@suse.de>
4404 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
4405 for pruning loop and prune defs feeding only already visited PHIs.
4407 2018-05-17 Richard Biener <rguenther@suse.de>
4409 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
4411 2018-05-17 Bin Cheng <bin.cheng@arm.com>
4412 Richard Biener <rguenther@suse.de>
4414 PR tree-optimization/85793
4415 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
4416 for VMAT_ELEMENTWISE.
4418 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4420 * internal-fn.h (lookup_internal_fn): Declare
4421 * internal-fn.c (lookup_internal_fn): New function.
4422 * gimple.c (gimple_build_call_from_tree): Handle calls to
4424 * gimple-pretty-print.c (dump_gimple_call): Print "." before
4425 internal function names.
4426 * tree-pretty-print.c (dump_generic_node): Likewise.
4427 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
4429 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4431 * gimple-fold.h (gimple_build): Make the function forms take
4432 combined_fn rather than built_in_function.
4433 (gimple_simplify): Likewise.
4434 * gimple-match-head.c (gimple_simplify): Likewise.
4435 * gimple-fold.c (gimple_build): Likewise.
4436 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
4437 rather than gimple_build_call_internal.
4438 (get_initial_defs_for_reduction): Likewise.
4439 (vect_create_epilog_for_reduction): Likewise.
4440 (vectorizable_live_operation): Likewise.
4442 2018-05-17 Martin Liska <mliska@suse.cz>
4444 * gimple-ssa-sprintf.c (format_directive): Do not use
4445 space in between 'G_' and '('.
4447 2018-05-17 Jakub Jelinek <jakub@redhat.com>
4450 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
4451 even if the mask is not all ones.
4454 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
4456 (ix86_gimple_fold_builtin): Likewise.
4459 * config/i386/i386.c: Include tree-vector-builder.h.
4460 (ix86_vector_shift_count): New function.
4461 (ix86_fold_builtin): Fold shift builtins by scalar count.
4462 (ix86_gimple_fold_builtin): Likewise.
4464 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
4465 _mm512_setzero): New intrinsics.
4467 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
4468 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4470 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
4471 code generation for cases where splatting a value is not useful.
4472 * simplify-rtx.c (simplify_ternary_operation): Simplify
4473 vec_merge across a vec_duplicate and a paradoxical subreg forming
4474 a vector mode to a vec_concat.
4476 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
4478 * config.gcc: Support "goldmont-plus".
4479 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4481 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4482 PROCESSOR_GOLDMONT_PLUS.
4483 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
4484 (processor_target_table): Add "goldmont-plus".
4485 (PTA_GOLDMONT_PLUS): Define.
4486 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
4487 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
4488 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
4489 (fold_builtin_cpu): Add "goldmont-plus".
4490 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
4491 (ix86_option_override_internal): Add "goldmont-plus".
4492 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
4493 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
4494 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
4495 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
4497 2018-05-17 Richard Biener <rguenther@suse.de>
4499 PR tree-optimization/85757
4500 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
4501 remove defs that only feed that PHI from further processing.
4503 2018-05-16 Jim Wilson <jimw@sifive.com>
4505 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
4507 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
4509 2018-05-16 Mark Wielaard <mark@klomp.org>
4511 * dwarf2out.c (count_index_strings): New function.
4512 (output_indirect_strings): Call count_index_strings and generate
4513 header for dwarf_version >= 5.
4515 2018-05-16 Mark Wielaard <mark@klomp.org>
4517 * dwarf2out.c (dwarf_FORM): New function.
4518 (set_indirect_string): Use dwarf_FORM.
4519 (reset_indirect_string): Likewise.
4520 (size_of_die): Likewise.
4521 (value_format): Likewise.
4522 (output_die): Likewise.
4523 (add_skeleton_AT_string): Likewise.
4524 (output_macinfo_op): Likewise.
4525 (index_string): Likewise.
4526 (output_index_string_offset): Likewise.
4527 (output_index_string): Likewise.
4528 (count_index_strings): Likewise.
4530 2018-05-16 Carl Love <cel@us.ibm.com>
4532 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4533 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
4535 2018-05-16 Martin Jambor <mjambor@suse.cz>
4537 * ipa-prop.c (ipa_free_all_edge_args): Remove.
4538 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
4540 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
4542 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
4543 (fnma<mode>4): Likewise.
4544 (fms<mode>4): Likewise.
4545 (fnms<mode>4): Likewise.
4546 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
4547 (aarch64_fnma<mode>4): Likewise.
4548 (aarch64_fms<mode>4): Likewise.
4549 (aarch64_fnms<mode>4): Likewise.
4550 (aarch64_fnmadd<mode>4): Likewise.
4552 2018-05-16 Jason Merrill <jason@redhat.com>
4554 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
4556 2018-05-16 Richard Biener <rguenther@suse.de>
4558 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
4559 (dump_stmt_cost): Declare.
4560 (add_stmt_cost): Dump cost we add.
4561 (add_stmt_costs): New function.
4562 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
4564 (vect_analyze_stmt): Adjust prototype.
4565 (vectorizable_condition): Likewise.
4566 (vectorizable_live_operation): Likewise.
4567 (vectorizable_reduction): Likewise.
4568 (vectorizable_induction): Likewise.
4569 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
4570 cost vector to pass to vectorizable_ and record afterwards.
4571 (vect_model_reduction_cost): Take cost vector argument and adjust.
4572 (vect_model_induction_cost): Likewise.
4573 (vectorizable_reduction): Likewise.
4574 (vectorizable_induction): Likewise.
4575 (vectorizable_live_operation): Likewise.
4576 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
4577 SLP_TREE_NUMBER_OF_VEC_STMTS.
4578 (vect_analyze_slp_cost_1): Remove.
4579 (vect_analyze_slp_cost): Likewise.
4580 (vect_slp_analyze_node_operations): Take visited args and
4581 a target cost vector. Avoid processing already visited stmt sets.
4582 (vect_slp_analyze_operations): Use a local cost vector to gather
4583 costs and register those of non-discarded instances.
4584 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
4585 (vect_schedule_slp_instance): Remove copying of
4586 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
4588 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
4589 adding cost. Record cost entry location.
4590 (vect_prologue_cost_for_slp_op): Function to compute cost of
4591 a constant or invariant generated for SLP vect in the prologue,
4592 split out from vect_analyze_slp_cost_1.
4593 (vect_model_simple_cost): Make static. Adjust for SLP costing.
4594 (vect_model_promotion_demotion_cost): Likewise.
4595 (vect_model_store_cost): Likewise, make static.
4596 (vect_model_load_cost): Likewise.
4597 (vectorizable_bswap): Add cost vector arg and adjust.
4598 (vectorizable_call): Likewise.
4599 (vectorizable_simd_clone_call): Likewise.
4600 (vectorizable_conversion): Likewise.
4601 (vectorizable_assignment): Likewise.
4602 (vectorizable_shift): Likewise.
4603 (vectorizable_operation): Likewise.
4604 (vectorizable_store): Likewise.
4605 (vectorizable_load): Likewise.
4606 (vectorizable_condition): Likewise.
4607 (vectorizable_comparison): Likewise.
4608 (can_vectorize_live_stmts): Likewise.
4609 (vect_analyze_stmt): Likewise.
4610 (vect_transform_stmt): Adjust calls to vectorizable_*.
4611 * tree-vectorizer.c: Include gimple-pretty-print.h.
4612 (dump_stmt_cost): New function.
4614 2018-05-16 Richard Biener <rguenther@suse.de>
4616 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
4617 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
4618 * tree-ssa-dse.c: Include tree-ssa-loop.h.
4619 (check_name): New callback.
4620 (dse_classify_store): Track cycles via a visited bitmap of PHI
4621 defs and simplify handling of in-loop and across loop dead stores
4622 and properly fail for loop-variant refs. Handle byte-tracking with
4623 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
4626 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
4628 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
4629 (vect_get_mask_type_for_stmt): Likewise.
4630 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
4632 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
4633 to determine the statement's vector type and the vector type that
4634 should be used for calculating nunits. Deal with cases in which
4635 the type has to be deferred.
4636 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
4637 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
4638 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
4639 (vect_determine_vf_for_stmt): New functions, split out from...
4640 (vect_determine_vectorization_factor): ...here.
4641 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
4642 (vect_get_mask_type_for_stmt): New functions, split out from
4643 vect_determine_vectorization_factor.
4645 2018-05-16 Richard Biener <rguenther@suse.de>
4647 * tree-cfg.c (verify_gimple_assign_ternary): Properly
4648 verify the [VEC_]COND_EXPR embedded comparison.
4650 2018-05-15 Martin Sebor <msebor@redhat.com>
4652 PR tree-optimization/85753
4653 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
4654 RECORD_TYPE in addition to ARRAY_TYPE.
4656 2018-05-15 Martin Sebor <msebor@redhat.com>
4659 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
4661 2018-05-15 Richard Biener <rguenther@suse.de>
4663 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
4664 add by_clobber_p one. Change algorithm to collect all defs
4665 representing uses we need to walk and try reducing them to
4666 a single one before failing.
4667 (dse_dom_walker::dse_optimize_stmt): Adjust.
4669 2018-05-13 Mark Wielaard <mark@klomp.org>
4671 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
4672 (size_of_loc_descr): Likewise.
4673 (output_loc_operands): Likewise.
4674 (output_loc_operands_raw): Likewise.
4675 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
4676 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
4677 (hash_loc_operands): Likewise.
4678 (compare_loc_operands): Likewise.
4680 2018-05-14 Mark Wielaard <mark@klomp.org>
4682 * dwarf2out.c (count_index_addrs): New function.
4683 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
4685 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4687 PR tree-optimization/83648
4688 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
4689 return value as malloc candidate.
4691 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4694 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
4695 param as true in call to suggest_attribute.
4697 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
4699 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
4700 -mreadonly-in-sdata.
4702 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4704 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
4706 (aarch64_crypto_aesd_fused): Likewise.
4708 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
4710 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
4711 (movsi_aarch64): Likewise.
4712 (load_pairsi): Likewise.
4713 (load_pairdi): Likewise.
4714 (store_pairsi): Likewise.
4715 (store_pairdi): Likewise.
4716 (load_pairsf): Likewise.
4717 (load_pairdf): Likewise.
4718 (store_pairsf): Likewise.
4719 (store_pairdf): Likewise.
4720 (zero_extend): Likewise.
4721 (trunc): Swap alternatives.
4722 (fcvt_target): Add '?' to prefer w over r.
4724 2018-05-14 Jakub Jelinek <jakub@redhat.com>
4727 * config/i386/i386.md: Disallow non-commutative arithmetics in
4728 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
4729 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
4730 in the peephole2 before it.
4732 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
4734 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
4735 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
4736 (ix86_handle_option): Handle -mcldemote.
4737 * config.gcc: New header.
4738 * config/i386/cldemoteintrin.h: New file.
4739 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
4740 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4742 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4743 OPTION_MASK_ISA_CLDEMOTE.
4744 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
4745 (ix86_valid_target_attribute_inner_p): Ditto.
4746 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
4747 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
4748 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
4749 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
4750 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
4752 * config/i386/i386.opt: Add -mcldemote.
4753 * config/i386/x86intrin.h: New header.
4754 * doc/invoke.texi: Add -mcldemote.
4756 2018-05-14 Richard Biener <rguenther@suse.de>
4758 * doc/match-and-simplify.texi: Adjust :s documentation.
4760 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
4762 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
4763 intended memcpy size.
4764 (REORDER_45): Likewise.
4766 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
4768 * sort.cc: New file.
4769 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
4770 * vec.c (qsort_chk): Use gcc_qsort.
4771 * Makefile.in (OBJS-libcommon): Add sort.o.
4772 (build/sort.o): New target. Use it...
4773 (BUILD_RTL): ... here, and...
4774 (build/gencfn-macros): ... here, and...
4775 (build/genmatch): ... here.
4777 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4778 Chung-Ju Wu <jasonwucj@gmail.com>
4780 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
4781 * config/nds32/nds32-graywolf.md: New file.
4782 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
4783 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
4785 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
4786 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
4787 * config/nds32/nds32.md (pipeline_model): Add graywolf.
4788 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
4789 * config/nds32/pipelines.md: Include n15 settings.
4791 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4792 Chung-Ju Wu <jasonwucj@gmail.com>
4794 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
4795 * config/nds32/nds32-n13.md: New file.
4796 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
4797 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
4799 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
4800 * config/nds32/nds32.md (pipeline_model): Add n13.
4801 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
4802 * config/nds32/pipelines.md: Include n13 settings.
4804 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4805 Chung-Ju Wu <jasonwucj@gmail.com>
4807 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
4808 * config/nds32/nds32-n10.md: New file.
4809 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
4810 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
4812 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
4813 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
4814 * config/nds32/nds32.md (pipeline_model): Add n10.
4815 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
4816 * config/nds32/pipelines.md: Include n10 settings.
4818 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
4819 Kito Cheng <kito.cheng@gmail.com>
4820 Chung-Ju Wu <jasonwucj@gmail.com>
4822 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
4823 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4824 Add enum values for DSP extension instructions.
4825 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
4827 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
4828 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
4830 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
4831 * config/nds32/nds32-dspext.md: New file for DSP implementation.
4832 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
4833 * config/nds32/nds32-intrinsic.md: Likewise.
4834 * config/nds32/nds32_intrinsic.h: Likewise.
4835 * config/nds32/nds32-md-auxiliary.c: Likewise.
4836 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
4837 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
4838 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
4839 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
4840 * config/nds32/nds32-protos.h: New declarations for DSP extension.
4841 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
4842 TYPE_DMAC in switch statement.
4843 * config/nds32/nds32.c: New checking and implementation for DSP
4844 extension instructions.
4845 * config/nds32/nds32.h: Likewise.
4846 * config/nds32/nds32.md: Likewise.
4847 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
4848 * config/nds32/predicates.md: Implement new predicates for DSP
4851 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4853 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
4854 Reformat alternatives and attributes so it is easier to identify
4855 which constraints/attributes go with which instruction.
4856 (mov<mode>_hardfloat32, FMOVE64): Likewise.
4857 (mov<mode>_softfloat32, FMOVE64): Likewise.
4858 (mov<mode>_hardfloat64, FMOVE64): Likewise.
4859 (mov<mode>_softfloat64, FMOVE64): Likewise.
4861 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
4863 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4865 (Basic PowerPC Built-in Functions): The new name of the
4866 subsection previously known as "PowerPC Built-in Functions".
4867 (Basic PowerPC Built-in Functions Available on all Configurations):
4869 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
4870 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
4871 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
4872 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
4874 2018-05-11 Martin Jambor <mjambor@suse.cz>
4877 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
4880 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4883 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
4885 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
4887 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
4888 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
4889 (ix86_handle_option): Handle -mwaitpkg.
4890 * config.gcc: New header.
4891 * config/i386/cpuid.h (bit_WAITPKG): New bit.
4892 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
4893 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
4895 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4896 OPTION_MASK_ISA_WAITPKG.
4897 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
4898 (ix86_option_override_internal): Add PTA_WAITPKG.
4899 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
4900 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
4901 IX86_BUILTIN_TPAUSE.
4902 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
4903 __builtin_ia32_umwait and __builtin_ia32_tpause.
4904 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
4905 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
4906 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
4907 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
4908 UNSPECV_TPAUSE): New.
4909 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
4910 * config/i386/i386.opt: Add -mwaitpkg.
4911 * config/i386/waitpkgintrin.h: New file.
4912 * config/i386/x86intrin.h: New header.
4913 * doc/invoke.texi: Add -mwaitpkg.
4915 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4918 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
4920 (cortex-m0): Use armv6s-m isa.
4921 (cortex-m0plus): Likewise.
4922 (cortex-m1): Likewise.
4923 (cortex-m0.small-multiply): Likewise.
4924 (cortex-m0plus.small-multiply): Likewise.
4925 (cortex-m1.small-multiply): Likewise.
4927 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
4928 Jakub Jelinek <jakub@redhat.com>
4930 PR tree-optimization/85692
4931 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
4932 source permute as well.
4934 2018-05-11 Martin Liska <mliska@suse.cz>
4937 * doc/extend.texi: Document LLVM style format for no_sanitize
4940 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
4942 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
4943 mode_supports_vsx_dform_quad to mode_supports_dq_form.
4944 (mode_supports_vsx_dform_quad): Likewise.
4945 (mode_supports_vmx_dform): Move these functions to be next to the
4946 other mode_supports functions.
4947 (mode_supports_dq_form): Likewise.
4948 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
4949 mode_supports_dq_form.
4950 (reg_offset_addressing_ok_p): Likewise.
4951 (offsettable_ok_by_alignment): Likewise.
4952 (rs6000_legitimate_offset_address_p): Likewise.
4953 (legitimate_lo_sum_address_p): Likewise.
4954 (rs6000_legitimize_address): Likewise.
4955 (rs6000_legitimize_reload_address): Likewise.
4956 (rs6000_secondary_reload_inner): Likewise.
4957 (rs6000_preferred_reload_class): Likewise.
4958 (rs6000_output_move_128bit): Likewise.
4960 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4962 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
4963 Generate SImode target register for null target.
4964 <case IX86_BUILTIN_XGETBV>: Ditto.
4965 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
4966 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
4968 2018-05-10 Carl Love <cel@us.ibm.com>
4970 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4971 dcbtt and dcbtstt if operands[2] is 0.
4973 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4976 * config/i386/sse.md (usadv64qi): New expander.
4978 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
4980 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
4981 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
4982 -maltivec=be support.
4983 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
4984 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
4985 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
4986 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
4987 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
4988 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
4989 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
4990 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
4991 altivec_vsumsws): Adjust.
4992 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
4993 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
4994 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
4996 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
4997 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
4998 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
4999 (altivec_lve<VI_char>x): Delete expand.
5000 (*altivec_lve<VI_char>x_internal): Rename to...
5001 (altivec_lve<VI_char>x): ... this.
5002 (altivec_lvxl_<mode>): Delete expand.
5003 (*altivec_lvxl_<mode>_internal): Rename to ...
5004 (altivec_lvxl_<mode>): ... this.
5005 (altivec_stvxl_<mode>): Delete expand.
5006 (*altivec_stvxl_<mode>_internal): Rename to ...
5007 (altivec_stvxl_<mode>): ... this.
5008 (altivec_stve<VI_char>x): Delete expand.
5009 (*altivec_stve<VI_char>x_internal): Rename to ...
5010 (altivec_stve<VI_char>x): ... this.
5011 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
5012 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
5013 reduc_plus_scal_<mode>): Adjust.
5014 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
5016 (rs6000_cpu_cpp_builtins): Adjust.
5017 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
5018 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
5019 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
5020 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
5021 -maltivec=be support.
5022 (rs6000_split_vec_extract_var): Adjust.
5023 (rs6000_split_v4si_init): Adjust.
5024 (swap_selector_for_mode): Delete.
5025 (altivec_expand_lvx_be, altivec_expand_stvx_be,
5026 altivec_expand_stvex_be): Delete.
5027 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
5028 -maltivec=be support.
5029 (rs6000_gimple_fold_builtin): Ditto.
5030 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
5032 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
5033 (TARGET_DIRECT_MOVE_64BIT): Adjust.
5034 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
5035 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
5036 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
5037 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
5038 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
5039 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
5040 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
5041 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
5042 anonymous split): Adjust.
5043 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
5044 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
5046 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
5048 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
5049 when --with-gxx-include-dir is also specified.
5050 * configure: Regenerate.
5052 2018-05-09 Jim Wilson <jimw@sifive.com>
5055 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
5056 * config/riscv/t-withmultilib: New.
5057 * config/riscv/withmultilib.h: New.
5058 * doc/install.texi: Document RISC-V --with-multilib-list support.
5060 2018-05-09 Richard Biener <rguenther@suse.de>
5062 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
5064 (vect_bb_vectorization_profitable_p): Adjust. Compute
5065 actual scalar cost using the cost vector and the add_stmt_cost
5068 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5070 PR rtl-optimization/85645
5071 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
5072 in the REG_CFA_REGISTER note for LR, don't leave it empty.
5074 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5076 PR rtl-optimization/85645
5077 * shrink-wrap.c (spread_components): Return a boolean saying if
5078 anything was changed.
5079 (try_shrink_wrapping_separate): Iterate spread_components until
5080 nothing changes anymore.
5082 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5084 PR rtl-optimization/85645
5085 * regrename.c (build_def_use): Also kill the chains that include the
5086 destination of a REG_CFA_REGISTER note.
5088 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5090 PR rtl-optimization/85645
5091 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
5092 insn that has a REG_CFA_REGISTER note.
5094 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
5096 * cfgexpand.c (expand_clobber): New function.
5097 (expand_gimple_stmt_1): Use it.
5098 * tree-vect-stmts.c (vect_clobber_variable): New function,
5100 (vectorizable_simd_clone_call): ...here.
5101 (vectorizable_store): Emit a clobber either side of an
5102 IFN_STORE_LANES sequence.
5103 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
5105 2018-05-09 Tom de Vries <tom@codesourcery.com>
5108 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
5109 (define_insn "trap_if_false"): Add exit after trap.
5111 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
5113 PR rtl-optimization/85638
5114 * bb-reorder.c: Include common/common-target.h.
5115 (create_forwarder_block): New function extracted from...
5116 (fix_up_crossing_landing_pad): ...here. Rename into...
5117 (dw2_fix_up_crossing_landing_pad): ...this.
5118 (sjlj_fix_up_crossing_landing_pad): New function.
5119 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
5120 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
5121 from both partitions and exit the loop after one iteration.
5123 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5126 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5128 (Basic PowerPC Built-in Functions): The new name of the
5129 subsection previously known as "PowerPC Built-in Functions".
5130 (Basic PowerPC Built-in Functions Available on all Configurations):
5132 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5134 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5136 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5138 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5141 2018-05-08 Jim Wilson <jimw@sifive.com>
5143 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
5144 (LD_EMUL_SUFFIX): New.
5145 (LINK_SPEC): Use it.
5147 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5149 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5151 (Basic PowerPC Built-in Functions): The new name of the
5152 subsection previously known as "PowerPC Built-in Functions".
5153 (Basic PowerPC Built-in Functions Available on all Configurations):
5155 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5157 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5159 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5161 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5164 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5167 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
5168 after cmpelim optimization.
5170 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
5172 * config.gcc: Support "goldmont".
5173 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
5174 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5176 * config/i386/i386.c (m_GOLDMONT): Define.
5177 (processor_target_table): Add "goldmont".
5178 (PTA_GOLDMONT): Define.
5179 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
5180 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
5181 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
5182 (fold_builtin_cpu): Add "goldmont".
5183 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
5184 (ix86_option_override_internal): Add "goldmont".
5185 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
5186 (processor_type): Add PROCESSOR_GOLDMONT.
5187 * config/i386/i386.md: Add CPU "glm".
5188 * config/i386/glm.md: New file.
5189 * config/i386/x86-tune.def: Add m_GOLDMONT.
5190 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
5192 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5195 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
5197 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
5198 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
5199 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
5202 * config/i386/i386.c (ix86_fold_builtin): Handle
5203 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
5206 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
5207 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
5209 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
5212 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
5213 (check_arch): Likewise.
5214 (check_fpu): Return the result rather than printing it.
5215 (end arch): Fix operator precedence.
5216 (end cpu): Likewise.
5217 (END): Print the result from check_fpu.
5219 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5220 Alan Hayward <alan.hayward@arm.com>
5221 David Sherwood <david.sherwood@arm.com>
5223 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
5224 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
5225 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
5226 (*fcmuo<mode>_and): New patterns.
5228 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5230 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
5231 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
5232 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
5233 (cmp_op, sve_imm_con): New code attributes.
5234 (SVE_COND_INT_CMP, imm_con): Delete.
5235 (cmp_op): Remove above unspecs from int attribute.
5236 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
5238 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
5239 comparison-specific unspecs.
5240 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
5241 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
5242 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
5243 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
5244 (*vec_fcm<cmp_op><mode>): Rename to...
5245 (*fcm<cmp_op><mode>): ...this and adjust likewise.
5246 (*vec_fcmuo<mode>): Rename to...
5247 (*fcmuo<mode>): ...this and adjust likewise.
5248 (*pred_fcm<cmp_op><mode>): New pattern.
5249 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
5250 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
5252 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
5254 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
5255 (aarch64_emit_sve_predicated_cond): New function.
5256 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
5257 (aarch64_emit_unspec_cond_or): Replace with...
5258 (aarch64_emit_sve_or_conds): ...this new function. Use
5259 aarch64_emit_sve_ptrue_op for the individual comparisons and
5260 aarch64_emit_binop to OR them together.
5261 (aarch64_emit_inverted_unspec_cond): Replace with...
5262 (aarch64_emit_sve_inverted_cond): ...this new function. Use
5263 aarch64_emit_sve_ptrue_op for the comparison and
5264 aarch64_emit_unop to invert the result.
5265 (aarch64_expand_sve_vec_cmp_float): Update after the above
5266 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
5268 2018-05-07 Nathan Sidwell <nathan@acm.org>
5270 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
5271 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
5272 (Backwards Compatibility): Likewise.
5274 2018-05-07 Luis Machado <luis.machado@linaro.org>
5278 2018-05-07 Luis Machado <luis.machado@linaro.org>
5280 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5281 <prefetch_dynamic_strides>: New const bool field.
5282 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5283 prefetch_dynamic_strides.
5284 (exynosm1_prefetch_tune): Likewise.
5285 (thunderxt88_prefetch_tune): Likewise.
5286 (thunderx_prefetch_tune): Likewise.
5287 (thunderx2t99_prefetch_tune): Likewise.
5288 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5290 (aarch64_override_options_internal): Update to set
5291 PARAM_PREFETCH_DYNAMIC_STRIDES.
5292 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5293 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5294 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5295 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5296 prefetch-dynamic-strides setting.
5298 2018-05-07 Luis Machado <luis.machado@linaro.org>
5300 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5301 <minimum_stride>: New const int field.
5302 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5303 minimum_stride field.
5304 (exynosm1_prefetch_tune): Likewise.
5305 (thunderxt88_prefetch_tune): Likewise.
5306 (thunderx_prefetch_tune): Likewise.
5307 (thunderx2t99_prefetch_tune): Likewise.
5308 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5309 (aarch64_override_options_internal): Update to set
5310 PARAM_PREFETCH_MINIMUM_STRIDE.
5311 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5312 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5313 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5314 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5315 stride is constant and is below the minimum stride threshold.
5317 2018-05-07 Luis Machado <luis.machado@linaro.org>
5319 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
5322 2018-05-07 Luis Machado <luis.machado@linaro.org>
5324 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5325 <prefetch_dynamic_strides>: New const bool field.
5326 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5327 prefetch_dynamic_strides.
5328 (exynosm1_prefetch_tune): Likewise.
5329 (thunderxt88_prefetch_tune): Likewise.
5330 (thunderx_prefetch_tune): Likewise.
5331 (thunderx2t99_prefetch_tune): Likewise.
5332 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5334 (aarch64_override_options_internal): Update to set
5335 PARAM_PREFETCH_DYNAMIC_STRIDES.
5336 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5337 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5338 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5339 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5340 prefetch-dynamic-strides setting.
5342 2018-05-07 Luis Machado <luis.machado@linaro.org>
5344 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5345 <minimum_stride>: New const int field.
5346 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5347 minimum_stride field.
5348 (exynosm1_prefetch_tune): Likewise.
5349 (thunderxt88_prefetch_tune): Likewise.
5350 (thunderx_prefetch_tune): Likewise.
5351 (thunderx2t99_prefetch_tune): Likewise.
5352 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5353 (aarch64_override_options_internal): Update to set
5354 PARAM_PREFETCH_MINIMUM_STRIDE.
5355 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5356 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5357 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5358 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5359 stride is constant and is below the minimum stride threshold.
5361 2018-05-06 Jakub Jelinek <jakub@redhat.com>
5364 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
5365 the type is addressable. Don't force op into register if it has
5368 2018-05-05 Roland McGrath <mcgrathr@google.com>
5371 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
5372 any section for which we don't know a specific type it should have,
5373 regardless of name. Previously this was done only for the exact
5374 names ".init_array", ".fini_array", and ".preinit_array".
5375 (default_elf_asm_named_section): Add comment about
5376 relationship with default_section_type_flags and SECTION_NOTYPE.
5377 (get_section): Don't consider it a type conflict if one side has
5378 SECTION_NOTYPE and the other doesn't, as long as neither has the
5379 SECTION_BSS et al used in the default_section_type_flags logic.
5381 2018-05-05 Tom de Vries <tom@codesourcery.com>
5384 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
5385 (workaround_barsyncs): New function.
5386 (nvptx_reorg): Use workaround_barsyncs.
5387 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5388 (define_expand "nvptx_membar_cta"): New define_expand.
5389 (define_insn "*nvptx_membar_cta"): New insn.
5391 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
5393 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
5394 To improve optimization opportunities.
5395 * builtin-types.def: The new needed builtin types for the above.
5397 2018-05-04 Richard Biener <rguenther@suse.de>
5399 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
5400 * gimple-ssa-store-merging.c
5401 (imm_store_chain_info::output_merged_store): Remove redundant create,
5402 release split_store vector contents on failure.
5403 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
5404 scalar stmt vector on cache hit.
5406 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
5408 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
5410 * config.gcc (powerpc-xilinx-eabi*): Remove.
5411 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
5413 (fusion_addis_mem_combo_load): Ditto.
5414 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
5416 (rs6000_cpu_cpp_builtins): Ditto.
5417 * config/rs6000/rs6000-linux.c
5418 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
5419 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
5420 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
5422 (rs6000_setup_reg_addr_masks): Ditto.
5423 (rs6000_init_hard_regno_mode_ok): Ditto.
5424 (rs6000_option_override_internal): Ditto.
5425 (legitimate_lo_sum_address_p): Ditto.
5426 (rs6000_legitimize_address): Ditto.
5427 (rs6000_legitimize_reload_address): Ditto.
5428 (rs6000_legitimate_address_p): Ditto.
5429 (abi_v4_pass_in_fpr): Ditto.
5430 (setup_incoming_varargs): Ditto.
5431 (rs6000_gimplify_va_arg): Ditto.
5432 (rs6000_split_multireg_move): Ditto.
5433 (rs6000_savres_strategy): Ditto.
5434 (rs6000_emit_prologue_components): Ditto.
5435 (rs6000_emit_epilogue_components): Ditto.
5436 (rs6000_emit_prologue): Ditto.
5437 (rs6000_emit_epilogue): Ditto.
5438 (rs6000_elf_file_end): Ditto.
5439 (rs6000_function_value): Ditto.
5440 (rs6000_libcall_value): Ditto.
5441 * config/rs6000/rs6000.h: Ditto.
5442 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
5443 (TARGET_MINMAX): ... this. New.
5444 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
5445 * config/rs6000/rs6000.md: Remove Xilinx FP support.
5446 (*movsi_internal1_single): Delete.
5447 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
5448 mfpu=, mxilinx-fpu): Delete.
5449 * config/rs6000/singlefp.h: Delete.
5450 * config/rs6000/sysv4.h: Remove Xilinx FP support.
5451 * config/rs6000/t-rs6000: Ditto.
5452 * config/rs6000/t-xilinx: Delete.
5453 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
5454 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
5455 (VStype_simple): Delete.
5456 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
5457 * config/rs6000/xfpu.h: Delete.
5458 * config/rs6000/xfpu.md: Delete.
5459 * config/rs6000/xilinx.h: Delete.
5460 * config/rs6000/xilinx.opt: Delete.
5461 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
5462 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
5464 2018-05-04 Tom de Vries <tom@codesourcery.com>
5467 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
5470 2018-05-04 Richard Biener <rguenther@suse.de>
5473 * tree-complex.c (update_complex_assignment): We are always in SSA form.
5474 (expand_complex_div_wide): Likewise.
5475 (expand_complex_operations_1): Likewise.
5476 (expand_complex_libcall): Preserve EH info of the original stmt.
5477 (tree_lower_complex): Handle removed blocks.
5478 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
5479 on complex multiplication and division libcall builtins.
5481 2018-05-04 Richard Biener <rguenther@suse.de>
5484 * fold-const.c (negate_expr_p): Restrict negation of operand
5485 zero of a division to when we know that can happen without
5487 (fold_negate_expr_1): Likewise.
5489 2018-05-04 Jakub Jelinek <jakub@redhat.com>
5492 * real.h (real_nextafter): Declare.
5493 * real.c (real_nextafter): New function.
5494 * fold-const-call.c (fold_const_nextafter): New function.
5495 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
5496 CASE_CFN_NEXTTOWARD.
5497 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
5498 even when arg1_mode is different from arg0_mode.
5500 2018-05-03 Nathan Sidwell <nathan@acm.org>
5502 * doc/extend.texi (Deprecated Features): Remove
5504 (Backwards Compatibility): Likewise.
5505 * doc/invoke.texi (C++ Language Options): Likewise.
5506 (C++ Dialect Options): Likewise.
5508 2018-05-03 Jakub Jelinek <jakub@redhat.com>
5511 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
5512 _mm512_mask_mullox_epi64): New intrinsics.
5514 2018-05-03 Tom de Vries <tom@codesourcery.com>
5517 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5518 dump files): Add offload-tree.
5520 2018-05-03 Richard Biener <rguenther@suse.de>
5522 PR tree-optimization/85615
5523 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
5524 to loops not nested in BBs loop father to avoid creating multi-entry
5527 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5529 PR tree-optimization/70291
5530 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
5531 arguments. Change return type to tree. Emit libcall as a new
5532 statement rather than replacing existing one when inplace_p is true.
5533 (expand_complex_multiplication_components): New function.
5534 (expand_complex_multiplication): Expand floating-point complex
5535 multiplication using the above.
5536 (expand_complex_division): Rename inner_type parameter to type.
5537 Update expand_complex_libcall call-site.
5538 (expand_complex_operations_1): Update expand_complex_multiplication
5539 and expand_complex_division call-sites.
5541 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5544 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5545 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5546 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
5547 the highest significant bit of the shift count mask is clear. In
5548 check whether and[sq]i3 is needed verify that all significant bits
5549 of the shift count other than the highest are set.
5551 2018-05-02 Tom de Vries <tom@codesourcery.com>
5554 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
5555 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
5556 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
5557 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
5558 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
5559 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
5560 __builtin_goacc_parlevel_size.
5562 2018-05-02 Richard Biener <rguenther@suse.de>
5564 PR tree-optimization/85597
5565 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
5566 do not use split vect_get_vec_defs call but call vect_get_slp_defs
5569 2018-05-02 Tom de Vries <tom@codesourcery.com>
5572 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5573 dump files): Add ltrans-tree.
5575 2018-05-02 Tom de Vries <tom@codesourcery.com>
5578 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5579 dump files): Add wpa-ipa.
5581 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
5583 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
5584 powerpc*-*-linux*paired* target.
5585 * config/rs6000/750cl.h: Delete.
5586 * config/rs6000/paired.h: Delete.
5587 * config/rs6000/paired.md: Delete.
5588 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
5590 * config/rs6000/rs6000-builtin.def: Remove paired float support.
5591 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
5592 comment. Remove paired float support.
5593 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
5594 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
5596 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
5597 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
5599 * config/rs6000/rs6000.c: Remove paired float support.
5600 (paired_expand_vector_init, paired_expand_vector_move,
5601 paired_emit_vector_compare, paired_emit_vector_cond_expr,
5602 (paired_expand_lv_builtin, paired_expand_stv_builtin,
5603 paired_expand_builtin, paired_expand_predicate_builtin,
5604 paired_init_builtins): Delete.
5605 * config/rs6000/rs6000.h: Remove paired float support.
5606 * config/rs6000/rs6000.md: Remove paired float support.
5607 (move_from_CR_ov_bit): Delete.
5608 * config/rs6000/rs6000.opt (mpaired): Delete.
5609 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
5610 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
5612 2018-05-02 Richard Biener <rguenther@suse.de>
5615 * gimplify.c (gimplify_save_expr): When in SSA form allow
5616 SAVE_EXPRs to compute to SSA vars.
5618 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5621 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5622 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5623 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
5624 clobber operands[2], instead use a new pseudo. Formatting fixes.
5626 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
5628 PR tree-optimization/85586
5629 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
5630 exit early for statements in the same group if the accesses are
5633 2018-05-02 Tom de Vries <tom@codesourcery.com>
5636 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
5639 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
5641 PR tree-optimization/85143
5642 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
5644 2018-05-01 Tom de Vries <tom@codesourcery.com>
5647 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
5648 not found" error message.
5650 2018-05-01 Tom de Vries <tom@codesourcery.com>
5653 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
5654 * vec.c (test_ordered_remove_if): New function.
5655 (vec_c_tests): Call test_ordered_remove_if.
5656 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
5657 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
5658 * tree-vect-patterns.c (vect_pattern_recog_1): Use
5659 VEC_ORDERED_REMOVE_IF.
5661 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5663 PR tree-optimization/82665
5664 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
5665 pointer subtraction where arguments come from a memchr call.
5667 2018-05-01 Jakub Jelinek <jakub@redhat.com>
5669 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
5670 --push-state --as-needed and --pop-state instead of --as-needed and
5671 --no-as-needed if ld supports it.
5672 * configure: Regenerated.
5675 * doc/install.texi2html: Replace _002d with - and _002a with * in
5676 generated html files using sed.
5678 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5681 * gcc-rich-location.c (blank_line_before_p): New function.
5682 (use_new_line): New function.
5683 (gcc_rich_location::add_fixit_insert_formatted): New function.
5684 * gcc-rich-location.h
5685 (gcc_rich_location::add_fixit_insert_formatted): New function.
5687 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5689 * selftest.c (assert_streq): Rename "expected" and "actual" to
5690 "val1" and "val2". Extend NULL-handling to cover both inputs
5691 symmetrically, while still requiring both to be non-NULL for a pass.
5692 * selftest.h (assert_streq): Rename "expected" and "actual" to
5694 (ASSERT_EQ): Likewise.
5695 (ASSERT_EQ_AT): Likewise.
5696 (ASSERT_KNOWN_EQ): Likewise.
5697 (ASSERT_KNOWN_EQ_AT): Likewise.
5698 (ASSERT_NE): Likewise.
5699 (ASSERT_MAYBE_NE): Likewise.
5700 (ASSERT_MAYBE_NE_AT): Likewise.
5701 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
5702 the assertion to pass.
5703 (ASSERT_STREQ_AT): Likewise.
5705 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5707 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
5708 interaction with -pie.
5710 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5712 * selftest.h: Fix alphabetization of per-source-file selftest
5715 2018-04-30 Jason Merrill <jason@redhat.com>
5717 PR c++/61982 - dead stores to destroyed objects.
5718 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
5721 2018-04-30 Jason Merrill <jason@redhat.com>
5723 * tree.c (build_clobber): New.
5724 * tree.h: Declare it.
5725 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
5727 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5729 * diagnostic-show-locus.c (layout::layout): Update for
5730 location_get_source_line returning a char_span.
5731 (struct char_span): Move to input.h.
5732 (struct correction): Update for fields in char_span becoming
5734 (struct source_line): Update for location_get_source_line
5735 returning a char_span.
5736 (layout::print_line): Likewise.
5737 * edit-context.c (edited_file::print_content): Likewise.
5738 (edited_file::print_diff_hunk): Likewise.
5739 (edited_file::print_run_of_changed_lines): Likewise.
5740 (edited_file::get_num_lines): Likewise.
5741 (edited_line::edited_line): Likewise.
5742 * final.c (asm_show_source): Likewise.
5743 * input.c (location_get_source_line): Convert return type
5744 from const char * to char_span, losing the final "line_len"
5746 (dump_location_info): Update for the above.
5747 (get_substring_ranges_for_loc): Likewise. Use a char_span
5748 when handling the literal within the line.
5749 (test_reading_source_line): Update for location_get_source_line
5750 returning a char_span.
5751 * input.h (class char_span): Move here from
5752 diagnostic-show-locus.c, converting from a struct to a class.
5753 Make data members private.
5754 (char_span::operator bool): New.
5755 (char_span::length): New.
5756 (char_span::get_buffer): New.
5757 (char_span::operator[]): New.
5758 (char_span::subspan): Make const.
5759 (char_span::xstrdup): New.
5760 (location_get_source_line): Convert return type from const char *
5761 to char_span, losing the final "line_size" param.
5763 2018-04-30 Jan Hubicka <jh@suse.cz>
5765 * lto-wrapper.c (ltrans_priorities): New static var.
5766 (cmp_priority): New.
5767 (run_gcc): Read priorities and if doing parallel build order
5768 the Makefile by them.
5770 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5772 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
5774 2018-04-30 Richard Biener <rguenther@suse.de>
5776 * tree-cfg.c (verify_address): Remove base argument, add
5777 flag whether to check TREE_ADDRESSABLE and do that.
5778 (verify_expr): Remove.
5779 (verify_types_in_gimple_reference): Add pieces from verify_expr.
5780 (verify_gimple_assign_single): Likewise.
5781 (verify_gimple_switch): Likewise.
5782 (verify_expr_location_1): Dereference tp once. Add (disabled)
5783 piece from verify_expr.
5784 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
5786 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5788 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
5790 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5792 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
5793 (small_data_pattern): Likewise.
5794 (arc_rewrite_small_data): Likewise.
5795 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
5796 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
5797 (get_symbol_alignment): New function.
5798 (legitimate_small_data_address_p): Likewise.
5799 (legitimate_scaled_address): Update, call
5800 legitimate_small_data_address_p.
5801 (output_sdata): New static variable.
5802 (arc_print_operand): Update how we handle small data operands.
5803 (arc_print_operand_address): Likewise.
5804 (arc_legitimate_address_p): Update, use
5805 legitimate_small_data_address_p.
5806 (arc_rewrite_small_data_p): Remove.
5807 (arc_rewrite_small_data_1): Likewise.
5808 (arc_rewrite_small_data): Likewise.
5809 (small_data_pattern): Likewise.
5810 (compact_sda_memory_operand): Update to use
5811 legitimate_small_data_address_p and get_symbol_alignment.
5812 (prepare_move_operands): Don't rewite sdata pattern.
5813 (prepare_extend_operands): Remove.
5814 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
5816 (zero_extendqisi2): Likewise.
5817 (zero_extendhisi2): Likewise.
5818 (extendqihi2): Likewise.
5819 (extendqisi2): Likewise.
5820 (extendhisi2): Likewise.
5824 * config/arc/constraints.md (Usd): Change it to memory constraint.
5826 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5828 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
5829 as source of std instructions.
5830 * config/arc/arc.md (movsi_insn): Update pattern predicate to
5831 allow 6-bit constants as source for store instructions.
5832 (movdi_insn): Update instruction pattern to allow 6-bit constants
5833 as source for store instructions.
5835 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5837 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
5839 2018-04-30 Nathan Sidwell <nathan@acm.org>
5840 Sandra Loosemore <sandra@codesourcery.com>
5842 * dumpfile.c (dump_open): Allow '-' for stdout.
5843 * doc/invoke.texi (Developer Options): Document dump filename
5844 determination early. Document stdin/stdout selection.
5846 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
5848 Microblaze Target: PIC data text relative
5850 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
5851 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
5853 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
5854 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
5855 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
5856 New addressing mode for data-text relative position indepenedent code.
5857 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
5858 'ADDRESS_SYMBOLIC_TXT_REL'.
5859 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
5860 (microblaze_legitimate_pic_operand): Exclude function calls from
5861 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
5862 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
5864 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5865 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5866 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
5867 for 'address + offset'.
5868 (microblaze_expand_prologue): Add new function prologue call for
5870 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
5871 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
5872 table in case of TARGET_PIC_DATA_TEXT_REL.
5873 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
5874 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
5875 Add new macros 'UNSPEC_TEXT',
5876 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
5877 + exclude function calls from 'UNSPEC_PLT' in case of data text
5879 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5880 new target hook for generating address diff vector tables in case of
5882 * doc/tm.texi : Regenerate.
5883 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
5884 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
5885 of addr diff vector generation.
5886 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5887 target hook definition.
5888 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
5889 Add default function for generate_pic_addr_diff_vec -> flag_pic.
5890 * doc/invoke.texi (Add new pic option): Add new microblaze pic
5891 option for data text relative.
5893 2018-04-30 Richard Biener <rguenther@suse.de>
5895 * tree-chrec.h (evolution_function_is_constant_p): Remove
5897 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
5899 2018-04-30 Richard Biener <rguenther@suse.de>
5902 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
5904 2018-04-30 Richard Biener <rguenther@suse.de>
5906 PR tree-optimization/28364
5907 PR tree-optimization/85275
5908 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
5909 copying first exit test.
5911 2018-04-28 Mark Wielaard <mark@klomp.org>
5913 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
5915 (dwarf_AT): Handle DW_AT_addr_base.
5916 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
5918 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
5921 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
5922 (*ashl<dwi>3_doubleword_mask_1): Ditto.
5923 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
5924 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
5926 2018-04-28 Richard Biener <rguenther@suse.de>
5928 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
5929 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
5930 to reflect use. Only add interesting stmts.
5932 2018-04-27 Martin Jambor <mjambor@suse.cz>
5935 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
5936 the jump function allows for passing through aggregate values.
5938 2018-04-27 David Malcolm <dmalcolm@redhat.com>
5940 * input.h (in_system_header_at): Convert from macro to inline
5942 (from_macro_expansion_at): Likewise.
5943 (from_macro_definition_at): Likewise.
5945 2018-04-27 Jeff Law <law@redhat.com>
5947 * config.gcc: Mark tile* targets as deprecated/obsolete.
5949 2018-04-27 Richard Biener <rguenther@suse.de>
5951 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
5954 2018-04-27 Richard Biener <rguenther@suse.de>
5956 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
5958 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
5960 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
5961 with Yd constraint. Set "preferred_for_speed" attribute from
5962 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
5964 (*movdi_internal): Ditto.
5965 (movti_interunit splitters): Remove
5966 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
5967 (movdi_interunit splitters): Ditto.
5968 * config/i386/constraints.md (Ye): Remove.
5969 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
5971 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5974 * config/aarch64/constraints.md (Usg): Limit to 31.
5977 2018-04-27 Jakub Jelinek <jakub@redhat.com>
5979 PR tree-optimization/85529
5980 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
5981 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
5982 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
5983 zero extension or masking of the MSB bit.
5984 (optimize_range_tests): Add FIRST_BB argument, pass it through
5985 to optimize_range_tests_var_bound.
5986 (maybe_optimize_range_tests, reassociate_bb): Adjust
5987 optimize_range_tests callers.
5989 2018-04-26 Richard Biener <rguenther@suse.de>
5990 Jakub Jelinek <jakub@redhat.com>
5992 * cgraph.h (symbol_table): Just declare debug method here.
5993 * symtab.c (symbol_table::debug): Define.
5995 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
5997 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
5999 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
6001 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
6002 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
6003 (*movdi_internal): Substitute Yi and Yj constraint with x
6004 and Ym and Yn constraint with y constraint. Update "isa"
6005 attribute and set "preferred_for_speed" attribute from
6006 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
6007 (*movsi_internal): Ditto.
6008 (*movdf_internal): Ditto.
6009 (*movsf_internal): Ditto.
6010 (*zero_extendsidi2): Ditto.
6011 * config/i386/sse.md (vec_set<mode>_0): Ditto.
6012 (sse2_loadld): Ditto.
6013 (*vec_extract<ssevecmodelower>_0): Ditto.
6014 (*vec_extractv4si_0_zext_sse4): Ditto.
6015 (vec_concatv2di): Ditto.
6016 (*vec_dup<mode>): Ditto.
6017 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
6018 * config/i386/constraints.md (Yi): Remove.
6023 2018-04-26 Nathan Sidwell <nathan@acm.org>
6025 * dumpfile.c (dump_open): New.
6026 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
6027 (dump_finish): Detect stdio/stderr by value not name.
6029 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
6031 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
6033 2018-04-26 Tom de Vries <tom@codesourcery.com>
6036 * config/nvptx/nvptx.c (verify_neutering_jumps)
6037 (verify_neutering_labels): New function
6038 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
6040 2018-04-26 Tom de Vries <tom@codesourcery.com>
6043 * config/nvptx/nvptx.c (needs_neutering_p): New function.
6044 (nvptx_single): Use needs_neutering_p to skip over insns that do not
6047 2018-04-26 Richard Biener <rguenther@suse.de>
6048 Tom de Vries <tom@codesourcery.com>
6051 * lto-streamer-out.c (output_function): Fixup loops if required to match
6052 discovery done in the reader.
6054 2018-04-26 Richard Biener <rguenther@suse.de>
6056 PR tree-optimization/85116
6057 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
6058 have a loop exit from the single latch predecessor. Remove
6059 case of header with just condition.
6060 (ch_base::copy_headers): Exclude infinite loops from any
6062 (pass_ch::execute): Record exits.
6064 2018-04-26 Richard Biener <rguenther@suse.de>
6066 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
6067 prologue cost vector and pass it to vect_get_load_cost.
6068 (vect_get_peeling_costs_all_drs): Likewise.
6069 (vect_peeling_hash_get_lowest_cost): Likewise.
6070 (vect_enhance_data_refs_alignment): Likewise.
6072 2018-04-26 Richard Biener <rguenther@suse.de>
6075 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
6076 checking of integer<->pointer conversions.
6077 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
6078 sign-/zero-extending pointer types.
6079 (expand_omp_for_static_chunk): Likewise.
6081 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
6082 Jean Lee <xiaoyur347@gmail.com>
6084 * config/mips/mips.c (mips_asan_shadow_offset): New function.
6085 (TARGET_ASAN_SHADOW_OFFSET): Define.
6086 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
6087 true for -fsanitize=address.
6089 2018-04-25 Mark Wielaard <mark@klomp.org>
6091 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
6094 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6096 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
6097 than "alu", remove explicit "memory" and "imm_disp" attributes.
6098 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
6101 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
6102 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
6103 gen_lowpart_no_emit.
6105 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
6108 * config/i386/i386.c (ix86_expand_builtin): Change memory
6109 operand to XI, extend p0 to Pmode.
6110 * config/i386/i386.md: Change unspec volatile and operand
6111 1 mode to XI, change operand 0 mode to P.
6113 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6115 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
6116 GET_MODE_MASK before any checking.
6117 (nds32_can_use_bset_p): Likewise.
6118 (nds32_can_use_btgl_p): Likewise.
6120 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6122 * config/nds32/nds32-doubleword.md: New define_split pattern for
6123 illegal register number.
6125 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6127 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
6129 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6131 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
6133 2018-04-25 Richard Biener <rguenther@suse.de>
6135 * lto-streamer.h (LTO_major_version): Bump to 8.
6137 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6139 * BASE-VER: Set to 9.0.0.
6141 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
6143 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
6144 in __abskf2 and __powikf2.
6146 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6149 * config/aarch64/constraints.md (Usg, Usj): New constraints.
6150 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
6151 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
6152 Use the above on operand 2. Reindent.
6153 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
6155 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6158 * common/config/i386/i386-common.c (ix86_handle_option): Don't
6160 * config/i386/i386.opt (mcet): Removed.
6161 * doc/install.texi: Remove -mcet documentation.
6162 * doc/invoke.texi: Likewise.
6164 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6167 * doc/install.texi: Remove -mcet from bootstrap-cet.
6169 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6172 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
6173 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
6177 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
6178 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
6179 containing a CONST_VECTOR.
6181 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
6183 * doc/install.texi: Update newlib dependency for nvptx.
6185 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6188 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
6189 instead of INTVAL when shifting x left.
6191 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
6193 PR tree-optimization/85478
6194 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
6195 vect_grouped_store_supported for single element vectors.
6197 2018-04-24 Richard Biener <rguenther@suse.de>
6200 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
6201 load cost increase to the case of non-constant step.
6203 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6206 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
6207 destination if any_malformed_asm.
6209 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
6212 * expr.c (store_field): In the bitfield case, if the value comes from
6213 a function call and is returned in registers by means of a PARALLEL,
6214 do not change the mode of the temporary unless BLKmode and VOIDmode.
6216 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
6218 PR rtl-optimization/85423
6219 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
6220 dependencies to debug insns when the previous insn is non-debug.
6222 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6224 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
6225 enums into a single definition.
6226 (fls): Fix predicates and printing.
6229 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6231 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
6232 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
6233 and short u6 immediate.
6234 (check_if_valid_sleep_operand): Remove.
6235 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
6237 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6239 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
6240 flag_always_save_lp condition.
6241 * config/nds32/nds32.opt (malways-save-lp): New option.
6243 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6245 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
6246 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
6247 * config/nds32/nds32.h
6248 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
6249 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
6251 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6253 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
6254 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
6256 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6257 Chung-Ju Wu <jasonwucj@gmail.com>
6259 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
6261 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
6262 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
6264 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6266 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
6268 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6270 * config/nds32/nds32-protos.h (nds32_data_alignment,
6271 nds32_local_alignment): Declare.
6272 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
6273 nds32_local_alignment): New functions.
6274 (TARGET_CONSTANT_ALIGNMENT): Define.
6275 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
6277 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6279 * config/nds32/nds32.c
6280 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
6281 (TARGET_MODES_TIEABLE_P): Likewise.
6283 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6285 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
6288 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
6289 Chung-Ju Wu <jasonwucj@gmail.com>
6291 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6292 for unaligned access.
6293 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6295 * config/nds32/nds32-intrinsic.md: Likewise.
6296 * config/nds32/nds32_intrinsic.h: Likewise.
6297 * config/nds32/nds32.h (nds32_builtins): Likewise.
6298 * config/nds32/nds32.opt (munaligned-access): New option.
6299 * config/nds32/nds32.c (nds32_asm_file_start): Display
6300 flag_unaligned_access status.
6302 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6304 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
6305 -mno-relax is present.
6306 * config/riscv/linux.h (LINK_SPEC): Ditto.
6308 2018-04-20 Martin Sebor <msebor@redhat.com>
6311 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
6313 (gimple_fold_builtin_stxcpy_chk): Same.
6314 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
6316 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
6319 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
6320 __powikf2 when long double is IEEE 128-bit.
6322 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6324 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
6325 step to make sure stack always aligned.
6327 2018-04-20 Carl Love <cel@us.ibm.com>
6330 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
6331 size check for arg0.
6333 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
6334 Tom de Vries <tom@codesourcery.com>
6337 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
6338 Emit insns for calls too.
6339 (nvptx_find_par): Always look for worker-level predecessor insn.
6340 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
6342 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
6343 (nvptx_process_pars): Propagate frames for calls.
6345 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
6348 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
6350 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
6351 (ix86_handle_option): Don't handle OPT_mibt.
6352 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
6354 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
6356 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
6357 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
6358 (ix86_target_macros): Define __CET__ with flag_cf_protection
6359 for -fcf-protection.
6360 * config/i386/i386.c (isa2_opts): Remove -mibt.
6361 * config/i386/i386.h (TARGET_IBT): Removed.
6362 (TARGET_IBT_P): Likewise.
6363 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
6364 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
6365 * config/i386/i386.opt (mcet): Update help message.
6368 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
6369 -mcet as an alias for -mshstk.
6371 2018-04-20 Richard Biener <rguenther@suse.de>
6374 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
6375 complexity by forcing a single use of the multiply operand.
6377 2018-04-20 Martin Jambor <mjambor@suse.cz>
6380 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
6381 recursion dependency to only apply to non-clones.
6383 2018-04-20 Martin Jambor <mjambor@suse.cz>
6386 * ipa-cp.c (create_specialized_node): Check that clones of
6387 self-recursive edges exist during IPA-CP.
6389 2018-04-19 Toon Moene <toon@moene.org>
6391 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
6394 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6396 PR tree-optimization/85467
6397 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
6398 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
6399 VECTOR_CST element to type.
6401 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6404 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
6405 * config/i386/i386.md (builtin_setjmp_setup): Removed.
6406 (builtin_longjmp): Likewise.
6407 (save_stack_nonlocal): New pattern.
6408 (restore_stack_nonlocal): Likewise.
6410 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6413 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6414 Replace ASM_OUTPUT_LABEL with fprintf.
6416 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6419 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6420 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
6421 * config/i386/i386-c.c (ix86_target_macros_internal): Also
6422 define __IBT__ and __SHSTK__ for -fcf-protection.
6423 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
6425 (ix86_trampoline_init): Likewise.
6426 (x86_output_mi_thunk): Likewise.
6427 (ix86_notrack_prefixed_insn_p): Likewise.
6428 (ix86_option_override_internal): Don't disallow -fcf-protection.
6429 * config/i386/i386.md (rdssp<mode>): Also enable for
6431 (incssp<mode>): Likewise.
6432 (nop_endbr): Likewise.
6433 * config/i386/i386.opt (mcet): Change help message to built-in
6437 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
6438 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
6439 enable CET built-in functions.
6441 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
6443 * common/config/i386/i386-common.c
6444 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
6445 OPTION_MASK_ISA_MOVDIRI_UNSET,
6446 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
6447 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
6448 * config.gcc (movdirintrin.h): New header.
6449 * config/i386/cpuid.h (bit_MOVDIRI,
6450 bit_MOVDIR64B): New bits.
6451 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
6453 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
6454 (VOID, PVOID, PCVOID)): New function types.
6455 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
6456 __builtin_ia32_directstoreu_u64,
6457 __builtin_ia32_movdir64b): New builtins.
6458 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
6459 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
6461 (ix86_valid_target_attribute_inner_p): Ditto.
6462 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
6463 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
6464 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
6465 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
6466 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
6467 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
6468 (movdiri<mode>, movdir64b_<mode>): New.
6469 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
6470 * config/i386/immintrin.h: Include movdirintrin.h.
6471 * config/i386/movdirintrin.h: New file.
6472 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
6474 2018-04-19 Richard Biener <rguenther@suse.de>
6477 * cfg.c (clear_bb_flags): When loop state says we have
6478 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
6480 2018-04-19 Richard Biener <rguenther@suse.de>
6482 PR tree-optimization/84737
6483 * tree-vect-data-refs.c (vect_copy_ref_info): New function
6484 copying restrict info.
6485 (vect_setup_realignment): Use it.
6486 * tree-vectorizer.h (vect_copy_ref_info): Declare.
6487 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
6488 the first DR to all generated stores.
6489 (vectorizable_load): Likewise for loads.
6491 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6493 PR tree-optimization/85446
6494 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
6495 the integral and pointer types to have the same precision.
6497 * doc/install.texi: Document --disable-cet being the default and
6500 2018-04-18 Martin Liska <mliska@suse.cz>
6502 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
6505 2018-04-18 Martin Liska <mliska@suse.cz>
6508 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6511 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6512 arguments if they are comparable.
6514 2018-04-18 Martin Liska <mliska@suse.cz>
6517 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6520 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6523 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
6526 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
6527 ENDBR after calling __morestack.
6529 2018-04-18 David Malcolm <dmalcolm@redhat.com>
6532 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
6533 by using gcc_base_ver to generate a gcc_driver_version, and use
6534 it when generating GCC_DRIVER_NAME.
6535 * configure: Regenerate.
6537 2018-04-18 Jakub Jelinek <jakub@redhat.com>
6540 * config.gcc: Obsolete powerpc*-*-*spe*.
6542 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6545 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
6546 (stabstr_D): Change type of unum from unsigned int to
6547 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
6550 2018-04-17 Jim Wilson <jimw@sifive.com>
6553 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
6554 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
6555 Set arg_pointer_offset after using pretend_args_size.
6557 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6559 PR rtl-optimization/85431
6560 * dse.c (record_store): Ignore zero width stores.
6563 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
6564 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
6565 __builtin_stack_restore rather than after it.
6566 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
6567 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
6568 argument instead of virtual_dynamic_stack_rtx.
6570 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
6572 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
6574 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
6575 Add note to error message to explain internal mapping of overloaded
6576 built-in function name to non-overloaded built-in function name.
6577 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
6580 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6583 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
6584 where the inputs overlap with the output.
6586 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6588 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
6589 (=v, v) alternative and explicit "memory" attribute.
6590 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
6591 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6593 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
6594 "sselog1" type instead of "sselog".
6595 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
6596 "sselog". Remove explicit "memory" attribute.
6597 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
6598 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6600 (vec_extract_hi_v32hi): Merge all alternatives into one, use
6601 "sselog1" type instead of "sselog". Remove explicit "memory"
6603 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
6604 use "sselog1" type instead of "sselog". Remove explicit "memory"
6606 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
6607 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6609 (vec_extract_hi_v64qi): Merge all alternatives into one, use
6610 "sselog1" type instead of "sselog". Remove explicit "memory"
6612 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
6613 use "sselog1" type instead of "sselog". Remove explicit "memory"
6617 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
6620 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
6623 2018-04-17 Martin Jambor <mjambor@suse.cz>
6626 * ipa-cp.c (create_specialized_node): Call
6627 expand_all_artificial_thunks if necessary.
6629 2018-04-17 Martin Liska <mliska@suse.cz>
6632 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
6633 in message, remote space in between '_G' and '('.
6635 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6638 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
6639 avx512f_vmcmp<mode>3<round_saeonly_name>,
6640 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
6641 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
6642 avx512f_rndscale<mode><round_saeonly_name>,
6643 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
6644 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
6645 Use %<iptr>2 instead of %2 for -masm=intel.
6646 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
6647 avx512f_vcvttss2usi<round_saeonly_name>,
6648 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
6650 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
6651 avx512f_vcvttsd2usi<round_saeonly_name>,
6652 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
6653 Use %q1 instead of %1 for -masm=intel.
6654 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
6655 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
6656 of %3 for -masm=intel.
6657 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
6659 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
6661 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
6663 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
6664 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
6666 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
6668 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
6669 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
6670 %g1 and one with %0 and %1.
6671 (avx512er_vmrcp28<mode><round_saeonly_name>,
6672 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
6674 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
6675 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
6676 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
6677 of %0 and %{%4%} for -masm=intel.
6678 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
6679 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
6680 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
6681 order of %0 and %{%5%}%{z%} for -masm=intel.
6683 2018-04-17 Jan Hubicka <jh@suse.cz>
6686 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
6688 2018-04-17 Martin Liska <mliska@suse.cz>
6691 * multiple_target.c (create_dispatcher_calls): Set apostrophes
6692 for target_clone error message. Make default implementation
6693 clone to be a local declaration.
6694 (separate_attrs): Add new argument and check for an empty
6696 (expand_target_clones): Handle it.
6697 (ipa_target_clone): Make redirection just for target_clones
6700 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
6701 Tom de Vries <tom@codesourcery.com>
6704 * omp-expand.c (expand_oacc_for): Add dummy false branch for
6705 tiled basic blocks without omp continue statements.
6707 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
6710 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
6711 vec_extract expression as having side effects to make sure it gets
6714 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
6717 * config/i386/i386.c (get_builtin_code_for_version): Check
6720 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
6723 * gcc/config.gcc: Support "skylake".
6724 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
6726 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
6727 (processor_target_table): Add "skylake".
6728 (ix86_option_override_internal): Add "skylake".
6729 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
6730 PROCESSOR_CANNONLAKE.
6731 (get_builtin_code_for_version): Fix priority for
6732 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
6733 PROCESSOR_SKYLAKE-AVX512.
6734 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
6735 (processor_type): Add PROCESSOR_SKYLAKE.
6737 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
6738 Jason Merrill <jason@redhat.com>
6741 * convert.c (convert_to_integer_1): Use direct recursion for
6742 enumeral types and types with a precision less than the number
6743 of bits in their mode.
6745 2018-04-16 Julia Koval <julia.koval@intel.com>
6748 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
6749 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
6751 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
6754 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
6755 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
6756 and -mno-direct-move.
6758 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
6761 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
6762 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
6763 Ensure negative shifts result in {0}.
6765 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
6767 PR rtl-optimization/79916
6768 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
6769 regs (if any) to define how to gnerate SD moves when LRA is in
6772 2018-04-13 Jakub Jelinek <jakub@redhat.com>
6774 PR rtl-optimization/85393
6775 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
6776 * except.c (expand_dw2_landing_pad_for_region): Make static.
6777 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
6778 a label and unconditional jump to old_bb, rather than
6779 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
6782 PR rtl-optimization/85376
6783 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
6784 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
6785 instead of a specific value.
6787 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6788 Bin Cheng <bin.cheng@arm.com>
6790 PR tree-optimization/82965
6791 PR tree-optimization/83991
6792 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
6793 by_profile_only parameter.
6794 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
6795 information if the loop was predicted to iterate too many times.
6796 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
6798 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6801 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
6804 2018-04-13 Martin Liska <mliska@suse.cz>
6805 Jakub Jelinek <jakub@redhat.com>
6808 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
6809 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
6810 * builtins.c (expand_builtin_memory_copy_args): Use
6811 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
6812 handle dest_addr == pc_rtx.
6814 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
6817 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
6818 asked to not generate direct moves.
6819 (fix_trunc<mode>si2_stfiwx): Similar.
6820 (fix_trunc<mode>si2_internal): Similar.
6822 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6825 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
6826 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
6827 lookup if dest in some wider mode is known to be const0_rtx and
6828 if so, record permanent equivalence for it to be ZERO_EXTEND of
6829 the narrower mode destination.
6831 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6833 * lto-streamer-out.c (output_function): Revert 259346.
6834 * omp-expand.c (expand_oacc_for): Likewise.
6836 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
6838 PR rtl-optimization/85354
6839 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
6840 * sel-sched.c (sel_global_init): ... here.
6842 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
6845 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
6846 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6847 mode for PE-COFF targets.
6848 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
6849 (i386_pe_asm_lto_end): Likewise.
6850 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
6851 (TARGET_ASM_LTO_END): Likewise.
6852 * config/i386/winnt.c (saved_debug_info_level): New static variable.
6853 (i386_pe_asm_lto_start): New function.
6854 (i386_pe_asm_lto_end): Likewise.
6856 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6857 Richard Biener <rguenther@suse.de>
6860 * lto-streamer-out.c (output_function): Fix CFG loop state before
6862 * omp-expand.c (expand_oacc_for): Handle calls to internal
6863 functions like regular functions.
6865 2018-04-12 Richard Biener <rguenther@suse.de>
6868 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
6869 for the early LTO debug to properly generate references to it
6870 during DIE emission. Do not re-use that for the skeleton for
6872 (dwarf2out_early_finish): Likewise.
6874 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6877 * config/i386/sse.md
6878 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
6879 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
6880 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
6881 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
6882 and output is a reg, avoid creating invalid lowpart subreg, but
6883 instead split into a 512-bit move. Don't split if not AVX512VL,
6884 input is xmm16+ reg and output is a mem.
6885 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
6886 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
6887 xmm16+ reg and output is a mem.
6889 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6891 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
6892 also for flag_dwarf2_cfi_asm.
6894 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6896 PR rtl-optimization/85342
6897 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
6898 a bool scalar var inside of the loop instead. Don't try to update
6899 recog_data.operand after failed apply_change_group.
6901 2018-04-12 Tom de Vries <tom@codesourcery.com>
6904 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
6905 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
6906 array with flexible array member as array without given dimension.
6907 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
6908 argument for undefined param to true.
6910 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
6913 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6914 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
6915 from PowerPC section.
6916 * config/rs6000/sysv4.opt (mcall-): Improve help text.
6917 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
6918 help text that is too long.
6919 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
6920 help text that is too long.
6921 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
6922 help text that is too long.
6924 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
6926 * config/alpha/alpha.md (stack_probe_internal): Rename
6927 from "probe_stack". Update all callers.
6929 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6931 PR rtl-optimization/84566
6932 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
6933 sched_macro_fuse_insns.
6935 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6938 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
6939 (compute_block_dependences): ... from here.
6941 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6943 PR tree-optimization/85331
6944 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
6945 from int to HOST_WIDE_INT.
6947 2018-04-11 Martin Jambor <mjambor@suse.cz>
6950 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
6951 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
6952 not the same as the source val.
6953 (cgraph_edge_brings_value_p): New parameter.
6954 (gather_edges_for_value): Pass destination value to
6955 cgraph_edge_brings_value_p.
6956 (perhaps_add_new_callers): Likewise.
6957 (get_info_about_necessary_edges): Likewise and exclude values brought
6958 only by self-recursive edges.
6959 (create_specialized_node): Redirect only clones of self-calling edges.
6960 (+self_recursive_pass_through_p): New function.
6961 (find_more_scalar_values_for_callers_subset): Use it.
6962 (find_aggregate_values_for_callers_subset): Likewise.
6963 (known_aggs_to_agg_replacement_list): Removed.
6964 (decide_whether_version_node): Re-calculate known constants for all
6965 remaining context clones.
6967 2018-04-11 Richard Biener <rguenther@suse.de>
6970 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
6971 from early DWARF output.
6972 (dwarf2out_early_finish): Output line info unconditionally into
6973 early DWARF and add reference to it.
6975 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6978 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
6979 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
6980 other than V2DFmode using iptr mode attribute.
6981 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
6983 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6985 PR rtl-optimization/84659
6986 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
6988 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6991 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
6993 (output_loc_list): Pass address of a dummy size variable even in the
6994 locview handling loop.
6995 (index_location_lists): Add comment on why skip_loc_list_entry can't
6998 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
7001 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
7004 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
7007 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7008 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
7009 and -mstring-compare-inline-limit.
7011 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7014 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
7015 for stack clash protection in a register whenever we need it to be in
7018 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7020 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
7021 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
7023 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7026 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
7028 (mlong-double-): Ditto.
7029 * config/rs6000/sysv4.opt (msdata=): Ditto.
7030 (mtls-size=): Ditto.
7032 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7034 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
7035 erroneous entries for
7036 "vector int vec_ldl (int, long int *)", and
7037 "vector unsigned int vec_ldl (int, unsigned long int *)".
7038 Add comments and entries for
7039 "vector bool char vec_ldl (int, bool char *)",
7040 "vector bool short vec_ldl (int, bool short *)",
7041 "vector bool int vec_ldl (int, bool int *)",
7042 "vector bool long long vec_ldl (int, bool long long *)",
7043 "vector pixel vec_ldl (int, pixel *)",
7044 "vector long long vec_ldl (int, long long *)",
7045 "vector unsigned long long vec_ldl (int, unsigned long long *)".
7046 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
7047 type tree bool_long_long_type_node and correct definition of
7048 bool_V2DI_type_node to make reference to this new type tree.
7049 (rs6000_mangle_type): Replace erroneous reference to
7050 bool_long_type_node with bool_long_long_type_node.
7051 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
7052 comments to emphasize sign distinctions for char and int types and
7053 replace RS6000_BTI_bool_long constant with
7054 RS6000_BTI_bool_long_long constant. Also add comment to restrict
7055 use of RS6000_BTI_pixel.
7056 (bool_long_type_node): Remove this macro definition.
7057 (bool_long_long_type_node): New macro definition
7059 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7061 PR rtl-optimization/85300
7062 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
7063 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
7064 simplify_unary_operation fails.
7066 2018-04-10 Martin Liska <mliska@suse.cz>
7068 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
7069 cgraph_edge and ipa_ref.
7071 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7075 * config/i386/sse.md
7076 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
7077 computation of the VEC_MERGE selector from mask.
7078 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
7079 Fix decoding of the VEC_MERGE selector into mask.
7081 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7083 PR tree-optimization/85286
7084 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
7086 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7088 * final.c (final_1): Set insn_last_address as well as
7089 insn_current_address.
7091 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7094 * explow.c (emit_stack_probe): Call validize_mem on memory location
7095 before passing it to gen_probe_stack. Create address operand and
7096 legitimize it for the probe_stack_address case.
7098 2018-04-09 Jan Hubicka <jh@suse.cz>
7101 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
7102 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
7103 * tree.c (free_lang_data_in_type): Fix handling of binfos;
7105 (free_lang_data): Rebuild type inheritance graph.
7107 2018-04-09 Martin Sebor <msebor@redhat.com>
7109 * invoke.texi (-finline-small-functions): Mention other optimization
7111 (-findirect-inlining, -fpartial-inlining): Same.
7112 (-finline-functions-called-once): Same.
7113 (-freorder-blocks-and-partition): Same.
7115 2018-04-09 Jan Hubicka <jh@suse.cz>
7118 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
7119 jumps; choose last target that matches the criteria (i.e.
7120 no partition changes for non-crossing jumps).
7121 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
7122 support for redirecting crossing jumps to non-crossing.
7124 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
7126 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
7127 also for naked functions.
7129 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
7131 * config/arc/arc.md (add_shift): New pattern.
7132 (add_shift2): Likewise.
7133 (sub_shift): Likewise.
7134 (sub_shift_cmp0_noout): Likewise.
7135 (compare_si_ashiftsi): Likewise.
7136 (xbfu_cmp0_noout): New combine pattern.
7137 (xbfu_cmp0"): Likewise.
7138 (movsi_set_cc_insn): Place the predicable variant first.
7139 (commutative_binary_cmp0_noout): Remove clobber.
7140 (commutative_binary_cmp0): New pattern.
7141 (noncommutative_binary_cmp0): Likewise.
7142 (noncommutative_binary_cmp0_noout): Likewise.
7143 (noncommutative_binary_comparison_result_used): Removed.
7144 (rsub_cmp0): New pattern.
7145 (rsub_cmp0_noout): Likewise.
7146 (extzvsi): Changed, keep only meaningful variants.
7147 (SQH, SEZ): New iterators.
7148 (SQH_postfix): New mode attribute.
7149 (SEZ_prefix): New code attribute.
7150 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
7151 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
7152 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
7154 (noncommutative_operator): Check the availability of barrel
7157 2018-04-09 Richard Biener <rguenther@suse.de>
7159 PR tree-optimization/85284
7160 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
7161 Only use the niter constraining form of simple_iv when the exit
7164 2018-04-09 Tom de Vries <tom@codesourcery.com>
7167 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7168 (define_expand "*memory_barrier"): New define_expand.
7169 (define_insn "memory_barrier"): New insn.
7171 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7173 PR rtl-optimization/80463
7174 PR rtl-optimization/83972
7175 PR rtl-optimization/83480
7177 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
7178 correct producer for the insn.
7179 (tidy_control_flow): Fixup seqnos in case of debug insns.
7181 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7183 PR rtl-optimization/83913
7185 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
7186 different sched-times when merging exprs.
7188 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7190 PR rtl-optimization/83962
7192 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
7193 tidy_fallthru_edge and tidy_control_flow.
7195 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7197 PR rtl-optimization/83530
7199 * sel-sched.c (force_next_insn): New global variable.
7200 (remove_insn_for_debug): When force_next_insn is true, also leave only
7201 next insn in the ready list.
7202 (sel_sched_region): When the region wasn't scheduled, make another pass
7203 over it with force_next_insn set to 1.
7205 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7207 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
7209 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7210 for interrupt control.
7211 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7212 functions for interrupt control.
7213 * config/nds32/nds32-intrinsic.md: Likewise.
7214 * config/nds32/nds32_intrinsic.h: Likewise.
7215 * config/nds32/nds32.h (nds32_builtins): Likewise.
7217 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7219 * config/nds32/nds32.c (nds32_init_machine_status,
7220 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
7221 strict_aligned_p field.
7222 (nds32_expand_to_rtl_hook): New function.
7223 (TARGET_EXPAND_TO_RTL_HOOK): Define.
7224 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
7226 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7227 Chung-Ju Wu <jasonwucj@gmail.com>
7229 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
7230 * config/nds32/nds32-n7.md: New file.
7231 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
7232 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
7234 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
7235 * config/nds32/nds32.md (pipeline_model): Add n7.
7236 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
7237 * config/nds32/pipelines.md: Include n7 settings.
7239 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7240 Chung-Ju Wu <jasonwucj@gmail.com>
7242 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
7243 * config/nds32/nds32-e8.md: New file.
7244 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
7245 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
7247 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
7248 * config/nds32/nds32.md (pipeline_model): Add e8.
7249 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
7250 * config/nds32/pipelines.md: Include e8 settings.
7252 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7253 Chung-Ju Wu <jasonwucj@gmail.com>
7255 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
7256 * config/nds32/nds32-n8.md: New file.
7257 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
7258 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
7260 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
7261 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
7262 * config/nds32/nds32.md (pipeline_model): Add n8.
7263 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
7264 * config/nds32/pipelines.md: Include n8 settings.
7266 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7267 Chung-Ju Wu <jasonwucj@gmail.com>
7269 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
7270 * config/nds32/nds32-n9-2r1w.md: New file.
7271 * config/nds32/nds32-n9-3r2w.md: New file.
7272 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
7273 nds32_register_ports): New or modify for cpu n9.
7274 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
7276 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
7277 * config/nds32/nds32-utils.c: New file.
7278 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
7279 TARGET_MUL_SLOW): Define.
7280 * config/nds32/nds32.md (pipeline_model): New attribute.
7281 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
7282 New options that support cpu n9.
7283 * config/nds32/pipelines.md: Include n9 settings.
7284 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
7286 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7288 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
7289 information if necessary.
7290 (output_cond_branch_compare_zero): Likewise.
7291 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
7292 (nds32_target_alignment): Refine for alignment.
7293 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
7294 (FUNCTION_BOUNDARY): Modify.
7295 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
7297 * config/nds32/nds32.opt (malways-align, malign-functions): New.
7299 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7301 * config/nds32/constants.md (unspec_volatile_element): Add values for
7302 TLB operation and data prefetch.
7303 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7304 functions for TLB operation and data prefetch.
7305 * config/nds32/nds32-intrinsic.md: Likewise.
7306 * config/nds32/nds32_intrinsic.h: Likewise.
7307 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
7308 (nds32_print_operand): Likewise.
7309 * config/nds32/nds32.h (nds32_builtins): Likewise.
7311 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
7312 Andrew Pinski <pinsika@gcc.gnu.org>
7315 * match.pd: Use constant_boolean_node of correct type instead of
7316 boolean_true_node or boolean_false_node for simplifying
7317 pointer comparisons to zero.
7319 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7321 PR tree-optimization/80021
7322 * tree.c (verify_type_variant): Make error call in verify_variant_match
7323 translatable and remove final full stop.
7325 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7327 * config/nds32/constants.md (unspec_volatile_element): Add
7328 UNSPEC_VOLATILE_EH_RETURN.
7329 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
7330 nds32_output_stack_pop): Support dwarf exception handling process.
7331 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
7332 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
7333 exception handling process.
7334 (nds32_compute_stack_frame): Likewise.
7335 (nds32_return_addr_rtx): Likewise.
7336 (nds32_initial_elimination_offset): Likewise.
7337 (nds32_expand_prologue): Likewise.
7338 (nds32_expand_epilogue): Likewise.
7339 (nds32_dynamic_chain_address): New function.
7340 * config/nds32/nds32.h (machine_function): Add fields for dwarf
7342 (DYNAMIC_CHAIN_ADDRESS): Define.
7343 (EH_RETURN_DATA_REGNO): Define.
7344 (EH_RETURN_STACKADJ_RTX): Define.
7345 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
7346 patterns for dwarf exception handling.
7348 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7350 * config/nds32/nds32.h: Clean up obsolete macros.
7352 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7354 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7355 Add enum values for particular instructions.
7356 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7357 particular intrinsic functions.
7358 * config/nds32/nds32-intrinsic.md: Likewise.
7359 * config/nds32/nds32_intrinsic.h: Likewise.
7360 * config/nds32/nds32.h (nds32_builtins): Likewise.
7361 * config/nds32/nds32.md (type): Add pbsad and pbsada.
7362 (btst, ave): New patterns for particular instructions.
7364 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7366 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7367 Add enum values for atomic load/store and memory sync.
7368 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
7370 * config/nds32/nds32-intrinsic.md: Likewise.
7371 * config/nds32/nds32_intrinsic.h: Likewise.
7372 * config/nds32/nds32.h (nds32_builtins): Likewise.
7374 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7376 PR tree-optimization/85257
7377 * fold-const.c (native_encode_vector): If not all elts could fit
7378 and off is -1, return 0 rather than offset.
7379 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
7380 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
7381 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
7382 adjust buffer in native_interpret_expr call.
7384 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7386 * config/nds32/constants.md (unspec_volatile_element): Add cache
7387 control enum values.
7388 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
7389 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
7390 * config/nds32/nds32.c (nds32_cctl_names): New.
7391 (nds32_print_operand): Handle cache control register names.
7392 * config/nds32/nds32.h (nds32_builtins): New enum values.
7393 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
7395 * config/nds32/nds32.md (type): Add mmu.
7396 * config/nds32/pipelines.md (simple_insn): Add mmu.
7398 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7400 * config/nds32/nds32.md (type): Remove call.
7401 * config/nds32/pipelines.md (simple_insn): Likewise.
7403 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7405 * config/nds32/constants.md (unspec_volatile_element): Add
7406 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
7407 UNSPEC_VOLATILE_FMFCFG.
7408 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
7409 description for fmfcfg and fmfcsr.
7410 (bdesc_1arg): Add fmtcsr.
7411 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
7412 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
7413 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
7414 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
7415 unspec_fmfcfg): New patterns.
7416 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
7417 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
7418 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
7419 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
7420 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
7421 __nds32__fmfcfg): Define.
7423 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7425 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
7426 intrinsic register names.
7427 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
7428 intrinsic register enum values and macros.
7430 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7432 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
7433 for load/store addressing form.
7434 (nds32_print_operand_address): Likewise.
7436 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
7439 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
7440 based on LABEL_REF. Remove useless assertion.
7441 (pic_address_needs_scratch): Fix formatting.
7442 (sparc_legitimize_pic_address): Minor tweaks.
7443 (sparc_delegitimize_address): Adjust assertion accordingly.
7444 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
7445 into symbolic_operand.
7446 (movsi_high_pic_label_ref): Likewise.
7447 (movsi_lo_sum_pic_label_ref): Likewise.
7448 (movdi_pic_label_ref): Likewise.
7449 (movdi_high_pic_label_ref): Likewise.
7450 (movdi_lo_sum_pic_label_ref): Likewise.
7452 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
7454 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
7455 custom LIB_SPEC setup.
7457 2018-04-06 Ruslan Bukin <br@bsdpad.com>
7458 Kito Cheng <kito.cheng@gmail.com>
7460 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
7461 * config/riscv/freebsd.h: New.
7463 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7465 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
7466 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
7469 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7470 Kito Cheng <kito.cheng@gmail.com>
7472 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
7473 nds32_output_call, nds32_symbol_binds_local_p): New functions.
7474 * config/nds32/nds32-protos.h (nds32_output_call,
7475 nds32_output_return): Declare.
7476 * config/nds32/nds32.md: Refine all the call and return patterns.
7478 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7481 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
7482 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
7484 PR rtl-optimization/84872
7485 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
7486 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
7489 2018-04-06 Tamar Christina <tamar.christina@arm.com>
7491 * expr.c (copy_blkmode_to_reg): Revert 254862.
7492 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
7494 2018-04-06 Richard Biener <rguenther@suse.de>
7497 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
7498 after seeing a component reference with an adjacent field. Treat
7499 refs to arrays at struct end of external decls similar to
7500 refs to unconstrained commons.
7502 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7505 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
7506 look through SAVE_EXPRs with non-side-effects argument. Adjust
7508 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
7511 2018-04-06 Richard Biener <rguenther@suse.de>
7514 * alias.c (find_base_term): New wrapper around find_base_term
7515 unwinding CSELIB_VAL_PTR changes.
7516 (find_base_term): Do not restore CSELIB_VAL_PTR during the
7519 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7521 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
7523 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
7524 constant definitions.
7525 ("nop"): lr 0,0 -> nopr r0
7526 ("nop_lr0", "nop_lr1"): New insn definitions.
7528 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7530 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
7531 NDS32_V3PUSH_AVAILABLE_P macro.
7533 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
7534 Chung-Ju Wu <jasonwucj@gmail.com>
7536 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
7537 (nds32*-*-*): Add float and fpu_config into supported_defaults.
7538 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
7539 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
7540 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
7541 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
7542 * config/nds32/constraints.md: New constraints and checking for hard
7543 float configuration.
7544 * config/nds32/iterators.md: New mode iterator and attribute for hard
7545 float configuration.
7546 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
7548 * config/nds32/nds32-fpu.md: New file.
7549 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
7550 deal with hard float code generation.
7551 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
7553 (abi_type, float_reg_number): New enum type.
7554 * config/nds32/nds32-predicates.c: New predicates for hard float.
7555 * config/nds32/nds32-protos.h: Declare functions for hard float.
7556 * config/nds32/nds32.c: Implementation for hard float configuration.
7557 * config/nds32/nds32.h: Definitions for hard float configuration.
7558 * config/nds32/nds32.md: Include hard float machine description and
7559 modify patterns for hard float configuration.
7560 * config/nds32/nds32.opt: New options for hard float configuration.
7561 * config/nds32/predicates.md: New predicates for hard float
7564 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7566 * common/config/nds32/nds32-common.c
7567 (nds32_option_optimization_table): Enable -mreleax-hint by default.
7569 2018-04-05 Jakub Jelinek <jakub@redhat.com>
7572 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
7573 CONSTRUCTOR_ELT (ctor, ...)->value.
7575 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
7578 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
7580 2018-04-05 Tom de Vries <tom@codesourcery.com>
7583 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
7586 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7587 Kito Cheng <kito.cheng@gmail.com>
7589 * config/nds32/constraints.md (U33): Fine-tune checking condition.
7590 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
7591 * config/nds32/nds32.h (nds32_16bit_address_type): Add
7592 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
7594 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7595 Kito Cheng <kito.cheng@gmail.com>
7597 * config/nds32/constraints.md (Ufe): New memory constraint.
7598 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
7599 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
7600 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
7602 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
7603 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
7605 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7607 * config/nds32/nds32.md: Use optimize_size in the condition for
7608 alu-shift instructions.
7610 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7612 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
7614 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7616 * config/nds32/nds32.md (negsi2): Refine pattern.
7618 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7619 Chung-Ju Wu <jasonwucj@gmail.com>
7621 * config/nds32/iterators.md (shift_rotate): New code iterator.
7622 (shift): New code attribute.
7623 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
7624 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
7625 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
7626 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
7627 bit-wise operations.
7628 (andsi3, *andsi3): Ditto.
7629 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
7630 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
7631 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
7632 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
7633 nds32_ior_operand, nds32_xor_operand): New predicates.
7635 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7637 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
7638 (addsi3, subsi3): ... this.
7640 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7642 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
7644 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7646 * config/nds32/nds32.md: Adjust indention.
7648 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7650 * config/nds32/nds32.md (feature): New attribute.
7652 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7654 * config/nds32/nds32.md (subtype): New attribute.
7656 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
7659 * config/arm/arm-builtins.c (arm_expand_builtin): Change
7660 expansion to perform a bitwise AND of the argument followed by a
7661 boolean negation of the result.
7663 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
7665 PR rtl-optimization/84878
7666 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
7667 the basic block. Assert the use reference is not artificial and that
7668 it has an associated insn.
7670 2018-04-04 Michael Matz <matz@suse.de>
7672 * builtins.c (compute_objsize): Pass correct operand
7673 to array_at_struct_end_p.
7675 2018-04-04 Richard Biener <rguenther@suse.de>
7678 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
7679 from contexts for DINFO_LEVEL_TERSE and below.
7681 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7683 * config/nds32/nds32-doubleword.md (move_<mode>): Require
7684 resiter_operand condition.
7685 * config/nds32/nds32.md (*move<mode>): Ditto.
7687 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7688 Monk Chiang <sh.chiang04@gmail.com>
7690 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
7692 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7694 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
7696 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7697 Kito Cheng <kito.cheng@gmail.com>
7699 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
7700 nds32_cond_code_str, output_cond_branch,
7701 output_cond_branch_compare_zero, nds32_expand_cbranch,
7702 nds32_expand_cstore, nds32_expand_movcc,
7703 nds32_output_cbranchsi4_equality_zero,
7704 nds32_output_cbranchsi4_equality_reg,
7705 nds32_output_cbranchsi4_equality_reg_or_const_int,
7706 nds32_output_cbranchsi4_greater_less_zero: New functions.
7707 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
7708 nds32_expand_cstore, nds32_expand_movcc,
7709 nds32_output_cbranchsi4_equality_zero,
7710 nds32_output_cbranchsi4_equality_reg,
7711 nds32_output_cbranchsi4_equality_reg_or_const_int,
7712 nds32_output_cbranchsi4_greater_less_zero): Declare.
7713 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
7714 nds32_rimm11s_operand): New predicates.
7715 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
7716 * config/nds32/nds32.md: Rewrite all the branch and conditional move
7719 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7721 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
7722 * config/nds32/nds32.md: Ditto.
7723 * config/nds32/pipelines.md: Ditto.
7725 2018-04-04 Richard Biener <rguenther@suse.de>
7727 PR tree-optimization/85168
7728 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
7729 propagating abnormals.
7731 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7733 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
7735 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7736 Kito Cheng <kito.cheng@gmail.com>
7738 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
7739 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
7740 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
7741 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
7742 * config/nds32/nds32.md (sibcall_internal): New.
7743 (sibcall_register): Remove.
7744 (sibcall_immediate): Remove.
7745 (sibcall_value_internal): New.
7746 (sibcall_value_register): Remove.
7747 (sibcall_value_immediate): Remove.
7748 * config/nds32/predicates.md (nds32_general_register_operand): New.
7749 (nds32_call_address_operand): New.
7751 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7753 PR rtl-optimization/85167
7754 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
7755 bb_defs if *split_p, instead preinitialize it to NULL.
7757 PR tree-optimization/85156
7758 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
7759 evaluating the argument multiple times.
7761 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
7763 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
7765 (_mm_cvtpd_ps): Likewise.
7766 (_mm_cvttpd_epi32): Likewise.
7767 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
7768 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
7769 vector, pixel, and bool following altivec.h include.
7771 2018-04-03 Martin Sebor <msebor@redhat.com>
7773 * doc/extend.texi (Common Function Attributes): Clarify.
7774 (const attribute): Likewise.
7775 (pure attribute): Likewise.
7777 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7780 * config/i386/i386.c (ix86_expand_vector_set): Use
7781 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
7783 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
7785 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
7786 instructions when changing rounding bits to preserve precision bits
7787 in the x87 control word.
7789 2018-04-03 Martin Liska <mliska@suse.cz>
7791 PR tree-optimization/82491
7792 * rtl.h (strip_offset_and_add): Replace += suboffset with
7793 poly_uint64 () + suboffset.
7795 2018-03-29 Martin Liska <mliska@suse.cz>
7796 Martin Jambor <mjambor@suse.cz>
7799 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
7800 param_type is not an integral or pointer type.
7802 2018-04-03 Richard Biener <rguenther@suse.de>
7804 * sese.h (recompute_all_dominators): Remove.
7806 2018-04-02 Martin Sebor <msebor@redhat.com>
7808 * doc/invoke.texi (-Wrestrict): Fix typos.
7810 2018-04-02 Jim Wilson <jimw@sifive.com>
7812 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
7813 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
7814 (<optab>di3, <optab>si3_extend): Likewise.
7815 (<optab>si3_mask, <optab>si3_mask_1): New.
7816 (<optab>di3_mask, <optab>di3_mask_1): New.
7817 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
7818 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
7819 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
7821 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
7823 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
7826 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
7828 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
7829 (nds32_canonicalize_comparison): New function.
7831 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7832 Kito Cheng <kito.cheng@gmail.com>
7833 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7835 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
7836 * config/nds32/constants.md (unspec_volatile_element): Add
7837 UNSPEC_VOLATILE_RELAX_GROUP.
7838 * config/nds32/nds32-relax-opt.c: New file.
7839 * config/nds32/nds32-predicates.c
7840 (nds32_symbol_load_store_p): New function.
7841 * config/nds32/nds32-protos.h
7842 (nds32_symbol_load_store_p): Declare function.
7843 (make_pass_nds32_relax_opt): Declare new rtl pass function.
7844 * config/nds32/nds32.c
7845 (nds32_register_pass): New function to register pass.
7846 (nds32_register_passes): New function to register passes.
7847 * config/nds32/nds32.md (relax_group): New pattern.
7848 * config/nds32/nds32.opt (mrelax-hint): New option.
7849 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
7851 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
7853 * config/nds32/t-nds32: Modify files dependency.
7855 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7857 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
7858 (PROFILE_HOOK): Define its implementation.
7860 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7862 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
7863 type and 32-bit size.
7865 2018-04-01 Jakub Jelinek <jakub@redhat.com>
7868 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
7869 (V_128_256): New mode iterator.
7870 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
7871 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
7872 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
7874 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
7875 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
7877 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
7880 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
7881 NaN inputs correctly.
7883 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
7886 * config/rs6000/vsx.md (??r): New mode attribute.
7887 (*vsx_mov<mode>_64bit): Use it.
7888 (*vsx_mov<mode>_32bit): Likewise.
7890 2018-03-30 Martin Sebor <msebor@redhat.com>
7892 PR tree-optimization/84818
7893 * builtins.c (check_access): Use warning_n.
7895 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7898 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
7900 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
7903 2018-03-30 Julia Koval <julia.koval@intel.com>
7906 * x86-tune.def (movx, partial_reg_dependency): Enable for
7909 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
7912 * lra-constraints.c (process_alt_operands): Move setting
7913 this_alternative_matches below.
7915 2018-03-29 Martin Liska <mliska@suse.cz>
7918 * doc/invoke.texi: Document how LTO works with debug info.
7919 Describe auto-load support of binutils. Mention 'x86-64'
7920 as valid option value of -march option.
7922 2018-03-29 Jakub Jelinek <jakub@redhat.com>
7924 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
7927 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
7928 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
7929 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
7932 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7935 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
7936 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
7937 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
7938 for RS6000_BTM_POWERPC64.
7939 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
7940 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
7941 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
7946 2018-03-28 Carl Love <cel@us.ibm.com>
7949 2017-09-27 Carl Love <cel@us.ibm.com>
7951 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
7952 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
7953 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
7956 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7958 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
7959 instead of __vector bool.
7960 (_mm_max_pu8): Likewise.
7961 (_mm_min_pi16): Likewise.
7963 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7966 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
7967 (DIVWEUO): Likewise.
7969 (DIVDEUO): Likewise.
7970 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
7971 DIVWEUO and DIVDEUO.
7972 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
7973 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
7974 (div_extend): Likewise.
7975 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
7977 (__builtin_divweuo): Likewise.
7978 (__builtin_divdeo): Likewise.
7979 (__builtin_divdeuo): Likewise.
7981 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7984 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
7985 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
7987 PR tree-optimization/82004
7988 * gimple-match-head.c (optimize_pow_to_exp): New function.
7989 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
7990 Don't fold to exp if optimize_pow_to_exp is false.
7992 2018-03-28 Martin Liska <mliska@suse.cz>
7995 * calls.c (initialize_argument_information): Fix trailing space.
7996 * common.opt: Fix typo and provide better explanation for
7997 -fsanitize-coverage option.
7998 * config/i386/i386.opt: Fix typo.
8000 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8001 Martin Liska <mliska@suse.cz>
8004 * gimplify.c (asan_poison_variable): Don't do the check for
8005 gimplify_omp_ctxp here.
8006 (gimplify_decl_expr): Do it here.
8007 (gimplify_target_expr): Likewise.
8009 2018-03-28 Martin Liska <mliska@suse.cz>
8012 * config/i386/i386.c (ix86_function_arg_advance): Do not call
8013 chkp_type_bounds_count if MPX is not enabled.
8015 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
8017 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
8019 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
8022 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
8023 function to create the function decl for complex long double
8024 multiply and divide for -mabi=ieeelongdouble.
8025 (init_float128_ieee): Call it.
8027 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
8030 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
8031 -fcf-protection=branch -mibt.
8032 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
8034 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8037 * config/arm/arm.c (arm_valid_symbolic_address): Handle
8038 arm_word_relocations.
8040 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
8043 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
8044 extern array declarations.
8046 2018-03-27 Richard Biener <rguenther@suse.de>
8049 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
8050 explicit single_use checks.
8052 2018-03-27 Richard Biener <rguenther@suse.de>
8054 PR tree-optimization/85082
8055 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
8058 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8060 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
8061 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
8062 Turn on fasynchronous-unwind-tables and funwind-tables.
8064 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
8067 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
8068 (*bmi_blsr_<mode>_ccz): Ditto.
8070 2018-03-26 Tom de Vries <tom@codesourcery.com>
8072 PR tree-optimization/85063
8073 * omp-general.c (offloading_function_p): New function. Factor out
8075 * omp-offload.c (pass_omp_target_link::gate): ... here.
8076 * omp-general.h (offloading_function_p): Declare.
8077 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
8078 with attribute omp declare target for offloading functions.
8080 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
8082 PR tree-optimization/84005
8083 * tree-data-ref.h (get_base_for_alignment): Declare.
8084 * tree-data-ref.c (get_base_for_alignment_1): New function.
8085 (get_base_for_alignment): Likewise.
8086 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
8087 get_base_for_alignment to find a suitable base object, instead
8088 of always using drb->base_address.
8090 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8093 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
8094 known size by default.
8096 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
8099 * lra-constraints.c (process_alt_operands): Don't match BLKmode
8100 and non BLKmode operands.
8102 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8105 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
8106 Clean up attributes.
8108 2018-03-23 Richard Biener <rguenther@suse.de>
8111 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
8112 we are going to emit early debug for LTO.
8114 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8117 * function.c (match_asm_constraints_1): Don't optimize if input
8118 doesn't satisfy general_operand predicate for output's mode.
8121 * alias.c (write_dependence_p): Don't require for x_canonicalized
8122 non-VOIDmode if x has VOIDmode.
8125 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
8126 just don't try to optimize it rather than assert it never happens.
8128 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8130 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
8131 macro expansions for definition of ST_INTERNAL_<mode> and
8132 LD_INTERNAL_<mode> builtins.
8133 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
8135 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
8137 (altivec_expand_st_builtin): Likewise.
8138 (altivec_expand_builtin): Remove calls to deleted functions.
8139 (rs6000_address_for_altivec): Delete this function.
8140 * config/rs6000/vector.md: Remove expands for
8141 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
8143 2018-03-22 Sudakshina Das <sudi.das@arm.com>
8146 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
8147 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
8148 re-computing once computed.
8149 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
8150 (arm_init_machine_status): Initialize
8151 machine->static_chain_stack_bytes.
8153 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8156 * doc/extend.texi: Add four new prototypes for vec_ld.
8157 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
8158 definitions for more logical presentation.
8159 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
8160 entries for V1TI variants of __builtin_altivec_ld builtin.
8161 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
8162 handling of V1TI variant of LVX icode pattern.
8163 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
8164 (rs6000_gimple_fold_builtin): Likewise.
8165 (altivec_init_builtins): Add code to define
8166 __builtin_altivec_lvx_v1ti function.
8168 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8171 * function.c (match_asm_constraints_1): Don't do the optimization
8172 if input isn't a REG, SUBREG, MEM or constant.
8174 2018-03-22 Tom de Vries <tom@codesourcery.com>
8176 PR tree-optimization/84956
8177 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
8178 bb_has_abnormal_pred.
8180 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8183 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
8184 DECL_INITIAL (decl) to decl at the end.
8185 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
8188 2018-03-21 Joseph Myers <joseph@codesourcery.com>
8190 * doc/extend.texi (__builtin_tgmath): Document when complex
8191 integer types are treated as _Complex _Float64.
8193 2018-03-21 Tom de Vries <tom@codesourcery.com>
8195 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
8197 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8199 PR tree-optimization/84960
8200 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
8201 if it is ENTRY block, move them into single succ of ENTRY in that case.
8203 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
8205 PR tree-optimization/84811
8206 * poly-int.h (poly_span_traits): Remove the T3 parameter and
8207 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
8208 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
8209 (known_subrange_p): Update accordingly. Cast each value involved
8210 in the size comparison, rather than casting the result of the
8213 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8215 PR tree-optimization/84982
8216 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
8217 by flipping the least significant bit rather than all bits from
8218 bitpos to bitpos + bitsize - 1.
8220 2018-03-21 Nathan Sidwell <nathan@acm.org>
8222 * doc/extend.texi (Deprecated Features): Remove mention of
8223 long-deleted deprecations.
8225 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8228 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
8229 * configure: Regenerate.
8231 2018-03-21 Tom de Vries <tom@codesourcery.com>
8233 PR tree-optimization/83126
8234 * tree-parloops.c (num_phis): New function.
8235 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
8237 2018-03-21 Nathan Sidwell <nathan@acm.org>
8239 * doc/extend.texi (Deprecated Features): Update deprecated flags,
8240 mention anon-struct/union members and trailing attributes.
8242 2018-03-21 Bin Cheng <bin.cheng@arm.com>
8244 PR tree-optimization/84969
8245 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
8246 builtin memset partitions if they set different rhs values.
8248 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8250 PR rtl-optimization/84989
8251 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
8252 VEC_DUPLICATE with scalar result mode.
8254 2018-03-21 Martin Liska <mliska@suse.cz>
8257 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
8258 not intended return statement.
8260 2018-03-21 Martin Liska <mliska@suse.cz>
8263 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
8264 (chkp_find_bound_slots_1): Limit number of iterations.
8266 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
8269 * Minor grammar fixes for x86 options.
8271 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8274 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
8275 holding REG_CFA_RESTORE notes, instead turn them into a USE.
8277 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
8280 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
8281 (altivec_lvx_<mode>_1op): Likewise.
8282 (altivec_stvx_<mode>_2op): Likewise.
8283 (altivec_stvx_<mode>_1op): Likewise.
8284 (altivec_lvx_<VM2:mode>): New define_expand.
8285 (altivec_stvx_<VM2:mode>): Likewise.
8286 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
8287 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8288 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
8289 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8290 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
8291 (rs6000_gen_lvx): Likewise.
8292 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
8293 (altivec_expand_stv_builtin): Likewise.
8294 (altivec_expand_builtin): Likewise.
8295 * config/rs6000/vector.md: Likewise.
8297 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8300 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
8303 2018-03-20 Richard Biener <rguenther@suse.de>
8306 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
8307 sign-conversions as zero, fall back to standard scalar_stmt
8310 2018-03-20 Martin Liska <mliska@suse.cz>
8313 * predict.c (rebuild_frequencies): Handle case when we have
8314 PROFILE_ABSENT, but flag_guess_branch_prob is false.
8316 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8319 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
8320 flag_section_anchors.
8321 * varasm.c (use_blocks_for_decl_p): Remove hack for
8322 dw2_force_const_mem.
8325 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
8327 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
8328 be created, use lowpart_subreg of operands[0] rather than operands[0]
8330 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
8331 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
8332 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
8333 and n constraint instead of aarch64_shift_imm_di and Usd.
8334 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
8335 (*aarch64_<optab>_reg_minus<mode>3): ... this.
8337 2018-03-20 Sudakshina Das <sudi.das@arm.com>
8340 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
8341 to favor GPR over NEON registers.
8342 (<shift>di3_neon): Likewise.
8344 2018-03-20 Tom de Vries <tom@codesourcery.com>
8347 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
8348 (nvptx_process_pars): Emit bar.sync asap and alap.
8350 2018-03-20 Tom de Vries <tom@codesourcery.com>
8353 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
8354 seen_label if seen_label is already set.
8356 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8359 * config/i386/i386.c (fold_builtin_cpu): For features above 31
8360 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
8361 Use 1U instead of 1. Formatting fixes.
8364 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
8365 instead of TREE_TYPE (s1) for the return value.
8367 2018-03-19 Jakub Jelinek <jakub@redhat.com>
8369 PR tree-optimization/84946
8370 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
8371 bitsize + bitsize in poly_uint64 rather than poly_int64.
8374 * dwarf2asm.c: Include fold-const.c.
8375 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
8376 of decl rather than decl itself.
8378 PR rtl-optimization/84643
8379 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
8381 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
8384 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
8385 calling assemble_variable.
8387 2018-03-19 Sudakshina Das <sudi.das@arm.com>
8390 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
8391 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
8393 2018-03-19 Jim Wilson <jimw@sifive.com>
8396 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
8397 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
8398 (riscv_first_stack_step): Likewise.
8399 (riscv_option_override): Use STACK_BOUNDARY instead of
8401 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
8403 (BIGGEST_ALIGNMENT): Set to 128.
8404 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
8405 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
8408 2018-03-19 Richard Biener <rguenther@suse.de>
8410 PR tree-optimization/84933
8411 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
8412 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
8414 2018-03-19 Richard Biener <rguenther@suse.de>
8416 PR tree-optimization/84859
8417 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
8418 (cond_if_else_store_replacement): Perform sinking operation on
8419 single-store BBs regardless of MAX_STORES_TO_SINK setting.
8420 Generalize what a BB with a single eligible store is.
8422 2018-03-19 Richard Biener <rguenther@suse.de>
8424 PR tree-optimization/84929
8425 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
8426 chrec_is_positive against non-chrec arg.
8428 2018-03-19 Tamar Christina <tamar.christina@arm.com>
8431 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
8433 2018-03-18 Martin Liska <mliska@suse.cz>
8435 PR rtl-optimization/84635
8436 * regrename.c (build_def_use): Use matches_mode only when
8439 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
8441 PR tree-optimization/84913
8442 * tree-vect-loop.c (vectorizable_reduction): Don't try to
8443 vectorize chains of COND_EXPRs.
8445 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8447 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
8449 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8451 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
8453 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8455 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
8457 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8458 Kito Cheng <kito.cheng@gmail.com>
8460 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
8461 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
8462 (nds32_adjust_reg_alloc_order): New function.
8463 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
8465 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
8467 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
8468 nds32_print_operand, nds32_print_operand_address): Use
8469 HOST_WIDE_INT_PRINT_DEC instead.
8471 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8473 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
8475 2018-03-17 Jakub Jelinek <jakub@redhat.com>
8478 * config/i386/i386.c (initial_ix86_tune_features,
8479 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
8481 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
8482 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
8483 rather than 1u << ix86_tune. Formatting fix.
8484 (ix86_option_override_internal): Change ix86_arch_mask from
8485 unsigned int to unsigned HOST_WIDE_INT, initialize to
8486 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
8487 (ix86_function_specific_restore): Likewise.
8489 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8492 * postreload.c (reload_combine_recognize_pattern): Perform
8493 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
8494 truncate_int_for_mode the result for the destination's mode.
8497 * hsa-gen.c (mem_type_for_type): Fix comment typo.
8498 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
8500 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8503 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
8506 * lra-assigns.c (lra_split_hard_reg_for): Don't use
8507 regno_allocno_class_array and sorted_pseudos.
8508 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
8509 insns where regno is used.
8511 2018-03-16 Martin Liska <mliska@suse.cz>
8514 * multiple_target.c (create_dispatcher_calls): Redirect
8515 reference in the symbol table.
8517 2018-03-16 Martin Liska <mliska@suse.cz>
8520 * multiple_target.c (create_dispatcher_calls): Redirect also
8523 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8527 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
8528 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
8529 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
8531 2018-03-16 Julia Koval <julia.koval@intel.com>
8533 * doc/invoke.texi (Skylake Server): Add CLWB.
8534 Cannonlake): Remove CLWB.
8536 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8538 PR tree-optimization/84841
8539 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
8541 (FLOAT_ONE_CONST_TYPE): Define.
8542 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
8543 (sort_by_operand_rank): Put entries with higher constant_type last
8544 rather than first to match comments.
8546 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
8548 * config/nios2/nios2.md (movsi_internal): Fix thinko in
8551 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8554 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
8555 check and use address of target always.
8557 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
8560 * config/i386/i386.c (indirect_thunk_needed): Update comments.
8561 (indirect_thunk_bnd_needed): Likewise.
8562 (indirect_thunks_used): Likewise.
8563 (indirect_thunks_bnd_used): Likewise.
8564 (indirect_return_needed): New.
8565 (indirect_return_bnd_needed): Likewise.
8566 (output_indirect_thunk_function): Add a bool argument for
8568 (output_indirect_thunk_function): Don't generate alias for
8569 function return thunk.
8570 (ix86_code_end): Call output_indirect_thunk_function to generate
8571 function return thunks.
8572 (ix86_output_function_return): Set indirect_return_bnd_needed
8573 and indirect_return_needed instead of indirect_thunk_bnd_needed
8574 and indirect_thunk_needed.
8576 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
8578 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
8579 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
8580 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
8582 2018-03-15 David Malcolm <dmalcolm@redhat.com>
8583 Paul Hua <paul.hua.gm@gmail.com>
8586 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
8588 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
8590 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
8591 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
8594 2018-03-15 Tamar Christina <tamar.christina@arm.com>
8597 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8598 instead of GET_MODE_SIZE when comparing Units.
8600 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
8603 * varasm.c (hash_section): Return an unchangeble hash value
8604 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
8605 Return !aarch64_can_use_per_function_literal_pools_p ().
8607 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8610 * optabs.c (emit_conditional_move): Pass address of cmode's copy
8611 rather than address of cmode as last argument to prepare_cmp_insn.
8613 2018-03-15 Julia Koval <julia.koval@intel.com>
8615 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
8616 F_AVX512VNNI, F_AVX512BITALG): New.
8618 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
8621 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
8622 insn for floating-point loads and stores.
8624 2018-03-14 Carl Love <cel@us.ibm.com>
8626 * config/rs6000/rs6000-c.c: Add macro definitions for
8627 ALTIVEC_BUILTIN_VEC_PERMXOR.
8628 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
8629 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
8630 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
8631 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
8633 * config/doc/extend.texi: Add prototypes for vec_permxor.
8635 2018-03-14 David Malcolm <dmalcolm@redhat.com>
8638 * diagnostic-show-locus.c (class layout_point): Convert m_line
8639 from int to linenum_type.
8640 (line_span::comparator): Use linenum "compare" function when
8641 comparing line numbers.
8642 (test_line_span): New function.
8643 (layout_range::contains_point): Convert param "row" from int to
8645 (layout_range::intersects_line_p): Likewise.
8646 (layout::will_show_line_p): Likewise.
8647 (layout::print_source_line): Likewise.
8648 (layout::should_print_annotation_line_p): Likewise.
8649 (layout::print_annotation_line): Likewise.
8650 (layout::print_leading_fixits): Likewise.
8651 (layout::annotation_line_showed_range_p): Likewise.
8652 (struct line_corrections): Likewise for field m_row.
8653 (line_corrections::line_corrections): Likewise for param "row".
8654 (layout::print_trailing_fixits): Likewise.
8655 (layout::get_state_at_point): Likewise.
8656 (layout::get_x_bound_for_row): Likewise.
8657 (layout::print_line): Likewise.
8658 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
8659 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
8660 * input.c (selftest::test_linenum_comparisons): New function.
8661 (selftest::input_c_tests): Call it.
8662 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
8663 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
8664 * selftest.h (ASSERT_GT): New macro.
8665 (ASSERT_GT_AT): New macro.
8666 (ASSERT_LT): New macro.
8667 (ASSERT_LT_AT): New macro.
8669 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
8671 PR rtl-optimization/84780
8672 * combine.c (distribute_links): Don't make a link based on pc_rtx.
8674 2018-03-14 Martin Liska <mliska@suse.cz>
8676 * tree.c (record_node_allocation_statistics): Use
8677 get_stats_node_kind.
8678 (get_stats_node_kind): New function extracted from
8679 record_node_allocation_statistics.
8680 (free_node): Use get_stats_node_kind.
8682 2018-03-14 Richard Biener <rguenther@suse.de>
8684 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
8685 that the value-set of ANTIC_IN doesn't grow.
8688 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8690 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8691 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8693 2018-03-14 Julia Koval <julia.koval@intel.com>
8695 * config.gcc (icelake-client, icelake-server): New.
8697 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
8698 (initial_ix86_arch_features): Ditto.
8699 (PTA_SKYLAKE): Add SGX.
8700 (PTA_ICELAKE): Remove.
8701 (PTA_ICELAKE_CLIENT): New.
8702 (PTA_ICELAKE_SERVER): New.
8703 (ix86_option_override_internal): Split up icelake on icelake client and
8705 (get_builtin_code_for_version): Ditto.
8706 (fold_builtin_cpu): Ditto.
8707 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
8708 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
8709 * config/i386/i386.h (processor_type): Ditto.
8710 * doc/invoke.texi: Ditto.
8712 2018-03-14 Jakub Jelinek <jakub@redhat.com>
8715 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
8716 INTEGER_CST offset, add it together with bitpos / 8 and
8717 sign extend based on POINTER_SIZE.
8721 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
8724 * config/i386/constraints.md (Yc): New register constraint.
8725 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
8726 Use Yc constraint for alternative 2 of operand 0. Remove
8727 preferred_for_speed attribute.
8729 2018-03-14 Richard Biener <rguenther@suse.de>
8731 PR tree-optimization/84830
8732 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
8733 with the old one to avoid oscillations.
8735 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
8738 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
8740 (assign_by_spills): Return a flag of reload assignment failure.
8741 Do not process the reload assignment failures. Do not spill other
8742 reload pseudos if they has the same reg class. Update n if
8744 (lra_assign): Add a return arg. Set up from the result of
8745 assign_by_spills call.
8746 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8747 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8748 usage_insns if it is not NULL.
8749 (spill_hard_reg_in_range): New function.
8750 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8751 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8752 function prototypes.
8753 (lra_assign): Change prototype.
8754 * lra.c (lra): Add code to deal with fails by splitting hard reg
8757 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
8759 * config/riscv/riscv.opt (mrelax): New option.
8760 * config/riscv/riscv.c (riscv_file_start): Emit ".option
8761 "norelax" when riscv_mrelax is disabled.
8762 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
8764 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8767 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
8768 reassociation for int modes.
8770 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8772 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
8773 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
8775 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
8776 * config/aarch64/aarch64-sve.md
8777 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
8778 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
8779 (*extend<mode><Vwide>2): Rename to...
8780 (aarch64_sve_extend<mode><Vwide>2): ...this.
8781 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
8782 renaming the old pattern to...
8783 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
8785 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
8786 define_expand, renaming the old pattern to...
8787 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
8788 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
8789 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
8790 account when deciding which SVE instruction the optab should use.
8791 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
8793 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8795 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
8796 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
8797 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
8798 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
8799 (tlsdesc_small_<mode>): Turn a define_expand and use
8800 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
8801 (tlsdesc_small_advsimd_<mode>): ...this.
8802 (tlsdesc_small_sve_<mode>): New pattern.
8804 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8806 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
8807 (UNSPEC_UMUL_HIGHPART): New constants.
8808 (MUL_HIGHPART): New int iteraor.
8809 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
8810 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
8812 (*<su>mul<mode>3_highpart): New define_insn.
8814 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8817 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8820 2018-03-13 Martin Liska <mliska@suse.cz>
8823 * (sem_item_optimizer::sem_item_optimizer): Initialize new
8825 (sem_item_optimizer::~sem_item_optimizer): Release it.
8826 (sem_item_optimizer::merge_classes): Register variable aliases.
8827 (sem_item_optimizer::fixup_pt_set): New function.
8828 (sem_item_optimizer::fixup_points_to_sets): Likewise.
8829 * ipa-icf.h: Declare new variables and functions.
8831 2018-03-13 Jakub Jelinek <jakub@redhat.com>
8834 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
8835 integer_pow2p@2 and test integer_pow2p in condition.
8836 (A < 0 ? C : 0): Similarly for @1.
8839 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
8840 characters starting at p contain '\0' character, don't look beyond
8844 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
8845 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
8848 * reg-stack.c (change_stack): Change update_end var from int to
8849 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
8850 also call set_block_for_insn on the newly added insns and rescan.
8853 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
8854 on the last operand.
8857 * tree.c (stabilize_reference_1): Return save_expr (e) for
8858 STATEMENT_LIST even if it doesn't have side-effects.
8860 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
8862 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
8864 2018-03-12 Renlin Li <renlin.li@arm.com>
8866 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
8867 aarch64_output_scalar_simd_mov_immediate.
8869 2018-03-12 Martin Sebor <msebor@redhat.com>
8871 PR tree-optimization/83456
8872 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
8873 for perfectly overlapping calls to memcpy.
8874 (gimple_fold_builtin_memory_chk): Same.
8875 (gimple_fold_builtin_strcpy): Handle no-warning.
8876 (gimple_fold_builtin_stxcpy_chk): Same.
8877 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
8879 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8881 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
8882 parameter. Use it for SFmode.
8883 (rs6000_function_arg_advance_1): Adjust.
8884 (rs6000_function_arg): Adjust.
8885 (rs6000_gimplify_va_arg): Pass false for that new parameter.
8887 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8889 PR rtl-optimization/84169
8890 PR rtl-optimization/84780
8891 * combine.c (can_combine_p): Check for a 2-insn combination whether
8892 the destination register is used between the two insns, too.
8894 2018-03-12 Richard Biener <rguenther@suse.de>
8896 PR tree-optimization/84803
8897 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
8898 for refs DR analysis didn't process.
8900 2018-03-12 Richard Biener <rguenther@suse.de>
8902 PR tree-optimization/84777
8903 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
8904 force-vectorize loops ignore whether we are optimizing for size.
8906 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
8908 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
8909 (TARGET_MD_ASM_ADJUST): Define.
8911 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
8912 Kito Cheng <kito.cheng@gmail.com>
8913 Chung-Ju Wu <jasonwucj@gmail.com>
8915 * config/nds32/nds32.c (nds32_compute_stack_frame,
8916 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
8917 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
8918 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
8919 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
8920 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
8921 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
8922 * config/nds32/nds32.md (prologue, epilogue): Use macro
8923 NDS32_V3PUSH_AVAILABLE_P to do checking.
8925 2018-03-11 Jakub Jelinek <jakub@redhat.com>
8928 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
8929 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
8930 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
8931 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
8932 addition of most attributes on !orig_type_die or the attribute not
8933 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
8935 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8936 Chung-Ju Wu <jasonwucj@gmail.com>
8938 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
8940 * config/nds32/nds32.opt (mvh): New option.
8942 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8943 Chung-Ju Wu <jasonwucj@gmail.com>
8945 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
8947 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
8948 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
8951 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8952 Chung-Ju Wu <jasonwucj@gmail.com>
8954 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
8956 * config/nds32/nds32-multiple.md (strlensi): New pattern.
8957 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
8959 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
8960 Kito Cheng <kito.cheng@gmail.com>
8961 Chung-Ju Wu <jasonwucj@gmail.com>
8963 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
8964 UNSPEC_FFMISM and UNSPEC_FLMISM.
8965 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
8966 for ffb, ffmism and flmism.
8967 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
8968 (unspec_ffmism): Ditto.
8969 (unspec_flmism): Ditto.
8970 (nds32_expand_builtin_impl): Check if string extension is available.
8971 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
8972 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
8974 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
8977 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8980 * lra-assigns.c (assign_by_spills): Return a flag of reload
8981 assignment failure. Do not process the reload assignment
8982 failures. Do not spill other reload pseudos if they has the same
8984 (lra_assign): Add a return arg. Set up from the result of
8985 assign_by_spills call.
8986 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8987 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8988 usage_insns if it is not NULL.
8989 (spill_hard_reg_in_range): New function.
8990 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8991 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8992 function prototypes.
8993 (lra_assign): Change prototype.
8994 * lra.c (lra): Add code to deal with fails by splitting hard reg
8997 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
9000 * config/i386/i386.opt: Replace Enforcment with Enforcement.
9002 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
9005 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
9006 (dw_val_node): Add val_symbolic_view.
9007 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
9008 (symview_upper_bound): New.
9009 (new_line_info_table): Initialize symviews_since_reset.
9010 (dwarf2out_source_line): Count symviews_since_reset and set
9011 symview_upper_bound.
9012 (dw_val_equal_p): Handle symview.
9013 (add_AT_symview): New.
9014 (print_dw_val): Handle symview.
9015 (attr_checksum, attr_checksum_ordered): Likewise.
9016 (same_dw_val_p, size_of_die): Likewise.
9017 (value_format, output_die): Likewise.
9018 (add_high_low_attributes): Use add_AT_symview for entry_view.
9019 (dwarf2out_finish): Reset symview_upper_bound, clear
9022 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
9025 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
9026 Add strict argument and use it.
9027 (rs6000_split_multireg_move): Update for new strict argument.
9028 (mem_operand_gpr): Disallow all non-offsettable addresses.
9029 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
9031 2018-03-09 Jakub Jelinek <jakub@redhat.com>
9034 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
9035 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
9036 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
9039 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
9040 decl, use remap_type if we want to use the type.
9042 2018-03-09 Martin Sebor <msebor@redhat.com>
9044 PR tree-optimization/84526
9045 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9047 (builtin_access::generic_overlap): Be prepared to handle non-array
9050 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
9052 PR rtl-optimization/84682
9053 * lra-constraints.c (process_address_1): Check is_address flag
9054 for address constraints.
9055 (process_alt_operands): Likewise.
9056 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
9057 preprocess_constraints.
9058 * recog.h (preprocess_constraints): Add oploc parameter.
9060 * recog.c (preprocess_constraints): Test address_operand for
9061 CT_ADDRESS constraints.
9063 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9066 * lra-assigns.c (assign_by_spills): Return a flag of reload
9067 assignment failure. Do not process the reload assignment
9068 failures. Do not spill other reload pseudos if they has the same
9070 (lra_assign): Add a return arg. Set up from the result of
9071 assign_by_spills call.
9072 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9073 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9074 usage_insns if it is not NULL.
9075 (spill_hard_reg_in_range): New function.
9076 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9077 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9078 function prototypes.
9079 (lra_assign): Change prototype.
9080 * lra.c (lra): Add code to deal with fails by splitting hard reg
9083 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9086 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
9087 Accept complain bool parameter. Only emit errors if it is true.
9088 (arm_parse_cpu_option_name): Likewise.
9089 (arm_target_thumb_only): Adjust callers of the above.
9090 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
9091 prototype to take a default true bool parameter.
9092 (arm_parse_arch_option_name): Likewise.
9094 2018-03-09 David Malcolm <dmalcolm@redhat.com>
9095 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
9099 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
9100 * configure: Regenerate.
9101 * configure.ac ("linker --version-script option"): New.
9102 ("linker soname option"): New.
9104 2018-03-09 Richard Biener <rguenther@suse.de>
9106 PR tree-optimization/84775
9107 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
9108 immediate uses of predicate stmts and mark them modified.
9111 PR tree-optimization/84178
9112 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9114 (version_loop_for_if_conversion): Delay update_ssa call.
9115 (tree_if_conversion): Delay update_ssa until after predicate
9118 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
9121 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
9122 when the function accesses prior frames.
9124 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9127 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
9128 gen_llsym, otherwise call maybe_gen_llsym.
9131 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
9132 has ',' character inside of it.
9134 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9137 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
9138 as clobbering CC_REGNUM.
9140 2018-03-08 Richard Biener <rguenther@suse.de>
9143 * tree-scalar-evolution.c: Include tree-into-ssa.h.
9144 (follow_copies_to_constant): Do not follow SSA names registered
9147 2018-03-08 Richard Biener <rguenther@suse.de>
9149 PR tree-optimization/84178
9150 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9152 (version_loop_for_if_conversion): Delay update_ssa call.
9153 (tree_if_conversion): Delay update_ssa until after predicate
9156 2018-03-08 David Malcolm <dmalcolm@redhat.com>
9158 PR tree-optimization/84178
9159 * tree-if-conv.c (release_bb_predicate): Remove the
9160 the assertion that the stmts have NULL use_ops.
9161 Discard the statements, asserting that they haven't
9162 yet been added to a BB.
9164 2018-03-08 Richard Biener <rguenther@suse.de>
9166 PR tree-optimization/84746
9167 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
9168 (phi_translate): Pass in destination ANTIC_OUT set.
9169 (phi_translate_1): Likewise. For a simplified result lookup
9170 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
9171 (phi_translate_set): Adjust.
9172 (do_pre_regular_insertion): Likewise.
9173 (do_pre_partial_partial_insertion): Likewise.
9175 2018-03-08 Martin Liska <mliska@suse.cz>
9177 PR gcov-profile/84735
9178 * doc/gcov.texi: Document usage of profile files.
9179 * gcov-io.h: Document changes in the format.
9181 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
9185 * dwarf2out.c (struct dw_line_info_table): Update comments for
9187 (FORCE_RESET_NEXT_VIEW): New.
9188 (FORCE_RESETTING_VIEW_P): New.
9189 (RESETTING_VIEW_P): Check for -1 too.
9190 (ZERO_VIEW_P): Likewise.
9191 (new_line_info_table): Force-reset next view.
9192 (dwarf2out_begin_function): Likewise.
9193 (dwarf2out_source_line): Simplify zero_view_p initialization.
9194 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
9195 view directly. Omit view when omitting .loc at line 0.
9197 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9199 PR tree-optimization/84740
9200 * tree-switch-conversion.c (process_switch): Call build_constructors
9201 only if info.phi_count is non-zero.
9203 PR tree-optimization/84739
9204 * tree-tailcall.c (find_tail_calls): Check call arguments against
9205 DECL_ARGUMENTS (current_function_decl) rather than
9206 DECL_ARGUMENTS (func) when checking for tail recursion.
9208 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9210 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
9211 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
9212 Volker Reichelt's entry and add entries for people that perform
9213 GCC fuzzy testing and report numerous bugs.
9215 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
9218 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
9219 readonly data in sdata, if that is disabled.
9220 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
9221 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
9222 -mreadonly-in-sdata option.
9224 2018-03-07 Martin Sebor <msebor@redhat.com>
9226 PR tree-optimization/84468
9227 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
9228 basic block when looking for nul assignment.
9230 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
9233 * except.h (output_function_exception_table): Adjust prototype.
9234 * except.c (output_function_exception_table): Remove FNNAME parameter
9235 and add SECTION parameter. Ouput one part of the table at a time.
9236 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
9237 the first part of the exception table and emit unwind directives.
9238 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
9239 (i386_pe_seh_cold_init): Likewise.
9240 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
9241 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
9242 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
9243 (ix86_output_call_insn): Emit a nop in one more case for SEH.
9244 * config/i386/winnt.c: Include except.h.
9245 (struct seh_frame_state): Add reg_offset, after_prologue and
9246 in_cold_section fields.
9247 (i386_pe_seh_end_prologue): Set seh->after_prologue.
9248 (i386_pe_seh_cold_init): New function.
9249 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
9250 to seh->in_cold_section.
9251 (seh_emit_push): Record the offset of the push.
9252 (seh_emit_save): Record the offet of the save.
9253 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
9254 Test seh->after_prologue to disregard the epilogue.
9255 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
9256 (i386_pe_end_cold_function): New function.
9258 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9261 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
9262 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
9265 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
9267 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
9268 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
9271 * multiple_target.c: Include tree-inline.h and intl.h.
9272 (expand_target_clones): Diagnose and fail if node->definition and
9273 !tree_versionable_function_p (node->decl).
9275 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
9277 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
9279 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
9280 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
9281 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
9283 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9286 * combine.c (try_combine): Use reg_or_subregno instead of handling
9287 just paradoxical SUBREGs and REGs.
9289 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
9291 * config/arc/arc.c (arc_finalize_pic): Remove function.
9292 (arc_must_save_register): We use single base PIC register, remove
9293 checks to save/restore the PIC register.
9294 (arc_expand_prologue): Likewise.
9295 * config/arc/arc-protos.h (arc_set_default_type_attributes):
9297 (arc_verify_short): Likewise.
9298 (arc_attr_type): Likewise.
9299 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
9300 (walk_stores): Likewise.
9301 (arc_address_cost): Make it static.
9302 (arc_verify_short): Likewise.
9303 (branch_dest): Likewise.
9304 (arc_attr_type): Likewise.
9305 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
9306 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
9307 (arc_final_prescan_insn): Remove inserting the nops due to
9308 hardware hazards. It is done in reorg step.
9309 (insn_length_variant_t): Remove.
9310 (insn_length_parameters_t): Likewise.
9311 (arc_insn_length_parameters): Likewise.
9312 (arc_get_insn_variants): Likewise.
9313 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
9315 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9318 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
9321 PR tree-optimization/84687
9322 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
9324 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
9326 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9328 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
9329 Rename to ppc_speculation_barrier.
9330 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
9331 __builtin_ppc_speculation_barrier.
9333 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9336 * combine.c (combine_simplify_rtx): Don't try to simplify if
9337 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
9340 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
9342 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
9343 to 32 bytes when compiling for POWER9.
9345 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9348 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
9349 regparm >= 3 with no arg reg available also for calls with
9350 flag_force_indirect_call. Pass decl to ix86_function_regparm.
9353 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
9355 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
9357 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
9360 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
9362 2018-03-05 Richard Biener <rguenther@suse.de>
9364 PR tree-optimization/84486
9365 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
9366 When inserting a __builtin_assume_aligned call set the LHS
9367 SSA name alignment info accordingly.
9369 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
9371 PR tree-optimization/84114
9372 * config/aarch64/aarch64.c (aarch64_reassociation_width)
9373 Avoid reassociation of FLOAT_MODE addition.
9375 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
9377 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
9378 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
9379 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
9380 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
9381 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
9382 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
9383 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
9385 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
9386 __builtin_ia32_wbinvd): New builtins.
9387 (SPECIAL_ARGS2): New.
9388 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
9389 (SPECIAL_ARGS2): New.
9390 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
9391 (ix86_valid_target_attribute_inner_p): Ditto.
9392 (ix86_init_mmx_sse_builtins): Add special_args2.
9393 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
9394 TARGET_WBNOINVD_P): New.
9395 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
9396 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
9397 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
9398 * config/i386/immintrin.h (_wbinvd): New intrinsic.
9399 * config/i386/pconfigintrin.h: New file.
9400 * config/i386/wbnoinvdintrin.h: Ditto.
9401 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
9403 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
9405 2018-03-05 Richard Biener <rguenther@suse.de>
9407 PR tree-optimization/84670
9408 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9410 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9411 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9412 (compute_antic_aux): Only assert the number of values in ANTIC_IN
9413 doesn't grow if all successors (recursively) were visited at least
9416 2018-03-05 Richard Biener <rguenther@suse.de>
9418 PR tree-optimization/84650
9419 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
9420 if executed in the loop pipeline.
9422 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
9424 * doc/configfiles.texi (Configuration Files): Move info about
9425 conditionalizing $target-protos.h to...
9426 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
9427 differs from $target-protos.h.
9429 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
9430 Chung-Ju Wu <jasonwucj@gmail.com>
9432 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
9433 * config/nds32/nds32-multiple.md (setmemsi): Define.
9434 * config/nds32/nds32-memory-manipulation.c
9435 (nds32_gen_dup_4_byte_to_word_value): New.
9436 (emit_setmem_word_loop): New.
9437 (emit_setmem_byte_loop): New.
9438 (nds32_expand_setmem_loop): New.
9439 (nds32_expand_setmem_loop_v3m): New.
9440 (nds32_expand_setmem_unroll): New.
9441 (nds32_expand_setmem): New.
9443 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9444 Chung-Ju Wu <jasonwucj@gmail.com>
9446 * config/nds32/nds32-memory-manipulation.c
9447 (nds32_emit_load_store): New.
9448 (nds32_emit_post_inc_load_store): New.
9449 (nds32_emit_mem_move): New.
9450 (nds32_emit_mem_move_block): New.
9451 (nds32_expand_movmemsi_loop_unknown_size): New.
9452 (nds32_expand_movmemsi_loop_known_size): New.
9453 (nds32_expand_movmemsi_loop): New.
9454 (nds32_expand_movmemsi_unroll): New.
9455 (nds32_expand_movmemqi): Rename ...
9456 (nds32_expand_movmemsi): ... to this.
9457 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
9458 (movmemsi): ... to this.
9459 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
9460 (nds32_expand_movmemsi): ... to this.
9462 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9463 Monk Chiang <sh.chiang04@gmail.com>
9464 Chung-Ju Wu <jasonwucj@gmail.com>
9466 * config/nds32/nds32-protos.h
9467 (nds32_expand_load_multiple): New arguments.
9468 (nds32_expand_store_multiple): Ditto.
9469 (nds32_valid_multiple_load_store): Rename ...
9470 (nds32_valid_multiple_load_store_p): ... to this.
9471 * config/nds32/nds32-memory-manipulation.c
9472 (nds32_expand_load_multiple): Refine implementation.
9473 (nds32_expand_store_multiple): Ditto.
9474 * config/nds32/nds32-multiple.md
9475 (load_multiple): Update nds32_expand_load_multiple interface.
9476 (store_multiple): Update nds32_expand_store_multiple interface.
9477 * config/nds32/nds32-predicates.c
9478 (nds32_valid_multiple_load_store): Rename ...
9479 (nds32_valid_multiple_load_store_p): ... to this and refine
9481 * config/nds32/predicates.md
9482 (nds32_load_multiple_and_update_address_operation): New predicate.
9483 (nds32_store_multiple_and_update_address_operation): New predicate.
9485 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9486 Chung-Ju Wu <jasonwucj@gmail.com>
9488 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
9489 (combo): New attribute.
9490 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
9492 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
9494 * config/nds32/nds32.opt: Change -mcmodel= default value.
9496 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
9497 Monk Chiang <sh.chiang04@gmail.com>
9498 Chung-Ju Wu <jasonwucj@gmail.com>
9500 * config/nds32/constants.md (unspec_element): New enum.
9501 * config/nds32/constraints.md (Umw): New constraint.
9502 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
9503 * config/nds32/nds32-intrinsic.md: Likewise.
9504 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
9505 (nds32_valid_smw_lwm_base_p): New.
9506 (nds32_output_smw_single_word): New.
9507 (nds32_output_lmw_single_word): New.
9508 (nds32_expand_unaligned_load): New.
9509 (nds32_expand_unaligned_store): New.
9510 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
9511 (nds32_output_smw_single_word): Declare.
9512 (nds32_output_lmw_single_word): Declare.
9513 (nds32_expand_unaligned_load): Declare.
9514 (nds32_expand_unaligned_store): Declare.
9515 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
9516 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
9517 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
9518 NDS32_BUILTIN_UASTORE_DW.
9519 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
9522 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
9523 Kito Cheng <kito.cheng@gmail.com>
9524 Chung-Ju Wu <jasonwucj@gmail.com>
9526 * config/nds32/nds32-intrinsic.c
9527 (nds32_expand_builtin_null_ftype_reg): Delete.
9528 (nds32_expand_builtin_reg_ftype_imm): Ditto.
9529 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
9530 (nds32_read_argument): New.
9531 (nds32_legitimize_target): Ditto.
9532 (nds32_legitimize_argument): Ditto.
9533 (nds32_check_constant_argument): Ditto.
9534 (nds32_expand_unop_builtin): Ditto.
9535 (nds32_expand_unopimm_builtin): Ditto.
9536 (nds32_expand_binop_builtin): Ditto.
9537 (nds32_builtin_decl_impl): Ditto.
9538 (builtin_description): Ditto.
9539 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
9540 (nds32_init_builtins_impl): Ditto.
9541 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
9542 (nds32_builtin_decl): New.
9543 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
9544 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
9546 2018-03-02 Jeff Law <law@redhat.com>
9548 * reorg.c (stop_search_p): Handle DEBUG_INSN.
9549 (redundant_insn, fill_simple_delay_slots): Likewise.
9550 (fill_slots_from_thread): Likewise.
9551 * resource.c (mark_referenced_resources): Likewise.
9552 (mark_set_resources, find_dead_or_set_registers): Likewise.
9554 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9556 * substring-locations.h (format_warning_va): Formatting fix for
9558 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
9560 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
9561 * substring-locations.c: Include intl.h.
9562 (format_warning_va): Turned into small wrapper around
9563 format_warning_n_va, renamed to ...
9564 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
9565 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
9567 (format_warning_at_substring_n): New function.
9568 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
9569 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
9570 format_warning_at_substring with just a shorter name instead of
9571 const function pointer.
9572 (fmtwarn_n): New function.
9573 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
9574 appropriate, get rid of all the fmtstr temporaries, move conditionals
9575 with G_() wrapped string literals directly into fmtwarn arguments,
9576 cast dir.len to (int), formatting fixes.
9578 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
9580 * doc/invoke.texi: Remove "Cilk Plus" references.
9582 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9583 Richard Biener <rguenther@suse.de>
9586 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
9587 for error or warning attributes if CALL_FROM_THUNK_P is set.
9590 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9593 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
9594 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
9597 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
9598 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
9600 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
9601 instead of -1U in last predictors element's probability member.
9603 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9606 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9607 arguments if they are comparable.
9609 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9611 PR tree-optimization/84634
9612 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
9613 masks and masked_loop_p with a single loop_masks, making sure it's
9614 null for bb vectorization.
9616 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9618 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9619 (vect_analyze_data_ref_access): Use loop->safe_len rather than
9620 loop->force_vectorize to check whether there is no alias.
9622 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9625 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
9627 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
9629 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
9630 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
9631 instead of a loop around prev_real_insn.
9632 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
9636 * config/i386/i386.c (ix86_print_operand): Use conditional
9637 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
9640 2018-03-02 Richard Biener <rguenther@suse.de>
9642 PR tree-optimization/84427
9643 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
9644 (bitmap_set_subtract_values): Rewrite to handle multiple
9647 (prune_clobbered_mems): Likewise.
9648 (phi_translate): Take edge instead of pred/phiblock.
9649 (phi_translate_1): Likewise.
9650 (phi_translate_set): Likewise. Insert all translated
9651 exprs for a value into the set, keeping possibly multiple
9652 expressions per value.
9653 (compute_antic_aux): Adjust for phi_translate changes.
9654 When intersecting union the expressions and prune those
9655 not in the final value set, keeping possibly multiple
9656 expressions per value. Do not use value-insertion
9657 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
9658 all expressions. Add verification that the value-sets
9659 only shrink during iteration.
9660 (compute_partial_antic_aux): Adjust for the phi_translate changes.
9661 (do_pre_regular_insertion): Likewise.
9662 (do_pre_partial_partial_insertion): Likewise.
9664 2018-03-02 Richard Biener <rguenther@suse.de>
9667 * config/darwin.c (saved_debug_info_level): New static global.
9668 (darwin_asm_lto_start): Disable debug info generation for LTO out.
9669 (darwin_asm_lto_end): Restore debug info generation settings.
9671 2018-03-01 Martin Liska <mliska@suse.cz>
9674 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
9677 2018-03-01 Richard Biener <rguenther@suse.de>
9680 * dwarf2out.c (gen_variable_die): Properly handle late VLA
9681 type annotation with LTO when debug was disabled at compile-time.
9683 2018-03-01 Matthew Fortune <mfortune@gmail.com>
9685 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
9687 (mips_final_postscan_insn): Likewise.
9689 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
9691 PR rtl-optimization/84528
9692 * alias.c (init_alias_target): Add commentary.
9693 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
9694 a unique base value if the frame pointer is not eliminated
9695 to the stack pointer.
9697 2018-03-01 Tom de Vries <tom@codesourcery.com>
9699 PR rtl-optimization/83327
9700 * lra-int.h (hard_regs_spilled_into): Declare.
9701 * lra.c (hard_regs_spilled_into): Define.
9702 (init_reg_info): Init hard_regs_spilled_into.
9703 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
9704 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
9705 (process_bb_lives): Handle hard_regs_spilled_into.
9706 (lra_create_live_ranges_1): Before doing liveness propagation, clear
9707 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
9709 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
9711 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
9712 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
9713 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
9714 * config/rs6000/aix72.h: New file.
9716 2018-02-28 Jakub Jelinek <jakub@redhat.com>
9718 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
9719 instead of warning_at with conditional singular and plural messages
9723 * stor-layout.c (update_alignment_for_field): For
9724 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
9725 && !DECL_PACKED (field), do the alignment update, just use
9726 only desired_align instead of MAX (type_align, desired_align)
9728 (place_field): Don't do known_align < desired_align handling
9729 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
9730 is non-NULL, instead do it after rli->prev_field handling and
9731 only if not within a bitfield word. For DECL_PACKED (field)
9732 use type_align of BITS_PER_UNIT.
9734 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
9736 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
9737 superfluous parentheses and trailing spaces.
9739 2018-02-28 Richard Biener <rguenther@suse.de>
9741 PR tree-optimization/84584
9742 * graphite-scop-detection.c (scop_detection::add_scop): Discard
9743 SCoPs with fake exit edge.
9745 2018-02-28 Martin Liska <mliska@suse.cz>
9748 * timevar.c (timer::print): Fix format to properly print 100%
9751 2018-02-28 Richard Biener <rguenther@suse.de>
9754 * genmatch.c (capture_info::walk_match): Do not mark
9755 captured expressions without operands as expr_p given
9756 they act more like predicates and should be subject to
9757 "lost tail" side-effect preserving.
9759 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
9761 PR rtl-optimization/81611
9762 * auto-inc-dec.c (attempt_change): Move dead note from
9763 mem_insn if it's the next use of regno
9764 (find_address): Take address use of reg holding
9765 non-incremented value. Add parm to limit search to the named
9767 (merge_in_block): Attempt to use a mem insn that is the next
9768 use of the original regno.
9770 2018-02-27 Martin Sebor <msebor@redhat.com>
9773 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
9774 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
9776 2018-02-27 Martin Sebor <msebor@redhat.com>
9778 PR translation/84207
9779 * diagnostic-core.h (warning_n, error_n, inform_n): Change
9780 n argument to unsigned HOST_WIDE_INT.
9781 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
9782 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
9783 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
9784 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
9786 2018-02-27 Richard Biener <rguenther@suse.de>
9788 PR tree-optimization/84512
9789 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
9790 Do not use the estimate returned from record_stmt_cost for
9791 the scalar iteration cost but sum properly using add_stmt_cost.
9793 2018-02-27 Richard Biener <rguenther@suse.de>
9795 PR tree-optimization/84466
9796 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9797 Adjust last change to less strictly validate use operands.
9799 2018-02-27 Martin Liska <mliska@suse.cz>
9801 PR gcov-profile/84548
9802 * gcov.c (process_file): Allow partial overlap and consider it
9803 also as group functions.
9804 (output_lines): Properly calculate range of lines for a group.
9806 2018-02-27 Martin Liska <mliska@suse.cz>
9808 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
9809 'ggc' suffixes. Change first column width.
9810 (timer::print): Fix formatting of the column.
9812 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
9814 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
9815 preserve inline entry blocks for the sake of debug inline
9816 entry point markers alone.
9817 (remove_unused_locals): Suggest in comments a better place to
9818 force the preservation of inline entry blocks that are
9819 otherwise unused, but do not preserve them.
9821 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9823 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
9825 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9828 * config/i386/constraints.md (Bs): Replace
9829 ix86_indirect_branch_register with
9830 TARGET_INDIRECT_BRANCH_REGISTER.
9832 * config/i386/i386.md (indirect_jump): Likewise.
9833 (tablejump): Likewise.
9834 (*sibcall_memory): Likewise.
9835 (*sibcall_value_memory): Likewise.
9836 Peepholes of indirect call and jump via memory: Likewise.
9837 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
9838 (*sibcall_value_GOT_32): Likewise.
9839 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9840 (GOT_memory_operand): Likewise.
9841 (call_insn_operand): Likewise.
9842 (sibcall_insn_operand): Likewise.
9843 (GOT32_symbol_operand): Likewise.
9844 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
9846 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9848 PR rtl-optimization/83496
9849 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
9850 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
9851 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
9852 redundant insn, if any.
9853 (relax_delay_slots): Likewise.
9854 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
9856 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
9858 PR tree-optimization/83965
9859 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
9860 that grouped statements are part of a reduction chain. Return
9861 true if the statement is not marked as a reduction itself but
9863 (vect_recog_dot_prod_pattern): Don't check whether the statement
9864 is part of a group here.
9865 (vect_recog_sad_pattern): Likewise.
9866 (vect_recog_widen_sum_pattern): Likewise.
9868 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9871 * final.c (rest_of_clean_state): Also look for calls inside sequences.
9873 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9876 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
9878 (ix86_output_indirect_function_return): New prototype.
9879 (ix86_split_simple_return_pop_internal): Likewise.
9880 * config/i386/i386.c (indirect_return_via_cx): New.
9881 (indirect_return_via_cx_bnd): Likewise.
9882 (indirect_thunk_name): Handle return va CX_REG.
9883 (output_indirect_thunk_function): Create alias for
9884 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
9885 (ix86_output_indirect_jmp): Remove the bool argument.
9886 (ix86_output_indirect_function_return): New function.
9887 (ix86_split_simple_return_pop_internal): Likewise.
9888 * config/i386/i386.md (*indirect_jump): Don't pass false
9889 to ix86_output_indirect_jmp.
9890 (*tablejump_1): Likewise.
9891 (simple_return_pop_internal): Change it to define_insn_and_split.
9892 Call ix86_split_simple_return_pop_internal to split it for
9894 (simple_return_indirect_internal): Call
9895 ix86_output_indirect_function_return instead of
9896 ix86_output_indirect_jmp.
9898 2018-02-26 Jakub Jelinek <jakub@redhat.com>
9901 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
9902 memset and value initialization afterwards.
9904 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
9906 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
9908 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9911 * common/config/aarch64/aarch64-common.c
9912 (aarch_option_optimization_table[]): Switch
9913 off fomit-frame-pointer
9915 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9916 Chung-Ju Wu <jasonwucj@gmail.com>
9918 * config/nds32/nds32-multiple.md (load_multiple): Disallow
9920 (store_multiple): Ditto.
9922 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9924 * config.gcc: Add --with-cpu support for nds32 target.
9925 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
9926 * config/nds32/nds32.opt: Add -mcpu= option.
9928 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
9930 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
9931 isel=yes): Warn for these deprecated options.
9933 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
9935 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
9936 ISA_2_5_MASKS_EMBEDDED.
9938 2018-02-23 Jakub Jelinek <jakub@redhat.com>
9940 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
9941 p->max as pointers rather than using iterative_hash_expr.
9943 2018-02-23 Carl Love <cel@us.ibm.com>
9945 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
9946 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
9948 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
9949 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
9950 P8V_BUILTIN_VEC_VUNSIGNED2.
9952 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
9955 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
9956 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
9958 * lra-constraints.c (curr_insn_transform): Set up
9959 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
9961 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
9962 * lra-eliminations.c (spill_pseudos): Ditto.
9963 (process_insn_for_elimination): Ditto.
9964 * lra-lives.c (reg_early_clobber_p): Use the new macros.
9965 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
9966 LRA_NON_CLOBBERED_ALT.
9968 2018-02-22 Martin Sebor <msebor@redhat.com>
9970 PR tree-optimization/84480
9971 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
9972 to maybe_diag_stxncpy_trunc. Call it.
9973 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
9974 from gimple_fold_builtin_strcpy. Print inlining stack.
9975 (handle_builtin_stxncpy): Print inlining stack.
9976 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
9978 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
9981 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
9982 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
9983 and -fcheck-pointer-bounds are used together.
9984 (indirect_thunk_prefix): New enum.
9985 (indirect_thunk_need_prefix): New function.
9986 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
9987 "_nt" instead of "_bnd" for NOTRACK prefix.
9988 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
9989 (output_indirect_thunk_function): Likewise.
9991 (ix86_code_end): Update output_indirect_thunk_function calls.
9992 (ix86_output_indirect_branch_via_reg): Replace
9993 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
9994 (ix86_output_indirect_branch_via_push): Likewise.
9995 (ix86_output_function_return): Likewise.
9996 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
9997 incompatible with -fcf-protection=branch and
9998 -fcheck-pointer-bounds.
10000 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10003 * config/aarch64/aarch64.c (aarch64_print_address_internal):
10004 Change gcc_assert call to output_operand_lossage.
10006 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10008 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
10010 2018-02-22 DJ Delorie <dj@redhat.com>
10011 Sebastian Perta <sebastian.perta@renesas.com>
10012 Oleg Endo <olegendo@gcc.gnu.org>
10014 * config/rx/rx.c (rx_rtx_costs): New function.
10015 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
10017 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
10019 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
10021 2018-02-22 Martin Liska <mliska@suse.cz>
10024 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
10025 Add "native" as a possible value.
10027 2018-02-22 Martin Liska <mliska@suse.cz>
10030 * config/i386/i386.c (ix86_option_override_internal):
10031 Add "native" as a possible value for -march and -mtune.
10033 2018-02-22 Jakub Jelinek <jakub@redhat.com>
10036 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
10037 to all type variants.
10039 PR tree-optimization/84503
10040 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
10041 width as info->bitpos + info->bitsize - start.
10042 (merged_store_group::merge_overlapping): Simplify width computation.
10043 (check_no_overlap): New function.
10044 (imm_store_chain_info::try_coalesce_bswap): Compute expected
10045 start + width and last_order of the group, fail if check_no_overlap
10047 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
10048 to group if check_no_overlap fails.
10050 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10052 * config/rs6000/altivec.md: Delete contraint arguments to
10053 define_expand, define_split, and define_peephole2, and in
10054 define_insn_and_split if always unused.
10055 * config/rs6000/darwin.md: Ditto.
10056 * config/rs6000/dfp.md: Ditto.
10057 * config/rs6000/rs6000.md: Ditto.
10058 * config/rs6000/sync.md: Ditto.
10059 * config/rs6000/vector.md: Ditto.
10060 * config/rs6000/vsx.md: Ditto.
10062 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10064 * config/rs6000/altivec.md: Write output control strings as braced
10065 blocks instead of double-quoted strings.
10066 * config/rs6000/darwin.md: Ditto.
10067 * config/rs6000/rs6000.md: Ditto.
10068 * config/rs6000/vector.md: Ditto.
10069 * config/rs6000/vsx.md: Ditto.
10071 2018-02-21 Jason Merrill <jason@redhat.com>
10073 PR c++/84314 - ICE with templates and fastcall attribute.
10074 * attribs.c (build_type_attribute_qual_variant): Remove assert.
10076 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10078 * ipa-cp.c (determine_versionability): Fix comment typos.
10080 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10083 * ipa-cp.c (determine_versionability): Do not version functions caling
10086 2018-02-21 Martin Liska <mliska@suse.cz>
10089 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
10090 Add "native" as a possible value.
10091 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
10092 the macro when native cpu detection is available.
10094 2018-02-21 Martin Liska <mliska@suse.cz>
10097 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
10098 Add "native" as a possible value.
10099 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
10100 when native cpu detection is available.
10102 2018-02-21 Jakub Jelinek <jakub@redhat.com>
10103 Martin Sebor <msebor@redhat.com>
10105 PR tree-optimization/84478
10106 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
10108 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
10109 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
10110 support which is conservatively correct, for 2 only stay conservative
10111 for maxlen. Formatting and comment capitalization fixes. Add STRICT
10112 argument to the 2 argument get_range_strlen, adjust 6 arg
10113 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
10115 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
10116 (gimple_fold_builtin_strlen): Pass true as last argument to
10119 2018-02-20 Martin Sebor <msebor@redhat.com>
10121 PR middle-end/84095
10122 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
10123 (builtin_memref::set_base_and_offset): Same. Handle inner references.
10124 (builtin_memref::builtin_memref): Factor out parts into
10125 set_base_and_offset and call it.
10127 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10129 PR middle-end/84406
10130 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
10131 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
10132 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
10133 search at the associated MODE_INT.
10135 2018-02-20 Jeff Law <law@redhat.com>
10137 PR middle-end/82123
10138 PR tree-optimization/81592
10139 PR middle-end/79257
10140 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
10141 for range data rather than using global data.
10142 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
10143 range data rather than using global data.
10144 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
10145 pass it to children as needed.
10146 (struct directive::fmtresult): Similarly.
10147 (struct directive::set_width): Similarly.
10148 (struct directive::set_precision): Similarly.
10149 (format_integer, format_directive, parse_directive): Similarly.
10150 (format_none): Accept unnamed vr_values parameter.
10151 (format_percent, format_floating, format_character): Similarly.
10152 (format_string, format_plain): Similarly.
10153 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
10154 the EVRP range analyzer for range data rather than using global data.
10155 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
10156 gimple-ssa-evrp-analyze.h
10157 (class sprintf_dom_walker): Add after_dom_children member function.
10158 Add evrp_range_analyzer member.
10159 (sprintf_dom_walker::before_dom_children): Call into the EVRP
10160 range analyzer as needed.
10161 (sprintf_dom_walker::after_dom_children): New member function.
10162 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
10164 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
10165 (evrp_range_analyzer::pop_to_marker): Likewise.
10167 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10169 PR tree-optimization/84419
10170 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
10171 with the required type if its current type is compatible but
10174 2018-02-20 Jakub Jelinek <jakub@redhat.com>
10176 PR middle-end/82004
10177 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
10178 after vectorization.
10180 2018-02-20 Martin Liska <mliska@suse.cz>
10183 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
10184 possible values if we don't have a hint.
10186 2018-02-20 Martin Liska <mliska@suse.cz>
10190 * final.c (shorten_branches): Build align_tab array with one
10192 * opts.c (finish_options): Add alignment option limit check.
10193 (MAX_CODE_ALIGN): Likewise.
10194 (MAX_CODE_ALIGN_VALUE): Likewise.
10195 * doc/invoke.texi: Document maximum allowed option value for
10196 all -falign-* options.
10198 2018-02-19 Jakub Jelinek <jakub@redhat.com>
10201 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
10202 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
10203 * var-tracking.c (emit_note_insn_var_location): Remove all references
10204 to NOTE_INSN_CALL_ARG_LOCATION.
10205 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
10206 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
10207 Use copy_rtx_if_shared.
10208 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
10209 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
10210 (dwarf2out_var_location): Remove handling of
10211 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
10213 * final.c (final_scan_insn): Remove all references to
10214 NOTE_INSN_CALL_ARG_LOCATION.
10215 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
10216 before dumping final insns.
10217 * except.c (emit_note_eh_region_end): Remove all references to
10218 NOTE_INSN_CALL_ARG_LOCATION.
10219 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
10220 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
10221 * config/arc/arc.c (hwloop_optimize): Likewise.
10222 * config/arm/arm.c (create_fix_barrier): Likewise.
10223 * config/s390/s390.c (s390_chunkify_start): Likewise.
10224 * config/sh/sh.c (find_barrier): Likewise.
10225 * config/i386/i386.c (rest_of_insert_endbranch,
10226 ix86_seh_fixup_eh_fallthru): Likewise.
10227 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
10228 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
10229 * config/frv/frv.c (frv_function_prologue): Likewise.
10230 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
10232 (note_outside_basic_block_p): Remove all references to
10233 NOTE_INSN_CALL_ARG_LOCATION.
10234 * gengtype.c (adjust_field_rtx_def): Likewise.
10235 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
10237 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
10238 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
10241 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
10244 PR tree-optimization/84452
10245 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
10246 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
10249 2018-02-19 Martin Liska <mliska@suse.cz>
10252 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
10254 2018-02-19 Martin Liska <mliska@suse.cz>
10255 Richard Sandiford <richard.sandiford@linaro.org>
10257 PR tree-optimization/82491
10258 * gimple-fold.c (get_base_constructor): Make earlier bail out
10261 2018-02-19 Carl Love <cel@us.ibm.com>
10263 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
10264 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
10266 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
10267 P8V_BUILTIN_VEC_NEG.
10269 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
10271 * config/rl78/rl78.md (movdf): New define expand.
10273 2018-02-19 Martin Liska <mliska@suse.cz>
10276 * doc/invoke.texi: Fix typo.
10277 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
10279 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
10281 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
10282 handle rs6000_single_float and rs6000_double_float specially for
10285 2018-02-16 Jeff Law <law@redhat.com>
10287 * config/rx/rx.c (add_pop_cfi_notes): New function.;
10288 (pop_regs): Use it.
10290 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10293 * ipa-inline.c (inline_small_functions): Fix a typo.
10295 2018-02-16 Nathan Sidwell <nathan@acm.org>
10297 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
10299 2018-02-16 Carl Love <cel@us.ibm.com>
10301 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
10302 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
10303 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
10304 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
10305 expansion to P8V_BUILTIN_VEC_FLOAT2.
10307 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
10309 PR rtl-optimization/70023
10310 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
10311 src_regno into account.
10313 2018-02-16 Carl Love <cel@us.ibm.com>
10315 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
10316 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
10317 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
10318 * config/rs6000/rs6000.c: Remove case statements for
10319 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
10320 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
10321 and P9V_BUILTIN_VEC_VINSERT4B.
10322 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
10323 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
10324 * config/rs6000/vsx.md:
10325 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
10328 2018-02-16 Carl Love <cel@us.ibm.com>
10330 * config/rs6000/altivec.h: Add builtin names vec_extract4b
10332 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
10334 * config/rs6000/rs6000-c.c: Add the definitions for
10335 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
10336 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
10337 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
10338 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
10339 definition for insert4b and define insn *insert3b_internal.
10340 * doc/extend.texi: Add documentation for vec_extract4b.
10342 2018-02-16 Nathan Sidwell <nathan@acm.org>
10344 * doc/extend.texi (Backwards Compatibility): Mention friend
10345 injection. Note for-scope is deprecated.
10346 * doc/invoke.texi (-ffriend-injection): Deprecate.
10348 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
10350 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
10351 that moved to I2, also allow destinations that are a paradoxical
10352 subreg (instead of a normal reg).
10354 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
10357 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
10360 2018-02-16 Richard Biener <rguenther@suse.de>
10362 PR tree-optimization/84037
10363 PR tree-optimization/84016
10365 * config/i386/i386.c (ix86_builtin_vectorization_cost):
10366 Adjust vec_construct for the fact we need additional higher latency
10367 128bit inserts for AVX256 and AVX512 vector builds.
10368 (ix86_add_stmt_cost): Scale vector construction cost for
10371 2018-02-16 Richard Biener <rguenther@suse.de>
10373 PR tree-optimization/84417
10374 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
10375 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
10376 (non_rewritable_lvalue_p): Likewise, use poly-ints.
10378 2018-02-16 Martin Liska <mliska@suse.cz>
10381 * internal-fn.def (ASAN_CHECK): Set proper flags.
10382 (ASAN_MARK): Likewise.
10384 2018-02-16 Julia Koval <julia.koval@intel.com>
10386 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
10387 from PTA_CANNONLAKE.
10389 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10392 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
10393 Use ++iter rather than iter++ for std::list iterators.
10394 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
10395 defer deleting them until all nodes in the forest are processed. Do
10396 free even leaf nodes. Change to_process into auto_vec.
10399 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
10400 * vec.h (vec_default_construct): Use memset instead of placement new
10401 if BROKEN_VALUE_INITIALIZATION is defined.
10402 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
10403 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
10406 PR rtl-optimization/83723
10407 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
10408 * lra.c (lra_substitute_pseudo): Likewise. If true, use
10409 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
10411 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
10413 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
10415 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
10417 PR rtl-optimization/81443
10418 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
10419 from inner REGs to paradoxical SUBREGs.
10421 2018-02-16 Richard Biener <rguenther@suse.de>
10423 PR tree-optimization/84399
10424 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10425 For operands we can analyze at their definition make sure we can
10426 analyze them at each use as well.
10428 2018-02-16 Richard Biener <rguenther@suse.de>
10430 PR tree-optimization/84190
10431 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
10432 volatile accesses if the decl isn't volatile.
10434 2018-02-15 Jason Merrill <jason@redhat.com>
10436 PR c++/84314 - ICE with templates and fastcall attribute.
10437 * attribs.c (build_type_attribute_qual_variant): Don't clobber
10438 TYPE_CANONICAL on an existing type.
10440 2018-02-15 Jakub Jelinek <jakub@redhat.com>
10442 PR tree-optimization/84383
10443 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
10444 dstoff nor call operand_equal_p if dstbase is NULL.
10446 PR tree-optimization/84334
10447 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
10448 also a CONSTANT_CLASS_P, punt.
10450 2018-02-14 Jim Wilson <jimw@sifive.com>
10452 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
10453 first SMALL_OPERAND check. New local min_second_step. Move assert
10454 to where locals are set. Add TARGET_RVC support.
10455 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
10457 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
10459 * doc/invoke.texi: Correct -Wformat-overflow code sample.
10461 2018-02-14 Martin Sebor <msebor@redhat.com>
10463 PR tree-optimization/83698
10464 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
10465 arrays constrain the offset range to their bounds.
10466 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
10467 (builtin_access::overlap): Avoid setting the size of overlap if it's
10469 (maybe_diag_overlap): Also consider arrays when deciding what values
10470 of offsets to include in diagnostics.
10472 2018-02-14 Martin Sebor <msebor@redhat.com>
10475 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
10476 that correspond to the kind of a declaration.
10478 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
10481 * config/pa/pa.md: Load address of PIC label using the linkage table
10482 if the label is nonlocal.
10484 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
10486 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
10487 warning message if user requests -maltivec=be.
10488 * doc/invoke.texi: Document deprecation of -maltivec=be.
10490 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
10493 * config/rs6000/rs6000-c.c: Update definitions for
10494 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
10495 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
10497 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10500 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
10501 add _get_ssp intrinsics. Remove argument from
10502 __builtin_ia32_rdssp[d|q].
10503 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
10504 * config/i386/i386-builtin.def: Remove argument from
10505 __builtin_ia32_rdssp[d|q].
10506 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
10507 ix86_expand_special_args_builtin for _rdssp[d|q].
10508 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
10509 Clear register before usage.
10510 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
10511 Add documentation for new _get_ssp and _inc_ssp intrinsics.
10513 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
10515 PR tree-optimization/84357
10516 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
10517 operand 1 of an ARRAY_REF too.
10519 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
10522 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
10523 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10525 (set_of_reg): New struct.
10526 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
10527 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
10528 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10530 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
10531 Split into bitclr, bitset, bitinvert patterns if appropriate.
10532 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
10533 use rx_fuse_in_memory_bitop.
10534 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
10535 to named insn, correct maximum insn length.
10537 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
10540 * machmode.def: Define a complex mode for PARTIAL_INT.
10541 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
10543 * doc/rtl.texi: Document CSPImode.
10544 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
10546 (msp430_hard_regno_nregs_with_padding): Likewise.
10548 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
10551 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
10553 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
10555 PR rtl-optimization/84169
10556 * combine.c (try_combine): New variable split_i2i3. Set it to true if
10557 we generated a parallel as new i3 and we split that to new i2 and i3
10558 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
10559 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
10560 those to i2, not i1. Partially rewrite this scan code.
10562 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10565 * stor-layout.c (place_field): For variable length fields, adjust
10566 offset_align afterwards not just based on the field's alignment,
10567 but also on the size.
10569 PR middle-end/84309
10570 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
10571 of exps and logs in the use_exp2 case.
10573 2018-02-13 Jeff Law <law@redhat.com>
10575 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
10576 entry for "vector".
10578 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
10581 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
10585 * common.opt (gas-loc-support, gas-locview-support): New.
10586 (ginline-points, ginternal-reset-location-views): New.
10587 * doc/invoke.texi: Document them. Use @itemx where intended.
10588 (gvariable-location-views): Adjust.
10589 * target.def (reset_location_view): New.
10590 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
10591 (TARGET_RESET_LOCATION_VIEW): New.
10592 * doc/tm.texi: Rebuilt.
10593 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
10594 (dwarf2out_default_as_locview_support): New.
10595 (output_asm_line_debug_info): Use option variables.
10596 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
10597 (output_loc_list): Likewise.
10598 (add_high_low_attributes): Check option variables.
10599 Don't output entry view attribute in strict mode.
10600 (gen_inlined_subroutine_die): Check option variables.
10601 (dwarf2out_inline_entry): Likewise.
10602 (init_sections_and_labels): Likewise.
10603 (dwarf2out_early_finish): Likewise.
10604 (maybe_reset_location_view): New, from...
10605 (dwarf2out_var_location): ... here. Call it.
10606 * debug.h (dwarf2out_default_as_loc_support): Declare.
10607 (dwarf2out_default_as_locview_support): Declare.
10608 * hooks.c (hook_int_rtx_insn_0): New.
10609 * hooks.h (hook_int_rtx_insn_0): Declare.
10610 * toplev.c (process_options): Take -gas-loc-support and
10611 -gas-locview-support from dwarf2out. Enable
10612 -gvariable-location-views by default only with locview
10613 assembler support. Enable -ginternal-reset-location-views by
10614 default only if the target defines the corresponding hook.
10615 Enable -ginline-points by default if location views are
10616 enabled; force it disabled if statement frontiers are
10618 * tree-inline.c (expand_call_inline): Check option variables.
10619 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
10621 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
10623 PR tree-optimization/84321
10624 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
10625 handling. Also check whether the anti-range contains any values
10626 that satisfy the mask; switch to a VR_RANGE if not.
10628 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
10631 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
10633 2018-02-13 Martin Jambor <mjambor@suse.cz>
10636 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
10637 of call statements, also set location of a load to a temporary.
10639 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
10641 * config/rl78/rl78.c (add_vector_labels): New function.
10642 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
10643 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
10644 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
10645 which checks that no arguments are passed.
10646 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
10647 * doc/extend.texi: Documentation for the new attribute.
10649 2018-02-13 Andreas Schwab <schwab@suse.de>
10651 * config/riscv/linux.h (CPP_SPEC): Define.
10653 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10656 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
10657 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
10658 OPTION_MASK_ISA_AES as first argument to def_builtin_const
10659 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
10660 instead of OPTION_MASK_ISA_PCLMUL as first argument to
10661 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
10662 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
10663 temporarily for AES and PCLMUL builtins.
10665 PR tree-optimization/84339
10666 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
10667 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
10670 PR middle-end/84309
10671 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
10672 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
10673 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
10675 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
10677 * omp-simd-clone.h: New file.
10678 * omp-simd-clone.c: Include omp-simd-clone.h.
10679 (expand_simd_clones): No longer static.
10680 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
10681 cgraph.h and omp-simd-clone.h.
10682 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
10683 (vect_recog_widen_shift_pattern): Formatting fix.
10684 (vect_pattern_recog_1): Don't check optab for calls.
10687 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
10688 operands[2] into a REG before using gen_lowpart on it.
10690 2018-02-12 Jeff Law <law@redhat.com>
10693 * config/sh/sh.c (find_barrier): Consider a sibling call
10696 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
10697 successfully back substituting a reg.
10699 2018-02-12 Richard Biener <rguenther@suse.de>
10701 PR tree-optimization/84037
10702 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
10703 parameter, move visited init to caller.
10704 (vect_slp_analyze_operations): Separate cost from validity
10705 check, initialize visited once for all instances.
10706 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
10708 * tree-vect-stmts.c (vect_model_simple_cost): Make early
10710 (vect_model_promotion_demotion_cost): Likewise.
10711 (vectorizable_bswap): Guard cost modeling with !slp_node
10712 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
10714 (vectorizable_call): Likewise.
10715 (vectorizable_conversion): Likewise.
10716 (vectorizable_assignment): Likewise.
10717 (vectorizable_shift): Likewise.
10718 (vectorizable_operation): Likewise.
10719 (vectorizable_store): Likewise.
10720 (vectorizable_load): Likewise.
10721 (vectorizable_condition): Likewise.
10722 (vectorizable_comparison): Likewise.
10724 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
10727 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
10728 (ASAN_MARK): Fix fnspec to account for return value, change pointer
10729 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
10731 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10733 PR middle-end/83665
10734 * params.def (inline-min-speedup): Increase from 8 to 15.
10735 (max-inline-insns-auto): Decrease from 40 to 30.
10736 * ipa-split.c (consider_split): Add some buffer for function to
10737 be considered inlining candidate.
10738 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
10741 2018-02-12 Richard Biener <rguenther@suse.de>
10743 PR tree-optimization/84037
10744 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
10745 matched stmts if we cannot swap the non-matched ones.
10747 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10749 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
10750 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
10751 _mm_maskz_scalef_round_ss): New intrinsics.
10752 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
10753 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
10754 __builtin_ia32_scalefss_round): Remove.
10755 (__builtin_ia32_scalefsd_mask_round,
10756 __builtin_ia32_scalefss_mask_round): New intrinsics.
10757 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
10758 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
10759 ((match_operand:VF_128 2 "<round_nimm_predicate>"
10760 "<round_constraint>")): Changed to ...
10761 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
10762 "<round_scalar_constraint>")): ... this.
10763 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
10764 %0, %1, %2<round_op3>}"): Changed to ...
10765 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
10766 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
10767 %2<round_scalar_mask_op3>}"): ... this.
10768 * config/i386/subst.md (round_scalar_nimm_predicate): New.
10770 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10772 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
10773 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
10774 (_mm_maskz_sqrt_round_ss): New intrinsics.
10775 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
10776 (__builtin_ia32_sqrtsd_mask_round)
10777 (__builtin_ia32_sqrtss_mask_round): New builtins.
10778 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
10779 (__builtin_ia32_sqrtss_round): Remove.
10780 (__builtin_ia32_sqrtsd_mask_round)
10781 (__builtin_ia32_sqrtss_mask_round): New builtins.
10782 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
10783 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
10784 ((match_operand:VF_128 1 "vector_operand"
10785 "xBm,<round_constraint>")): Changed to ...
10786 ((match_operand:VF_128 1 "vector_operand"
10787 "xBm,<round_scalar_constraint>")): ... this.
10788 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
10789 %0, %2, %<iptr>1<round_op3>}): Changed to ...
10790 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
10791 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
10792 %<iptr>1<round_scalar_mask_op3>}): ... this.
10793 ((set_attr "prefix" "<round_prefix>")): Changed to ...
10794 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
10796 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
10799 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
10800 Cast vec_cmpeq result to correct type.
10801 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
10802 Cast vec_cmpgt result to correct type.
10804 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
10806 * final.c (final_scan_insn_1): Renamed from...
10807 (final_scan_insn): ... this. New wrapper, to recover
10808 seen from the outermost call in recursive ones.
10809 * config/sparc/sparc.c (output_return): Drop seen from call.
10810 (output_sibcall): Likewise.
10811 * config/visium/visium.c (output_branch): Likewise.
10813 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
10815 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
10818 2018-02-10 Alan Modra <amodra@gmail.com>
10821 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
10822 Specify LR as an input.
10824 2018-02-10 Jakub Jelinek <jakub@redhat.com>
10827 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
10828 remove_member_access_dummy_vars): New functions.
10829 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
10830 lower_omp_1, execute_lower_omp): Use them.
10832 PR rtl-optimization/84308
10833 * shrink-wrap.c (spread_components): Release todo vector.
10835 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
10837 PR rtl-optimization/57193
10838 * ira-color.c (struct allocno_color_data): Add member
10839 conflict_allocno_hard_prefs.
10840 (update_conflict_allocno_hard_prefs): New.
10841 (bucket_allocno_compare_func): Add a preference based on
10842 conflict_allocno_hard_prefs.
10843 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
10844 (color_allocnos): Remove a dead code. Initiate
10845 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
10847 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10850 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
10851 constraint from =wa to wa. Avoid a subreg on the output operand,
10852 instead use a pseudo and subreg it in a move.
10853 (p9_xxbrd_<mode>): Changed to ...
10854 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
10855 (p9_xxbrd_v2df): New expander.
10856 (p9_xxbrw_<mode>): Changed to ...
10857 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
10858 (p9_xxbrw_v4sf): New expander.
10860 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10862 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
10864 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
10867 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
10868 multiply in 32-bit mode.
10869 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
10870 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
10873 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10875 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
10876 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
10877 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
10878 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
10880 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
10883 * dwarf2out.c (is_trivial_indirect_ref): New function.
10884 (dwarf2out_late_global_decl): Do not generate a location
10885 attribute for variables that have a non-trivial DECL_VALUE_EXPR
10886 and that are not defined in the current unit.
10888 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10890 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
10891 instead of a libcall for UNORDERED.
10893 2018-02-09 Tamar Christina <tamar.christina@arm.com>
10896 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
10897 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
10899 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10902 * config/s390/s390.c (s390_set_current_function): Invoke
10903 s390_indirect_branch_settings also if fndecl didn't change.
10905 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10907 * config/rs6000/rs6000.md (blockage): Set length to zero.
10909 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10911 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
10913 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10916 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
10917 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
10921 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
10922 PARALLEL incoming that failed vt_get_decl_and_offset check.
10924 PR middle-end/84237
10925 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
10926 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
10928 (get_variable_section): For decls in named .bss* sections pass true as
10929 second argument to bss_initializer_p.
10931 2018-02-09 Marek Polacek <polacek@redhat.com>
10932 Jakub Jelinek <jakub@redhat.com>
10935 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
10936 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
10937 Sync some changes from cxx_fold_indirect_ref.
10939 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10941 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
10943 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
10944 (BLOCK_INLINE_ENTRY_LABEL): New.
10945 (dwarf2out_var_location): Disregard inline entry markers.
10946 (inline_entry_data): New struct.
10947 (inline_entry_data_hasher): New hashtable type.
10948 (inline_entry_data_hasher::hash): New.
10949 (inline_entry_data_hasher::equal): New.
10950 (inline_entry_data_table): New variable.
10951 (add_high_low_attributes): Add DW_AT_entry_pc and
10952 DW_AT_GNU_entry_view attributes if a pending entry is found
10953 in inline_entry_data_table. Add old entry_pc attribute only
10954 if debug nonbinding markers are disabled.
10955 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
10956 markers are enabled.
10957 (block_within_block_p, dwarf2out_inline_entry): New.
10958 (dwarf2out_finish): Check that no entries remained in
10959 inline_entry_data_table.
10960 * final.c (reemit_insn_block_notes): Handle inline entry notes.
10961 (final_scan_insn, notice_source_line): Likewise.
10962 (rest_of_clean_state): Skip inline entry markers.
10963 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
10965 * gimple.c (gimple_build_debug_inline_entry): New.
10966 * gimple.h (enum gimple_debug_subcode): Add
10967 GIMPLE_DEBUG_INLINE_ENTRY.
10968 (gimple_build_debug_inline_entry): Declare.
10969 (gimple_debug_inline_entry_p): New.
10970 (gimple_debug_nonbind_marker_p): Adjust.
10971 * insn-notes.def (INLINE_ENTRY): New.
10972 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
10973 inline entry marker notes.
10974 (print_insn): Likewise.
10975 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
10976 (INSN_DEBUG_MARKER_KIND): Likewise.
10977 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
10978 * tree-inline.c (expand_call_inline): Build and insert
10979 debug_inline_entry stmt.
10980 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
10981 inline entry blocks early, if nonbind markers are enabled.
10982 (dump_scope_block): Dump fragment info.
10983 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
10984 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
10985 (gimple_build_debug_inline_entry): New.
10986 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
10987 Enable/disable inline entry points too.
10988 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
10989 (DEBUG_INSN): Describe inline entry markers.
10991 * common.opt (gvariable-location-views): New.
10992 (gvariable-location-views=incompat5): New.
10993 * config.in: Rebuilt.
10994 * configure: Rebuilt.
10995 * configure.ac: Test assembler for view support.
10996 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
10997 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
10998 * dwarf2out.c (var_loc_view): New typedef.
10999 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
11000 (dwarf2out_locviews_in_attribute): New.
11001 (dwarf2out_locviews_in_loclist): New.
11002 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
11003 (enum dw_line_info_opcode): Add LI_adv_address.
11004 (struct dw_line_info_table): Add view.
11005 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
11006 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
11007 (zero_view_p): New variable.
11008 (ZERO_VIEW_P): New macro.
11009 (output_asm_line_debug_info): New.
11010 (struct var_loc_node): Add view.
11011 (add_AT_view_list, AT_loc_list): New.
11012 (add_var_loc_to_decl): Add view param. Test it against last.
11013 (new_loc_list): Add view params. Record them.
11014 (AT_loc_list_ptr): Handle loc and view lists.
11015 (view_list_to_loc_list_val_node): New.
11016 (print_dw_val): Handle dw_val_class_view_list.
11017 (size_of_die): Likewise.
11018 (value_format): Likewise.
11019 (loc_list_has_views): New.
11020 (gen_llsym): Set vl_symbol too.
11021 (maybe_gen_llsym, skip_loc_list_entry): New.
11022 (dwarf2out_maybe_output_loclist_view_pair): New.
11023 (output_loc_list): Output view list or entries too.
11024 (output_view_list_offset): New.
11025 (output_die): Handle dw_val_class_view_list.
11026 (output_dwarf_version): New.
11027 (output_compilation_unit_header): Use it.
11028 (output_skeleton_debug_sections): Likewise.
11029 (output_rnglists, output_line_info): Likewise.
11030 (output_pubnames, output_aranges): Update version comments.
11031 (output_one_line_info_table): Output view numbers in asm comments.
11032 (dw_loc_list): Determine current endview, pass it to new_loc_list.
11033 Call maybe_gen_llsym.
11034 (loc_list_from_tree_1): Adjust.
11035 (add_AT_location_description): Create view list attribute if
11036 needed, check it's absent otherwise.
11037 (convert_cfa_to_fb_loc_list): Adjust.
11038 (maybe_emit_file): Call output_asm_line_debug_info for test.
11039 (dwarf2out_var_location): Reset views as needed. Precompute
11040 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
11041 attribute. Set view.
11042 (new_line_info_table): Reset next view.
11043 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
11044 (dwarf2out_source_line): Likewise. Output view resets and labels to
11045 the assembler, or select appropriate line info opcodes.
11046 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
11047 (optimize_string_length): Catch it. Adjust.
11048 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
11049 dw_val_class_view_list, and remove it if no longer needed.
11050 (hash_loc_list): Hash view numbers.
11051 (loc_list_hasher::equal): Compare them.
11052 (optimize_location_lists): Check whether a view list symbol is
11053 needed, and whether the locview attribute is present, and
11054 whether they match. Remove the locview attribute if no longer
11056 (index_location_lists): Call skip_loc_list_entry for test.
11057 (dwarf2out_finish): Call output_asm_line_debug_info for test.
11058 Use output_dwarf_version.
11059 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
11060 (struct dw_val_node): Add val_view_list.
11061 * final.c (SEEN_NEXT_VIEW): New.
11062 (set_next_view_needed): New.
11063 (clear_next_view_needed): New.
11064 (maybe_output_next_view): New.
11065 (final_start_function): Rename to...
11066 (final_start_function_1): ... this. Take pointer to FIRST,
11067 add SEEN parameter. Emit param bindings in the initial view.
11068 (final_start_function): Reintroduce SEEN-less interface.
11069 (final): Rename to...
11070 (final_1): ... this. Take SEEN parameter. Output final pending
11071 next view at the end.
11072 (final): Reintroduce seen-less interface.
11073 (final_scan_insn): Output pending next view before switching
11074 sections or ending a block. Mark the next view as needed when
11075 outputting variable locations. Notify debug backend of section
11076 changes, and of location view changes.
11077 (rest_of_handle_final): Adjust.
11078 * toplev.c (process_options): Autodetect value for debug variable
11079 location views option. Warn on incompat5 without -gdwarf-5.
11080 * doc/invoke.texi (gvariable-location-views): New.
11081 (gvariable-location-views=incompat5): New.
11082 (gno-variable-location-views): New.
11084 2018-02-08 David Malcolm <dmalcolm@redhat.com>
11086 PR tree-optimization/84136
11087 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
11088 that the result of find_edge is non-NULL.
11090 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11093 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
11094 storing integer register in SImode. Fix cost of 256 and 512
11095 byte aligned SSE register store.
11097 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11099 * config/i386/i386.c (ix86_multiplication_cost): Fix
11100 multiplication cost for TARGET_AVX512DQ.
11102 2018-02-08 Marek Polacek <polacek@redhat.com>
11104 PR tree-optimization/84238
11105 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
11108 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11110 PR tree-optimization/84265
11111 * tree-vect-stmts.c (vectorizable_store): Don't treat
11112 VMAT_CONTIGUOUS accesses as grouped.
11113 (vectorizable_load): Likewise.
11115 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11117 PR tree-optimization/81635
11118 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
11119 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
11120 (test_round_for_mask): New functions.
11121 (wide_int_cc_tests): Call test_round_for_mask.
11122 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
11123 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
11124 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
11125 range returned by get_range_info.
11127 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11130 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
11131 * symtab.c: Include builtins.h
11132 (symtab_node::output_to_lto_symbol_table_p): Move here
11133 from lto-streamer-out.c:output_symbol_p.
11134 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
11135 (output_symbol_p): Move all logic to symtab.c
11136 (produce_symtab): Update.
11138 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11140 * config/s390/s390-opts.h (enum indirect_branch): Define.
11141 * config/s390/s390-protos.h (s390_return_addr_from_memory)
11142 (s390_indirect_branch_via_thunk)
11143 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
11144 (enum s390_indirect_branch_type): Define.
11145 * config/s390/s390.c (struct s390_frame_layout, struct
11146 machine_function): Remove.
11147 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
11148 (indirect_branch_table_label_no, indirect_branch_table_name):
11150 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
11151 (enum s390_indirect_branch_option): Define.
11152 (s390_return_addr_from_memory): New function.
11153 (s390_handle_string_attribute): New function.
11154 (s390_attribute_table): Add new attribute handler.
11155 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
11156 (s390_indirect_branch_via_thunk): New function.
11157 (s390_indirect_branch_via_inline_thunk): New function.
11158 (s390_function_ok_for_sibcall): When jumping via thunk disallow
11159 sibling call optimization for non z10 compiles.
11160 (s390_emit_call): Force indirect branch target to be a single
11161 register. Add r1 clobber for non-z10 compiles.
11162 (s390_emit_epilogue): Emit return jump via return_use expander.
11163 (s390_reorg): Handle JUMP_INSNs as execute targets.
11164 (s390_option_override_internal): Perform validity checks for the
11165 new command line options.
11166 (s390_indirect_branch_attrvalue): New function.
11167 (s390_indirect_branch_settings): New function.
11168 (s390_set_current_function): Invoke s390_indirect_branch_settings.
11169 (s390_output_indirect_thunk_function): New function.
11170 (s390_code_end): Implement target hook.
11171 (s390_case_values_threshold): Implement target hook.
11172 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
11174 * config/s390/s390.h (struct s390_frame_layout)
11175 (struct machine_function): Move here from s390.c.
11176 (TARGET_INDIRECT_BRANCH_NOBP_RET)
11177 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
11178 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
11179 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
11180 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
11181 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
11182 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
11183 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
11184 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
11185 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
11186 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
11187 (mnemonic attribute): Add values which aren't recognized
11189 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
11190 pattern for branch conversion. Fix mnemonic attribute.
11191 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
11192 indirect branch via thunk if requested.
11193 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
11194 ("*indirect_jump"): Disable for branch conversion using out of
11196 ("indirect_jump_via_thunk<mode>_z10")
11197 ("indirect_jump_via_thunk<mode>")
11198 ("indirect_jump_via_inlinethunk<mode>_z10")
11199 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
11200 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
11201 ("casesi_jump_via_inlinethunk<mode>_z10")
11202 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
11203 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
11204 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
11205 ("*indirect2_jump"): Disable for branch conversion.
11206 ("casesi_jump"): Turn into expander and expand patterns for branch
11208 ("return_use"): New expander.
11209 ("*return"): Emit return via thunk and rename it to ...
11210 ("*return<mode>"): ... this one.
11211 * config/s390/s390.opt: Add new options and and enum for the
11214 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11216 * lra-constraints.c (match_reload): Unconditionally use
11217 gen_lowpart_SUBREG, rather than selecting between that
11218 and equivalent gen_rtx_SUBREG code.
11220 2018-02-08 Richard Biener <rguenther@suse.de>
11222 PR tree-optimization/84233
11223 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
11224 changed flag instead of boguously re-using phi_inserted.
11226 2018-02-08 Martin Jambor <mjambor@suse.cz>
11228 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
11229 static local variables.
11231 2018-02-08 Richard Biener <rguenther@suse.de>
11233 PR tree-optimization/84278
11234 * tree-vect-stmts.c (vectorizable_store): When looking for
11235 smaller vector types to perform grouped strided loads/stores
11236 make sure the mode is supported by the target.
11237 (vectorizable_load): Likewise.
11239 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11241 * config/aarch64/aarch64.c (aarch64_components_for_bb):
11242 Increase LDP/STP opportunities by adding adjacent callee-saves.
11244 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11246 PR rtl-optimization/84068
11247 PR rtl-optimization/83459
11248 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
11250 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
11252 PR tree-optimization/84224
11253 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
11254 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
11255 non-zero arguments.
11257 2018-02-07 Iain Sandoe <iain@codesourcery.com>
11260 * config/rs6000/altivec.md (*restore_world): Remove LR use.
11261 * config/rs6000/predicates.md (restore_world_operation): Adjust op
11262 count, remove one USE.
11264 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
11266 * doc/install.texi (Configuration): Document the
11267 --with-long-double-format={ibm,ieee} PowerPC configuration
11271 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
11272 Convert from define_expand to be define_insn_and_split. Rework
11273 float/double/_Float128 conversions to QI/HI/SImode to work with
11274 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
11275 conversions to QI/HImode types did a store and then a load to
11276 truncate the value. For conversions to VSX registers, don't split
11277 the insn, instead emit the code directly. Use the code iterator
11278 any_fix to combine signed and unsigned conversions.
11279 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
11280 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
11281 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11282 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
11283 (fix_<mode>di2_hw): Likewise.
11284 (fixuns_<mode>di2_hw): Likewise.
11285 (fix_<mode>si2_hw): Likewise.
11286 (fixuns_<mode>si2_hw): Likewise.
11287 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
11288 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11289 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
11290 fix<uns>_trunc<SFDF:mode>si2_p8.
11291 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
11292 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
11293 (fix<uns>_<mode>_mem): Likewise.
11294 (fctiw<u>z_<mode>_mem): Likewise.
11295 (fix<uns>_<mode>_mem): Likewise.
11296 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
11297 the register allocator from doing a direct move to the GPRs to do
11298 a store, and instead use the ISA 3.0 store byte/half-word from
11299 vector register instruction. For IEEE 128-bit floating point,
11300 also optimize stores of 32-bit ints.
11301 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
11303 2018-02-07 Alan Hayward <alan.hayward@arm.com>
11305 * genextract.c (push_pathstr_operand): New function to support
11307 (walk_rtx): Call push_pathstr_operand.
11308 (print_path): Support [a-zA-Z].
11310 2018-02-07 Richard Biener <rguenther@suse.de>
11312 PR tree-optimization/84037
11313 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
11314 (cse_and_gimplify_to_preheader): Declare.
11315 (vect_get_place_in_interleaving_chain): Likewise.
11316 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11318 (_loop_vec_info::~_loop_vec_info): Delete it.
11319 (cse_and_gimplify_to_preheader): New function.
11320 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
11321 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
11322 (vectorizable_load): Likewise. For grouped stores always base
11323 the IV on the first element.
11324 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
11325 condition before gimplifying.
11327 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11329 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
11330 *DIV_EXPR and *MOD_EXPR.
11332 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
11335 * config/i386/i386.c (ix86_option_override_internal): Mask out
11336 the CF_SET bit when checking -fcf-protection.
11338 2018-02-07 Tom de Vries <tom@codesourcery.com>
11341 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
11344 2018-02-07 Richard Biener <rguenther@suse.de>
11346 PR tree-optimization/84204
11347 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
11350 PR tree-optimization/84205
11351 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
11352 special-case isl_ast_op_zdiv_r.
11354 PR tree-optimization/84223
11355 * graphite-scop-detection.c (gather_bbs::before_dom_children):
11356 Only add conditions from within the region.
11357 (gather_bbs::after_dom_children): Adjust.
11359 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
11362 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
11363 * config/avr/avr.md: Only post-reload split REG-REG moves if
11364 either register is GENERAL_REG_P.
11366 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11368 PR tree-optimization/84235
11369 * tree-ssa-scopedtables.c
11370 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
11371 if the subtraction is performed in floating point type where NaNs are
11372 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
11373 build 1. Formatting fix.
11375 2018-02-06 Jakub Jelinek <jakub@redhat.com>
11378 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
11379 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
11380 and skip it regardless of bb boundaries. Use CALL_P macro,
11381 don't test INSN_P (insn) together with CALL_P or JUMP_P check
11382 unnecessarily, formatting fix.
11384 2018-02-06 Michael Collison <michael.collison@arm.com>
11386 * config/arm/thumb2.md:
11387 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
11388 (*thumb_mov_notscc): Ditto.
11390 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
11393 * config/rs6000/rs6000.md (su code attribute): Use "u" for
11394 unsigned_fix, not "s".
11396 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11398 * configure.ac (gcc_fn_eh_frame_ro): New function.
11399 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
11400 correct .eh_frame permissions.
11401 * configure: Regenerate.
11403 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
11405 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
11406 irrelevant options.
11408 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11410 * config/rs6000/rs6000.c (rs6000_option_override_internal):
11411 Display warning message for -mno-speculate-indirect-jumps.
11413 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
11415 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
11417 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
11419 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
11421 PR tree-optimization/84225
11422 * tree-eh.c (find_trapping_overflow): Only call
11423 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
11425 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11428 * config/i386/i386.c: Reimplement the check of possible options
11429 -mibt/-mshstk conbination. Change error messages.
11430 * doc/invoke.texi: Fix a typo: remove extra '='.
11432 2018-02-06 Marek Polacek <polacek@redhat.com>
11434 PR tree-optimization/84228
11435 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
11437 2018-02-06 Tamar Christina <tamar.christina@arm.com>
11440 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
11441 emitted arch directives.
11442 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
11443 __ARM_FEATURE_COPROC before changing architectures.
11445 2018-02-06 Richard Biener <rguenther@suse.de>
11447 * config/i386/i386.c (print_reg): Fix typo.
11448 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
11450 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
11452 * configure: Regenerate.
11454 2018-02-05 Martin Sebor <msebor@redhat.com>
11456 PR tree-optimization/83369
11457 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
11460 2018-02-05 Martin Liska <mliska@suse.cz>
11462 * doc/invoke.texi: Cherry-pick upstream r323995.
11464 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
11466 * ira.c (ira_init_register_move_cost): Adjust comment.
11468 2018-02-05 Martin Liska <mliska@suse.cz>
11470 PR gcov-profile/84137
11471 * doc/gcov.texi: Fix typo in documentation.
11473 2018-02-05 Martin Liska <mliska@suse.cz>
11475 PR gcov-profile/83879
11476 * doc/gcov.texi: Document necessity of --dynamic-list-data when
11477 using dlopen functionality.
11479 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
11481 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
11482 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
11483 _mm_maskz_range_ss, _mm_mask_range_round_ss,
11484 _mm_maskz_range_round_ss): New intrinsics.
11485 (__builtin_ia32_rangesd128_round)
11486 (__builtin_ia32_rangess128_round): Remove.
11487 (__builtin_ia32_rangesd128_mask_round,
11488 __builtin_ia32_rangess128_mask_round): New builtins.
11489 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
11490 __builtin_ia32_rangess128_round): Remove.
11491 (__builtin_ia32_rangesd128_mask_round,
11492 __builtin_ia32_rangess128_mask_round): New builtins.
11493 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
11494 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
11495 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
11496 "<round_saeonly_constraint>")): Changed to ...
11497 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
11498 "<round_saeonly_scalar_constraint>")): ... this.
11499 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
11500 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
11501 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
11502 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
11503 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
11505 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
11507 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
11509 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
11510 Remove all values except native, 8540 and 8548.
11512 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
11514 * config/i386/i386.c (ix86_output_function_return): Pass
11515 INVALID_REGNUM, instead of -1, as invalid register number to
11516 indirect_thunk_name and output_indirect_thunk.
11518 2018-02-02 Julia Koval <julia.koval@intel.com>
11520 * config.gcc: Add -march=icelake.
11521 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
11522 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
11523 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
11524 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
11525 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
11526 (processor_target_table): Add icelake.
11527 (ix86_option_override_internal): Handle new PTAs.
11528 (get_builtin_code_for_version): Handle icelake.
11529 (M_INTEL_COREI7_ICELAKE): New.
11530 (fold_builtin_cpu): Handle icelake.
11531 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
11532 * doc/invoke.texi: Add -march=icelake.
11534 2018-02-02 Julia Koval <julia.koval@intel.com>
11536 * config/i386/i386.c (ix86_option_override_internal): Change flags type
11537 to wide_int_bitmask.
11538 * wide-int-bitmask.h: New.
11540 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11543 * config/i386/i386.md: Replace Pmode with word_mode in
11544 builtin_setjmp_setup and builtin_longjmp to support x32.
11546 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
11550 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
11552 (rs6000_supported_cpu_names): New static variable.
11553 (linux_cpu_translation_table): Likewise.
11554 (elf_platform) <cpu>: Define new static variable and use it.
11555 Translate kernel AT_PLATFORM name to canonical name if needed.
11556 Error if platform name is unknown.
11558 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
11561 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
11563 2018-02-01 Jeff Law <law@redhat.com>
11566 * config/i386/i386.c (release_scratch_register_on_entry): Add new
11567 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
11568 the scratch if RELEASE_VIA_POP is false.
11569 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
11570 If we have to save a temporary register, decrement SIZE appropriately.
11571 Pass new arguments to release_scratch_register_on_entry.
11572 (ix86_adjust_stack_and_probe): Likewise.
11573 (ix86_emit_probe_stack_range): Pass new arguments to
11574 release_scratch_register_on_entry.
11576 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
11578 PR rtl-optimization/84157
11579 * combine.c (change_zero_ext): Use REG_P predicate in
11580 front of HARD_REGISTER_P predicate.
11582 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
11584 * config/avr/avr.c (avr_option_override): Move disabling of
11585 -fdelete-null-pointer-checks to...
11586 * common/config/avr/avr-common.c (avr_option_optimization_table):
11589 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11591 PR tree-optimization/81635
11592 * tree-data-ref.c (split_constant_offset_1): For types that
11593 wrap on overflow, try to use range info to prove that wrapping
11596 2018-02-01 Renlin Li <renlin.li@arm.com>
11599 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
11600 TAILCALL_ADDR_REGS.
11601 (aarch64_register_move_cost): Likewise.
11602 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
11603 TAILCALL_ADDR_REGS.
11604 (REG_CLASS_NAMES): Likewise.
11605 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
11606 TAILCALL_ADDR_REGS. Remove IP registers.
11607 * config/aarch64/aarch64.md (Ucs): Update register constraint.
11609 2018-02-01 Richard Biener <rguenther@suse.de>
11611 * domwalk.h (dom_walker::dom_walker): Add additional constructor
11612 for specifying RPO order and allow NULL for that.
11613 * domwalk.c (dom_walker::dom_walker): Likewise.
11614 (dom_walker::walk): Handle NULL RPO order.
11615 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
11617 (rewrite_update_dom_walker): Likewise.
11618 (mark_def_dom_walker): Likewise.
11620 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11622 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
11623 (aarch64_maybe_expand_sve_subreg_move): Declare.
11624 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
11625 * config/aarch64/predicates.md (aarch64_any_register_operand): New
11627 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
11628 that are semantically a reverse operation.
11629 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
11630 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
11631 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
11633 (aarch64_can_change_mode_class): For big-endian, forbid changes
11634 between two SVE modes if they have different element sizes.
11636 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11638 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
11639 the TImode handling for big-endian targets.
11641 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11643 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
11644 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
11646 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
11647 Remove BSWAP handing for big-endian targets and use the form of
11648 LD1RQ appropariate for the mode.
11650 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11652 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
11653 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
11654 duplicated element.
11656 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11659 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
11660 check for operands that need to go through aarch64_sve_reload_be.
11662 2018-02-01 Jakub Jelinek <jakub@redhat.com>
11664 PR tree-optimization/81661
11665 PR tree-optimization/84117
11666 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
11667 * tree-eh.c: Include gimplify.h.
11668 (find_trapping_overflow, replace_trapping_overflow,
11669 rewrite_to_non_trapping_overflow): New functions.
11670 * tree-vect-loop.c: Include tree-eh.h.
11671 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
11672 * tree-data-ref.c: Include tree-eh.h.
11673 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
11675 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
11677 PR rtl-optimization/84123
11678 * combine.c (change_zero_ext): Check if hard register satisfies
11679 can_change_dest_mode before calling gen_lowpart_SUBREG.
11681 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
11684 * ira.c (ira_init_register_move_cost): Remove assert.
11686 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11688 PR rtl-optimization/84071
11689 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
11690 * doc/tm.texi: Regenerate.
11692 2018-01-31 Richard Biener <rguenther@suse.de>
11694 PR tree-optimization/84132
11695 * tree-data-ref.c (analyze_miv_subscript): Properly
11696 check whether evolution_function_is_affine_multivariate_p
11697 before calling gcd_of_steps_may_divide_p.
11699 2018-01-31 Julia Koval <julia.koval@intel.com>
11702 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
11703 * config/i386/i386.md (rdpid_rex64) New.
11704 (rdpid): Make 32bit only.
11706 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
11709 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
11710 an IDENTIFIER_NODE for FUNCTION_TYPE's.
11712 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11715 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11717 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11719 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11721 PR rtl-optimization/84071
11722 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
11723 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
11725 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11727 * config/arc/arc.c (arc_handle_aux_attribute): New function.
11728 (arc_attribute_table): Add 'aux' attribute.
11729 (arc_in_small_data_p): Consider aux like variables.
11730 (arc_is_aux_reg_p): New function.
11731 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
11732 (arc_get_aux_arg): New function.
11733 (prepare_move_operands): Handle aux-register access.
11734 (arc_handle_aux_attribute): New function.
11735 * doc/extend.texi (ARC Variable attributes): Add subsection.
11737 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11739 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
11740 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
11741 (arc_attribute_table): Add 'uncached' attribute.
11742 (arc_print_operand): Print '.di' flag for uncached memory
11744 (arc_in_small_data_p): Do not consider for small data the uncached
11746 (arc_is_uncached_mem_p): New function.
11747 * config/arc/predicates.md (compact_store_memory_operand): Check
11748 for uncached memory accesses.
11749 (nonvol_nonimm_operand): Likewise.
11750 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
11752 2018-01-31 Jakub Jelinek <jakub@redhat.com>
11755 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
11756 falign-loops=): Add Optimization flag.
11758 2018-01-30 Jeff Law <law@redhat.com>
11761 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
11762 INT_REGISTERS_SAVED. Check it prior to calling
11763 get_scratch_register_on_entry.
11764 (ix86_adjust_stack_and_probe): Similarly.
11765 (ix86_emit_probe_stack_range): Similarly.
11766 (ix86_expand_prologue): Corresponding changes.
11768 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11771 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
11772 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
11774 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
11777 * lra-constraints.c (curr_insn_transform): Process AND in the
11780 2018-01-30 Jakub Jelinek <jakub@redhat.com>
11782 PR rtl-optimization/83986
11783 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
11784 dependence against last_pending_memory_flush in addition to
11785 pending_jump_insns.
11787 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
11789 PR tree-optimization/81611
11790 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
11793 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11796 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
11799 2018-01-30 Richard Biener <rguenther@suse.de>
11800 Jakub Jelinek <jakub@redhat.com>
11802 PR tree-optimization/84111
11803 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
11804 inner loops added during recursion, as they don't have up-to-date
11807 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11810 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
11811 (can_inline_edge_by_limits_p): ... here.
11812 (can_early_inline_edge_p, check_callers,
11813 update_caller_keys, update_callee_keys, recursive_inlining,
11814 add_new_edges_to_heap, speculation_useful_p,
11815 inline_small_functions,
11816 inline_small_functions, flatten_function,
11817 inline_to_all_callers_1): Update.
11819 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11821 * profile-count.c (profile_count::combine_with_ipa_count): Handle
11824 2018-01-30 Richard Biener <rguenther@suse.de>
11826 PR tree-optimization/83008
11827 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
11828 invariant and constant vector uses in stmts when they need
11829 more than one stmt.
11831 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11834 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
11835 * configure: Regenerate.
11837 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11839 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
11841 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
11842 Use gen_rtx_REG rather than gen_lowpart.
11844 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11846 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
11847 rather than 0 when creating partial subregs.
11849 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11851 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
11854 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
11857 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
11858 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
11859 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
11860 flags. This restores the settings used before the 2017-07-24.
11861 Turning off pre increment/decrement/modify allows IVOPTS to
11862 optimize DF/SF loops where the index is an int.
11864 2018-01-29 Richard Biener <rguenther@suse.de>
11865 Kelvin Nilsen <kelvin@gcc.gnu.org>
11868 * tree-vect-stmts.c (vectorizable_call): Don't call
11869 targetm.vectorize_builtin_md_vectorized_function if callee is
11872 2018-01-22 Carl Love <cel@us.ibm.com>
11874 * doc/extend.tex: Fix typo in second arg in
11875 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
11877 2018-01-29 Richard Biener <rguenther@suse.de>
11879 PR tree-optimization/84086
11880 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
11881 (flush_ssaname_freelist): When SSA names were released reset
11882 the SCEV hash table.
11884 2018-01-29 Richard Biener <rguenther@suse.de>
11886 PR tree-optimization/84057
11887 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
11888 removed paths when removing edges.
11890 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
11892 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
11893 -mfunction-return=@var{choice}.
11895 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
11897 PR diagnostic/84034
11898 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
11899 Handle CR like TAB.
11900 (layout::print_source_line): Likewise.
11901 (test_get_line_width_without_trailing_whitespace): Add test cases.
11903 2018-01-27 Jakub Jelinek <jakub@redhat.com>
11905 PR middle-end/84040
11906 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
11909 2018-01-26 Jim Wilson <jimw@sifive.com>
11911 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
11913 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
11916 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11918 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
11919 and CMP + SUB-immediate -> SUBS.
11921 2018-01-26 Martin Sebor <msebor@redhat.com>
11923 PR tree-optimization/83896
11924 * tree-ssa-strlen.c (get_string_len): Rename...
11925 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
11926 Avoid assuming length is constant.
11927 (handle_char_store): Use HOST_WIDE_INT for string length.
11929 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
11932 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
11933 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
11935 2018-01-26 Richard Biener <rguenther@suse.de>
11937 PR rtl-optimization/84003
11938 * dse.c (record_store): Only record redundant stores when
11939 the earlier store aliases at least all accesses the later one does.
11941 2018-01-26 Jakub Jelinek <jakub@redhat.com>
11943 PR rtl-optimization/83985
11944 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
11945 REG_CFA_RESTORE insns.
11946 (delete_unmarked_insns): Don't ignore separate shrink wrapping
11947 REG_CFA_RESTORE insns here.
11950 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
11951 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
11953 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11955 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
11956 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
11957 (arc_init): Likewise.
11958 (arc_override_options): Likewise.
11959 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
11961 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
11963 * config/arc/arc.h (TARGET_DBNZ): Define.
11964 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
11965 properly set the tune attribute.
11966 (dbnz): Use TARGET_DBNZ guard.
11967 * config/arc/arc.opt (mtune): Add core3 option.
11969 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11971 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
11972 recognize new pic like addresses.
11973 (arc_delegitimize_address): Clean up.
11975 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11977 * config/arc/arc-arches.def: Option mrf16 valid for all
11979 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
11980 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
11981 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
11982 * config/arc/arc-tables.opt: Regenerate.
11983 * config/arc/arc.c (arc_conditional_register_usage): Handle
11984 reduced register file case.
11985 (arc_file_start): Set must have build attributes.
11986 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
11987 mrf16 option value.
11988 * config/arc/arc.opt (mrf16): Add new option.
11989 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
11990 * config/arc/genmultilib.awk: Handle new mrf16 option.
11991 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
11992 * config/arc/t-multilib: Regenerate.
11993 * doc/invoke.texi (ARC Options): Document mrf16 option.
11995 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11997 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
11998 * config/arc/arc.c (arc_handle_secure_attribute): New function.
11999 (arc_attribute_table): Add 'secure_call' attribute.
12000 (arc_print_operand): Print secure call operand.
12001 (arc_function_ok_for_sibcall): Don't optimize tail calls when
12003 (arc_is_secure_call_p): New function. * config/arc/arc.md
12004 (call_i): Add support for sjli instruction.
12005 (call_value_i): Likewise.
12006 * config/arc/constraints.md (Csc): New constraint.
12008 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12009 John Eric Martin <John.Martin@emmicro-us.com>
12011 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
12012 * config/arc/arc.c (_arc_jli_section): New struct.
12013 (arc_jli_section): New type.
12014 (rc_jli_sections): New static variable.
12015 (arc_handle_jli_attribute): New function.
12016 (arc_attribute_table): Add jli_always and jli_fixed attribute.
12017 (arc_file_end): New function.
12018 (TARGET_ASM_FILE_END): Define.
12019 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
12020 (arc_add_jli_section): New function.
12021 (jli_call_scan): Likewise.
12022 (arc_reorg): Call jli_call_scan.
12023 (arc_output_addsi): Remove 'S' from printing asm operand.
12024 (arc_is_jli_call_p): New function.
12025 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
12027 (movhi_insn): Likewise.
12028 (movsi_insn): Likewise.
12029 (movsi_set_cc_insn): Likewise.
12030 (loadqi_update): Likewise.
12031 (load_zeroextendqisi_update): Likewise.
12032 (load_signextendqisi_update): Likewise.
12033 (loadhi_update): Likewise.
12034 (load_zeroextendhisi_update): Likewise.
12035 (load_signextendhisi_update): Likewise.
12036 (loadsi_update): Likewise.
12037 (loadsf_update): Likewise.
12038 (movsicc_insn): Likewise.
12039 (bset_insn): Likewise.
12040 (bxor_insn): Likewise.
12041 (bclr_insn): Likewise.
12042 (bmsk_insn): Likewise.
12043 (bicsi3_insn): Likewise.
12044 (cmpsi_cc_c_insn): Likewise.
12045 (movsi_ne): Likewise.
12046 (movsi_cond_exec): Likewise.
12047 (clrsbsi2): Likewise.
12048 (norm_f): Likewise.
12058 (call_i): Remove 'S' asm letter, add jli instruction.
12059 (call_value_i): Likewise.
12060 * config/arc/arc.op (mjli-always): New option.
12061 * config/arc/constraints.md (Cji): New constraint.
12062 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
12064 (subsf3_fpx): Likewise.
12065 (mulsf3_fpx): Likewise.
12066 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
12068 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
12069 function attrbutes.
12070 * doc/invoke.texi (ARC): Document mjli-always option.
12072 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
12074 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
12075 avoid addition with 0 and use incw and decw where possible.
12077 2018-01-26 Richard Biener <rguenther@suse.de>
12079 PR tree-optimization/81082
12080 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
12081 association if it requires casting to unsigned.
12082 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
12083 from fold_plusminus_mult_expr to catch important cases late when
12084 range info is available.
12086 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12088 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
12089 * configure.ac (hidden_linkonce): New test.
12090 * configure: Regenerate.
12091 * config.in: Regenerate.
12093 2018-01-26 Julia Koval <julia.koval@intel.com>
12095 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
12096 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
12097 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
12098 _mm_mask_bitshuffle_epi64_mask): Fix type.
12099 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
12100 USI_FTYPE_V4DI_V4DI_USI): Remove.
12101 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
12102 __builtin_ia32_vpshufbitqmb256_mask,
12103 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
12104 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
12105 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
12107 2018-01-26 Alan Modra <amodra@gmail.com>
12110 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
12111 UNSPEC_VBPERMQ. Sort other unspecs.
12113 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
12115 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
12117 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
12119 PR middle-end/83055
12120 * predict.c (drop_profile): Do not push/pop cfun; update also
12122 (handle_missing_profiles): Fix logic looking for zero profiles.
12124 2018-01-25 Jakub Jelinek <jakub@redhat.com>
12126 PR middle-end/83977
12127 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
12128 on functions with #pragma omp declare simd or functions with simd
12130 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
12131 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
12132 Remove trailing \n from warning_at calls.
12134 2018-01-25 Tom de Vries <tom@codesourcery.com>
12137 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12138 for neutered workers.
12140 2018-01-24 Joseph Myers <joseph@codesourcery.com>
12143 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
12144 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
12146 2018-01-24 Jeff Law <law@redhat.com>
12149 * i386.c (get_probe_interval): Move to earlier point.
12150 (ix86_compute_frame_layout): If -fstack-clash-protection and
12151 the frame is larger than the probe interval, then use pushes
12152 to save registers rather than reg->mem moves.
12153 (ix86_expand_prologue): Remove conditional for int_registers_saved
12156 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
12159 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
12160 min/max for never referenced object.
12162 2018-01-24 Jakub Jelinek <jakub@redhat.com>
12164 PR middle-end/83977
12165 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
12167 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
12168 attributes from DECL_ATTRIBUTES (decl) without affecting
12169 DECL_ATTRIBUTES (current_function_decl).
12170 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
12171 functions with non-NULL DECL_ABSTRACT_ORIGIN.
12173 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
12175 PR tree-optimization/83979
12176 * fold-const.c (fold_comparison): Use constant_boolean_node
12177 instead of boolean_{true,false}_node.
12179 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
12181 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
12184 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12186 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12187 Simplify the clause that sets the length attribute.
12188 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12189 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
12190 clause that sets the length attribute.
12191 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12193 2018-01-24 Tom de Vries <tom@codesourcery.com>
12196 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
12197 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
12198 Add strict parameter.
12199 (prevent_branch_around_nothing): Insert dummy insn between branch to
12200 label and label with no ptx insn inbetween.
12201 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
12203 2018-01-24 Tom de Vries <tom@codesourcery.com>
12206 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12207 for neutered threads in warp.
12208 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
12210 2018-01-24 Richard Biener <rguenther@suse.de>
12212 PR tree-optimization/83176
12213 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
12216 2018-01-24 Richard Biener <rguenther@suse.de>
12218 PR tree-optimization/82819
12219 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
12220 code generating pluses that are no-ops in the target precision.
12222 2018-01-24 Richard Biener <rguenther@suse.de>
12224 PR middle-end/84000
12225 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
12227 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12229 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
12230 to merge probabilities.
12231 * predict.c (probably_never_executed): Also mark as cold functions
12232 with global 0 profile and guessed local profile.
12233 * profile-count.c (profile_probability::combine_with_count): New
12235 * profile-count.h (profile_probability::operator*,
12236 profile_probability::operator*=, profile_probability::operator/,
12237 profile_probability::operator/=): Reduce precision to adjusted
12238 and set value to guessed on contradictory divisions.
12239 (profile_probability::combine_with_freq): Remove.
12240 (profile_probability::combine_wiht_count): Declare.
12241 (profile_count::force_nonzero):: Set to adjusted.
12242 (profile_count::probability_in):: Set quality to adjusted.
12243 * tree-ssa-tail-merge.c (replace_block_by): Use
12244 combine_with_count.
12246 2018-01-23 Andrew Waterman <andrew@sifive.com>
12247 Jim Wilson <jimw@sifive.com>
12249 * config/riscv/riscv.c (riscv_stack_boundary): New.
12250 (riscv_option_override): Set riscv_stack_boundary. Handle
12251 riscv_preferred_stack_boundary_arg.
12252 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
12253 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
12254 (STACK_BOUNDARY): Set to riscv_stack_boundary.
12255 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
12256 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
12257 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
12259 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
12262 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
12263 of struct ix86_frame.
12264 (ix86_expand_epilogue): Likewise. Add a local variable for
12265 the reg_save_offset field in struct ix86_frame.
12267 2018-01-23 Bin Cheng <bin.cheng@arm.com>
12269 PR tree-optimization/82604
12270 * tree-loop-distribution.c (enum partition_kind): New enum item
12271 PKIND_PARTIAL_MEMSET.
12272 (partition_builtin_p): Support above new enum item.
12273 (generate_code_for_partition): Ditto.
12274 (compute_access_range): Differentiate cases that equality can be
12275 proven at all loops, the innermost loops or no loops.
12276 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
12277 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
12278 (finalize_partitions, distribute_loop): Don't fuse partition of
12279 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
12280 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
12281 parloop is enabled.
12283 2018-01-23 Martin Liska <mliska@suse.cz>
12285 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
12286 order to ignore the predictor.
12287 (PRED_POLYMORPHIC_CALL): Likewise.
12288 (PRED_RECURSIVE_CALL): Likewise.
12290 2018-01-23 Martin Liska <mliska@suse.cz>
12292 * tree-profile.c (tree_profiling): Print function header to
12293 aware reader which function we are working on.
12294 * value-prof.c (gimple_find_values_to_profile): Do not print
12295 not interesting value histograms.
12297 2018-01-23 Martin Liska <mliska@suse.cz>
12299 * profile-count.h (enum profile_quality): Add
12300 profile_uninitialized as the first value. Do not number values
12301 as they are zero based.
12302 (profile_count::verify): Update sanity check.
12303 (profile_probability::verify): Likewise.
12305 2018-01-23 Nathan Sidwell <nathan@acm.org>
12307 * doc/invoke.texi (ffor-scope): Deprecate.
12309 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12311 PR tree-optimization/83510
12312 * domwalk.c (set_all_edges_as_executable): New function.
12313 (dom_walker::dom_walker): Convert bool param
12314 "skip_unreachable_blocks" to enum reachability. Move setup of
12315 edge flags to set_all_edges_as_executable and only do it when
12316 reachability is REACHABLE_BLOCKS.
12317 * domwalk.h (enum dom_walker::reachability): New enum.
12318 (dom_walker::dom_walker): Convert bool param
12319 "skip_unreachable_blocks" to enum reachability.
12320 (set_all_edges_as_executable): New decl.
12321 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
12322 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
12324 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
12325 but converting true to REACHABLE_BLOCKS.
12326 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
12328 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
12329 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
12330 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
12332 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
12333 if check_all_array_refs will be called.
12335 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12337 * tree.c (selftest::test_location_wrappers): Add more test
12340 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12342 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
12343 (selftest::test_bit_in_range): Likewise.
12345 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12348 * doc/sourcebuild.texi (vect_float): Say that the selector
12349 only describes the situation when -funsafe-math-optimizations is on.
12350 (vect_float_strict): Document.
12352 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12354 PR tree-optimization/83965
12355 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
12356 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
12357 instead of checking only for a reduction.
12358 (vect_recog_widen_sum_pattern): Likewise.
12360 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12362 * predict.c (probably_never_executed): Only use precise profile info.
12363 (compute_function_frequency): Skip after inlining hack since we now
12364 have quality checking.
12366 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12368 * profile-count.h (profile_probability::very_unlikely,
12369 profile_probability::unlikely, profile_probability::even): Set
12370 precision to guessed.
12372 2018-01-23 Richard Biener <rguenther@suse.de>
12374 PR tree-optimization/83963
12375 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
12376 Properly terminate dominator walk when crossing the exit edge not
12377 when visiting its source block.
12379 2018-01-23 Jakub Jelinek <jakub@redhat.com>
12382 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
12383 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
12385 2018-01-22 Jakub Jelinek <jakub@redhat.com>
12387 PR tree-optimization/83957
12388 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
12389 semicolon after for body surrounded by braces.
12391 PR tree-optimization/83081
12392 * profile-count.h (profile_probability::split): New method.
12393 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
12394 Use profile_probability::split.
12395 (do_compare_rtx_and_jump): Fix adjustment of probabilities
12396 when splitting a single conditional jump into 2.
12398 2018-01-22 David Malcolm <dmalcolm@redhat.com>
12400 PR tree-optimization/69452
12401 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
12404 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12406 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
12407 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
12408 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
12410 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12412 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
12414 * config/rl78/rl78.md (movdi): New define_expand.
12415 * config/rl78/rl78.c (rl78_split_movdi): New function.
12417 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
12420 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
12422 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
12423 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
12424 128-bit to produce an UNSPEC move to get the double word with the
12425 signbit and then a shift directly to do signbit.
12426 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
12427 implementation with a new version that just does either a direct
12428 move or a regular move. Move memory interface to separate insns.
12429 Move insns so they are next to the expander.
12430 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
12431 with signbit move. Split big and little endian case.
12432 (signbit<mode>2_dm_mem_le): Likewise.
12433 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
12434 (signbit<mode>2_dm2): Likewise.
12436 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12438 * config/rl78/rl78.md (anddi3): New define_expand.
12440 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12442 * config/rl78/rl78.md (umindi3): New define_expand.
12444 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12446 * config/rl78/rl78.md (smindi3): New define_expand.
12448 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12450 * config/rl78/rl78.md (smaxdi3): New define_expand.
12452 2018-01-22 Carl Love <cel@us.ibm.com>
12454 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
12455 LVX_V1TI): Add macro expansion.
12456 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
12457 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
12458 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
12459 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
12460 Change check to determine if the instruction is a byte reversing
12461 entry. Fix typo in comment.
12462 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
12463 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
12464 Add def_builtin calls for new builtins.
12465 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
12466 Add define_insn expansion.
12468 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12470 * config/rl78/rl78.md (umaxdi3): New define_expand.
12472 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12474 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
12475 for non-QImode registers.
12477 2018-01-22 Richard Biener <rguenther@suse.de>
12479 PR tree-optimization/83963
12480 * graphite-scop-detection.c (scop_detection::get_sese): Delay
12481 including the loop exit block.
12482 (scop_detection::merge_sese): Likewise.
12483 (scop_detection::add_scop): Do it here instead.
12485 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12487 * doc/sourcebuild.texi (arm_softfloat): Document.
12489 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
12492 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
12493 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
12494 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
12496 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12497 David Edelsohn <dje.gcc@gmail.com>
12500 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12501 Change "crset eq" to "crset 2".
12502 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12503 (*call_indirect_aix<mode>_nospec): Likewise.
12504 (*call_value_indirect_aix<mode>_nospec): Likewise.
12505 (*call_indirect_elfv2<mode>_nospec): Likewise.
12506 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
12507 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
12508 change assembly output from . to $.
12509 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12510 (indirect_jump<mode>_nospec): Change assembly output from . to $.
12511 (*tablejump<mode>_internal1_nospec): Likewise.
12513 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
12516 * config/sh/sh_optimize_sett_clrt.cc:
12517 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
12519 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
12521 PR tree-optimization/83940
12522 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
12523 offset_dt to vect_constant_def rather than vect_unknown_def_type.
12524 (vect_check_load_store_mask): Add a mask_dt_out parameter and
12525 use it to pass back the definition type.
12526 (vect_check_store_rhs): Likewise rhs_dt_out.
12527 (vect_build_gather_load_calls): Add a mask_dt argument and use
12528 it instead of a call to vect_is_simple_use.
12529 (vectorizable_store): Update calls to vect_check_load_store_mask
12530 and vect_check_store_rhs. Use the dt returned by the latter instead
12531 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
12532 instead of calls to vect_is_simple_use. Pass the scalar rather
12533 than the vector operand to vect_is_simple_use when handling
12534 second and subsequent copies of an rhs value.
12535 (vectorizable_load): Update calls to vect_check_load_store_mask
12536 and vect_build_gather_load_calls. Use the cached mask_dt and
12537 gs_info.offset_dt instead of calls to vect_is_simple_use.
12539 2018-01-20 Jakub Jelinek <jakub@redhat.com>
12541 PR middle-end/83945
12542 * tree-emutls.c: Include gimplify.h.
12543 (lower_emutls_2): New function.
12544 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
12545 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
12546 it before further processing.
12549 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
12550 UINTVAL (trueop1) instead of INTVAL (op1).
12552 2018-01-19 Jakub Jelinek <jakub@redhat.com>
12556 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
12557 INCOMING_FRAME_SP_OFFSET if not defined.
12558 (scan_trace): Add ENTRY argument. If true and
12559 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
12560 emit a note to adjust the CFA offset.
12561 (create_cfi_notes): Adjust scan_trace callers.
12562 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
12563 INCOMING_FRAME_SP_OFFSET in the CIE.
12564 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
12565 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
12567 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
12568 * doc/tm.texi: Regenerated.
12570 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12572 PR rtl-optimization/83147
12573 * lra-constraints.c (remove_inheritance_pseudos): Use
12574 lra_substitute_pseudo_within_insn.
12576 2018-01-19 Tom de Vries <tom@codesourcery.com>
12577 Cesar Philippidis <cesar@codesourcery.com>
12580 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
12582 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
12585 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
12586 spaces for function labels.
12588 2018-01-19 Martin Liska <mliska@suse.cz>
12590 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
12591 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
12592 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
12593 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
12594 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
12595 (PRED_CONST_RETURN): Change from 69 to 65.
12596 (PRED_NULL_RETURN): Change from 91 to 71.
12597 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
12598 (PRED_LOOP_GUARD): Change from 66 to 73.
12600 2018-01-19 Martin Liska <mliska@suse.cz>
12602 * predict.c (predict_insn_def): Add new assert.
12603 (struct branch_predictor): Change type to signed integer.
12604 (test_prediction_value_range): Amend test to cover
12605 PROB_UNINITIALIZED.
12606 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
12607 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
12608 (PRED_LOOP_ITERATIONS_MAX): Likewise.
12609 (PRED_LOOP_IV_COMPARE): Likewise.
12610 * predict.h (PROB_UNINITIALIZED): Define new constant.
12612 2018-01-19 Martin Liska <mliska@suse.cz>
12614 * predict.c (dump_prediction): Add new format for
12615 analyze_brprob.py script which is enabled with -details
12617 * profile-count.h (precise_p): New function.
12619 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12621 PR tree-optimization/83922
12622 * tree-vect-loop.c (vect_verify_full_masking): Return false if
12623 there are no statements that need masking.
12624 (vect_active_double_reduction_p): New function.
12625 (vect_analyze_loop_operations): Use it when handling phis that
12626 are not in the loop header.
12628 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12630 PR tree-optimization/83914
12631 * tree-vect-loop.c (vectorizable_induction): Don't convert
12632 init_expr or apply the peeling adjustment for inductions
12633 that are nested within the vectorized loop.
12635 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12637 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
12640 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12644 * function.h (gimplify_parameters): Add gimple_seq * argument.
12645 * function.c: Include gimple.h and options.h.
12646 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
12647 for the added local temporaries if needed.
12648 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
12649 if there are any parameter cleanups, wrap whole body into a
12650 try/finally with the cleanups.
12652 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
12655 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
12656 Use GET_MODE_CLASS for scalar floating point.
12658 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12662 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
12663 Fix call of call_cgraph_insertion_hooks.
12665 2018-01-18 Martin Sebor <msebor@redhat.com>
12667 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
12669 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12672 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
12675 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
12678 * common.opt: (-ffile-prefix-map): New option.
12679 * opts.c (common_handle_option): Defer it.
12680 * opts-global.c (handle_common_deferred_options): Handle it.
12681 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
12682 * file-prefix-map.h: New file.
12683 (remap_debug_filename, add_debug_prefix_map): ...here.
12684 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
12685 * final.c (debug_prefix_map, add_debug_prefix_map
12686 remap_debug_filename): Move to...
12687 * file-prefix-map.c: New file.
12688 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
12689 generalize, get rid of alloca(), use strrchr() instead of strchr().
12690 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
12691 Implement in terms of add_prefix_map().
12692 (remap_macro_filename, remap_debug_filename): Implement in term of
12694 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
12695 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
12696 * dbxout.c: Include file-prefix-map.h.
12697 * varasm.c: Likewise.
12698 * vmsdbgout.c: Likewise.
12699 * xcoffout.c: Likewise.
12700 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
12701 * doc/cppopts.texi (-fmacro-prefix-map): Document.
12702 * doc/invoke.texi (-ffile-prefix-map): Document.
12703 (-fdebug-prefix-map): Update description.
12705 2018-01-18 Martin Liska <mliska@suse.cz>
12707 * config/i386/i386.c (indirect_thunk_name): Document that also
12709 (output_indirect_thunk): Document why both instructions
12710 (pause and lfence) are generated.
12712 2018-01-18 Richard Biener <rguenther@suse.de>
12714 PR tree-optimization/83887
12715 * graphite-scop-detection.c
12716 (scop_detection::get_nearest_dom_with_single_entry): Remove.
12717 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
12718 (scop_detection::merge_sese): Re-implement with a flood-fill
12719 algorithm that properly finds a SESE region if it exists.
12721 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12724 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
12725 pointer_diff optimizations use view_convert instead of convert.
12727 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12729 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12730 Generate different code for -mno-speculate-indirect-jumps.
12731 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12732 (*call_indirect_aix<mode>): Disable for
12733 -mno-speculate-indirect-jumps.
12734 (*call_indirect_aix<mode>_nospec): New define_insn.
12735 (*call_value_indirect_aix<mode>): Disable for
12736 -mno-speculate-indirect-jumps.
12737 (*call_value_indirect_aix<mode>_nospec): New define_insn.
12738 (*sibcall_nonlocal_sysv<mode>): Generate different code for
12739 -mno-speculate-indirect-jumps.
12740 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12742 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12744 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
12745 long double type, set the flags for noting the default long double
12746 type, even if we don't pass or return a long double type.
12748 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
12751 * ipa-inline.c (flatten_function): Do not overwrite final inlining
12754 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
12756 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
12757 support for merge[hl].
12758 (fold_mergehl_helper): New helper function.
12759 (tree-vector-builder.h): New #include for tree_vector_builder usage.
12760 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
12761 (altivec_vmrglw_direct): Add xxmrglw insn.
12763 2018-01-17 Andrew Waterman <andrew@sifive.com>
12765 * config/riscv/riscv.c (riscv_conditional_register_usage): If
12766 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
12768 2018-01-17 David Malcolm <dmalcolm@redhat.com>
12771 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
12772 call the lto_location_cache before reading the
12773 DECL_SOURCE_LOCATION of the types.
12775 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
12776 Richard Sandiford <richard.sandiford@linaro.org>
12778 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
12779 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
12780 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
12781 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
12782 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
12784 * config/aarch64/constraints.md (aarch64_movti_operand):
12786 * config/aarch64/predicates.md (Uti): Add new constraint.
12788 2018-01-17 Carl Love <cel@us.ibm.com>
12790 * config/rs6000/vsx.md (define_expand xl_len_r,
12791 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
12792 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
12794 (define_expand, define_insn): Move the shift left from the
12795 define_insn to the define_expand for lxvl and stxvl instructions.
12796 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
12797 and XL_LEN_R definitions to PURE.
12799 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
12801 * config/i386/i386.c (indirect_thunk_name): Declare regno
12802 as unsigned int. Compare regno with INVALID_REGNUM.
12803 (output_indirect_thunk): Ditto.
12804 (output_indirect_thunk_function): Ditto.
12805 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
12806 in the call to output_indirect_thunk_function.
12808 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
12810 PR middle-end/83884
12811 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
12812 rather than the size of inner_type to determine the stack slot size
12813 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
12815 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
12818 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
12821 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
12823 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
12824 endian Linux systems to optionally enable multilibs for selecting
12825 the long double type if the user configured an explicit type.
12826 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
12827 have no long double multilibs if not defined.
12828 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
12829 warn if the user used -mabi={ieee,ibm}longdouble and we built
12830 multilibs for long double.
12831 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
12832 appropriate multilib option.
12833 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
12835 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
12836 for building long double multilibs.
12837 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
12839 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
12841 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
12844 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
12846 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
12849 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
12852 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
12855 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12857 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
12858 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
12860 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
12862 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
12863 different rtl trees depending on TARGET_64BIT.
12864 (rs6000_gen_lvx): Likewise.
12866 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12868 * config/visium/visium.md (nop): Tweak comment.
12869 (hazard_nop): Likewise.
12871 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12873 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
12874 -mspeculate-indirect-jumps.
12875 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
12876 for -mno-speculate-indirect-jumps.
12877 (*call_indirect_elfv2<mode>_nospec): New define_insn.
12878 (*call_value_indirect_elfv2<mode>): Disable for
12879 -mno-speculate-indirect-jumps.
12880 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
12881 (indirect_jump): Emit different RTL for
12882 -mno-speculate-indirect-jumps.
12883 (*indirect_jump<mode>): Disable for
12884 -mno-speculate-indirect-jumps.
12885 (*indirect_jump<mode>_nospec): New define_insn.
12886 (tablejump): Emit different RTL for
12887 -mno-speculate-indirect-jumps.
12888 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
12889 (tablejumpsi_nospec): New define_expand.
12890 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
12891 (tablejumpdi_nospec): New define_expand.
12892 (*tablejump<mode>_internal1): Disable for
12893 -mno-speculate-indirect-jumps.
12894 (*tablejump<mode>_internal1_nospec): New define_insn.
12895 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
12898 2018-01-16 Artyom Skrobov tyomitch@gmail.com
12900 * caller-save.c (insert_save): Drop unnecessary parameter. All
12903 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12904 Richard Biener <rguenth@suse.de>
12907 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
12908 return early, inline manually is_gimple_sizepos. Make sure if we
12909 call gimplify_expr we don't end up with a gimple constant.
12910 * tree.c (variably_modified_type_p): Don't return true for
12911 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
12912 * gimplify.h (is_gimple_sizepos): Remove.
12914 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12916 PR tree-optimization/83857
12917 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
12918 vectorizable_live_operation for pure SLP statements.
12919 (vectorizable_live_operation): Handle PHIs.
12921 2018-01-16 Richard Biener <rguenther@suse.de>
12923 PR tree-optimization/83867
12924 * tree-vect-stmts.c (vect_transform_stmt): Precompute
12925 nested_in_vect_loop_p since the scalar stmt may get invalidated.
12927 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12930 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
12931 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
12932 If off is not INTEGER_CST, issue a may not be aligned warning
12933 rather than isn't aligned. Use isn%'t rather than isn't.
12934 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
12936 <case MULT_EXPR>: Improve the case when bottom and one of the
12937 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
12938 operand, in that case check if the other operand is multiple of
12939 bottom divided by the INTEGER_CST operand.
12941 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12944 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
12945 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
12946 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
12947 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
12948 * config/pa/pa.c (pa_function_arg_advance): Likewise.
12949 (pa_function_arg, pa_arg_partial_bytes): Likewise.
12950 (pa_function_arg_size): New function.
12952 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12954 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
12955 in a separate statement.
12957 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12959 PR tree-optimization/83847
12960 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
12961 group gathers and scatters.
12963 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12965 PR rtl-optimization/86620
12966 * params.def (max-sched-ready-insns): Bump minimum value to 1.
12968 PR rtl-optimization/83213
12969 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
12970 to last if both are JUMP_INSNs.
12972 PR tree-optimization/83843
12973 * gimple-ssa-store-merging.c
12974 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
12975 store_immediate_info for bswap/nop orig_stores.
12977 2018-01-15 Andrew Waterman <andrew@sifive.com>
12979 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
12981 <UDIV>: Increase cost if !TARGET_DIV.
12983 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
12985 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
12986 (define_attr "cr_logical_3op"): New.
12987 (cceq_ior_compare): Adjust.
12988 (cceq_ior_compare_complement): Adjust.
12989 (*cceq_rev_compare): Adjust.
12990 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
12991 (is_cracked_insn): Adjust.
12992 (insn_must_be_first_in_group): Adjust.
12993 * config/rs6000/40x.md: Adjust.
12994 * config/rs6000/440.md: Adjust.
12995 * config/rs6000/476.md: Adjust.
12996 * config/rs6000/601.md: Adjust.
12997 * config/rs6000/603.md: Adjust.
12998 * config/rs6000/6xx.md: Adjust.
12999 * config/rs6000/7450.md: Adjust.
13000 * config/rs6000/7xx.md: Adjust.
13001 * config/rs6000/8540.md: Adjust.
13002 * config/rs6000/cell.md: Adjust.
13003 * config/rs6000/e300c2c3.md: Adjust.
13004 * config/rs6000/e500mc.md: Adjust.
13005 * config/rs6000/e500mc64.md: Adjust.
13006 * config/rs6000/e5500.md: Adjust.
13007 * config/rs6000/e6500.md: Adjust.
13008 * config/rs6000/mpc.md: Adjust.
13009 * config/rs6000/power4.md: Adjust.
13010 * config/rs6000/power5.md: Adjust.
13011 * config/rs6000/power6.md: Adjust.
13012 * config/rs6000/power7.md: Adjust.
13013 * config/rs6000/power8.md: Adjust.
13014 * config/rs6000/power9.md: Adjust.
13015 * config/rs6000/rs64.md: Adjust.
13016 * config/rs6000/titan.md: Adjust.
13018 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13020 * config/i386/predicates.md (indirect_branch_operand): Rewrite
13021 ix86_indirect_branch_register logic.
13023 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13025 * config/i386/constraints.md (Bs): Update
13026 ix86_indirect_branch_register check. Don't check
13027 ix86_indirect_branch_register with GOT_memory_operand.
13029 * config/i386/predicates.md (GOT_memory_operand): Don't check
13030 ix86_indirect_branch_register here.
13031 (GOT32_symbol_operand): Likewise.
13033 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13035 * config/i386/predicates.md (constant_call_address_operand):
13036 Rewrite ix86_indirect_branch_register logic.
13037 (sibcall_insn_operand): Likewise.
13039 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13041 * config/i386/constraints.md (Bs): Replace
13042 ix86_indirect_branch_thunk_register with
13043 ix86_indirect_branch_register.
13045 * config/i386/i386.md (indirect_jump): Likewise.
13046 (tablejump): Likewise.
13047 (*sibcall_memory): Likewise.
13048 (*sibcall_value_memory): Likewise.
13049 Peepholes of indirect call and jump via memory: Likewise.
13050 * config/i386/i386.opt: Likewise.
13051 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13052 (GOT_memory_operand): Likewise.
13053 (call_insn_operand): Likewise.
13054 (sibcall_insn_operand): Likewise.
13055 (GOT32_symbol_operand): Likewise.
13057 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13059 PR middle-end/83837
13060 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
13061 type rather than type addr's type points to.
13062 (expand_omp_atomic_mutex): Likewise.
13063 (expand_omp_atomic): Likewise.
13065 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13068 * config/i386/i386.c (output_indirect_thunk_function): Use
13069 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
13070 for __x86_return_thunk.
13072 2018-01-15 Richard Biener <rguenther@suse.de>
13074 PR middle-end/83850
13075 * expmed.c (extract_bit_field_1): Fix typo.
13077 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13080 * config/arm/iterators.md (VF): New mode iterator.
13081 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
13082 Remove integer-related logic from pattern.
13083 (neon_vabd<mode>_3): Likewise.
13085 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13087 PR middle-end/82694
13088 * common.opt (fstrict-overflow): No longer an alias.
13089 (fwrapv-pointer): New option.
13090 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
13091 also for pointer types based on flag_wrapv_pointer.
13092 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
13093 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
13094 opts->x_flag_wrapv got set.
13095 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
13096 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
13097 POINTER_TYPE_OVERFLOW_UNDEFINED.
13098 * match.pd: Likewise in address comparison pattern.
13099 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
13101 2018-01-15 Richard Biener <rguenther@suse.de>
13104 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
13105 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
13106 Reset type names to their identifier if their TYPE_DECL doesn't
13107 have linkage (and thus is used for ODR and devirt).
13108 (save_debug_info_for_decl): Remove.
13109 (save_debug_info_for_type): Likewise.
13110 (add_tree_to_fld_list): Adjust.
13111 * tree-pretty-print.c (dump_generic_node): Make dumping of
13112 type names more robust.
13114 2018-01-15 Richard Biener <rguenther@suse.de>
13116 * BASE-VER: Bump to 8.0.1.
13118 2018-01-14 Martin Sebor <msebor@redhat.com>
13121 * builtins.c (check_access): Avoid warning when the no-warning bit
13124 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
13126 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
13127 * ira-color (allocno_hard_regs_compare): Likewise.
13129 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
13132 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
13133 Use .pushsection/.popsection.
13135 2018-01-14 Martin Sebor <msebor@redhat.com>
13138 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
13140 2018-01-14 Jakub Jelinek <jakub@redhat.com>
13142 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
13143 entry from extra_headers.
13144 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
13145 extra_headers, make the list bitwise identical to the i?86-*-* one.
13147 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13149 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
13150 -mcmodel=large with -mindirect-branch=thunk,
13151 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
13152 -mfunction-return=thunk-extern.
13153 * doc/invoke.texi: Document -mcmodel=large is incompatible with
13154 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
13155 -mfunction-return=thunk and -mfunction-return=thunk-extern.
13157 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13159 * config/i386/i386.c (print_reg): Print the name of the full
13160 integer register without '%'.
13161 (ix86_print_operand): Handle 'V'.
13162 * doc/extend.texi: Document 'V' modifier.
13164 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13166 * config/i386/constraints.md (Bs): Disallow memory operand for
13167 -mindirect-branch-register.
13169 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13170 (GOT_memory_operand): Likewise.
13171 (call_insn_operand): Likewise.
13172 (sibcall_insn_operand): Likewise.
13173 (GOT32_symbol_operand): Likewise.
13174 * config/i386/i386.md (indirect_jump): Call convert_memory_address
13175 for -mindirect-branch-register.
13176 (tablejump): Likewise.
13177 (*sibcall_memory): Likewise.
13178 (*sibcall_value_memory): Likewise.
13179 Disallow peepholes of indirect call and jump via memory for
13180 -mindirect-branch-register.
13181 (*call_pop): Replace m with Bw.
13182 (*call_value_pop): Likewise.
13183 (*sibcall_pop_memory): Replace m with Bs.
13184 * config/i386/i386.opt (mindirect-branch-register): New option.
13185 * doc/invoke.texi: Document -mindirect-branch-register option.
13187 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13189 * config/i386/i386-protos.h (ix86_output_function_return): New.
13190 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
13191 set function_return_type.
13192 (indirect_thunk_name): Add ret_p to indicate thunk for function
13194 (output_indirect_thunk_function): Pass false to
13195 indirect_thunk_name.
13196 (ix86_output_indirect_branch_via_reg): Likewise.
13197 (ix86_output_indirect_branch_via_push): Likewise.
13198 (output_indirect_thunk_function): Create alias for function
13199 return thunk if regno < 0.
13200 (ix86_output_function_return): New function.
13201 (ix86_handle_fndecl_attribute): Handle function_return.
13202 (ix86_attribute_table): Add function_return.
13203 * config/i386/i386.h (machine_function): Add
13204 function_return_type.
13205 * config/i386/i386.md (simple_return_internal): Use
13206 ix86_output_function_return.
13207 (simple_return_internal_long): Likewise.
13208 * config/i386/i386.opt (mfunction-return=): New option.
13209 (indirect_branch): Mention -mfunction-return=.
13210 * doc/extend.texi: Document function_return function attribute.
13211 * doc/invoke.texi: Document -mfunction-return= option.
13213 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13215 * config/i386/i386-opts.h (indirect_branch): New.
13216 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
13217 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
13218 with local indirect jump when converting indirect call and jump.
13219 (ix86_set_indirect_branch_type): New.
13220 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
13221 (indirectlabelno): New.
13222 (indirect_thunk_needed): Likewise.
13223 (indirect_thunk_bnd_needed): Likewise.
13224 (indirect_thunks_used): Likewise.
13225 (indirect_thunks_bnd_used): Likewise.
13226 (INDIRECT_LABEL): Likewise.
13227 (indirect_thunk_name): Likewise.
13228 (output_indirect_thunk): Likewise.
13229 (output_indirect_thunk_function): Likewise.
13230 (ix86_output_indirect_branch_via_reg): Likewise.
13231 (ix86_output_indirect_branch_via_push): Likewise.
13232 (ix86_output_indirect_branch): Likewise.
13233 (ix86_output_indirect_jmp): Likewise.
13234 (ix86_code_end): Call output_indirect_thunk_function if needed.
13235 (ix86_output_call_insn): Call ix86_output_indirect_branch if
13237 (ix86_handle_fndecl_attribute): Handle indirect_branch.
13238 (ix86_attribute_table): Add indirect_branch.
13239 * config/i386/i386.h (machine_function): Add indirect_branch_type
13240 and has_local_indirect_jump.
13241 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
13243 (tablejump): Likewise.
13244 (*indirect_jump): Use ix86_output_indirect_jmp.
13245 (*tablejump_1): Likewise.
13246 (simple_return_indirect_internal): Likewise.
13247 * config/i386/i386.opt (mindirect-branch=): New option.
13248 (indirect_branch): New.
13251 (thunk-inline): Likewise.
13252 (thunk-extern): Likewise.
13253 * doc/extend.texi: Document indirect_branch function attribute.
13254 * doc/invoke.texi: Document -mindirect-branch= option.
13256 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
13259 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
13261 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
13263 * ipa-inline.c (want_inline_small_function_p): Return false if
13264 inlining has already failed with CIF_FINAL_ERROR.
13265 (update_caller_keys): Call want_inline_small_function_p before
13267 (update_callee_keys): Likewise.
13269 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
13271 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
13273 (rs6000_quadword_masked_address_p): Likewise.
13274 (quad_aligned_load_p): Likewise.
13275 (quad_aligned_store_p): Likewise.
13276 (const_load_sequence_p): Add comment to describe the outer-most loop.
13277 (mimic_memory_attributes_and_flags): New function.
13278 (rs6000_gen_stvx): Likewise.
13279 (replace_swapped_aligned_store): Likewise.
13280 (rs6000_gen_lvx): Likewise.
13281 (replace_swapped_aligned_load): Likewise.
13282 (replace_swapped_load_constant): Capitalize argument name in
13283 comment describing this function.
13284 (rs6000_analyze_swaps): Add a third pass to search for vector loads
13285 and stores that access quad-word aligned addresses and replace
13286 with stvx or lvx instructions when appropriate.
13287 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
13288 New function prototype.
13289 (rs6000_quadword_masked_address_p): Likewise.
13290 (rs6000_gen_lvx): Likewise.
13291 (rs6000_gen_stvx): Likewise.
13292 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
13293 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
13294 when memory address is aligned.
13295 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
13296 this split to select lvx instruction when memory address is aligned.
13297 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
13298 instruction when memory address is aligned.
13299 (*vsx_le_perm_load_v16qi): Likewise.
13300 (four unnamed splitters): Modify to select the stvx instruction
13301 when memory is aligned.
13303 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
13305 * predict.c (determine_unlikely_bbs): Handle correctly BBs
13306 which appears in the queue multiple times.
13308 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13309 Alan Hayward <alan.hayward@arm.com>
13310 David Sherwood <david.sherwood@arm.com>
13312 * tree-vectorizer.h (vec_lower_bound): New structure.
13313 (_loop_vec_info): Add check_nonzero and lower_bounds.
13314 (LOOP_VINFO_CHECK_NONZERO): New macro.
13315 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
13316 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
13317 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
13318 fields. Make seg_len the distance travelled, not including the
13320 (dr_direction_indicator): Declare.
13321 (dr_zero_step_indicator): Likewise.
13322 (dr_known_forward_stride_p): Likewise.
13323 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
13325 (runtime_alias_check_p): Allow runtime alias checks with
13327 (operator ==): Compare access_size and align.
13328 (prune_runtime_alias_test_list): Rework for new distinction between
13329 the access_size and seg_len.
13330 (create_intersect_range_checks_index): Likewise. Cope with polynomial
13332 (get_segment_min_max): New function.
13333 (create_intersect_range_checks): Use it.
13334 (dr_step_indicator): New function.
13335 (dr_direction_indicator): Likewise.
13336 (dr_zero_step_indicator): Likewise.
13337 (dr_known_forward_stride_p): Likewise.
13338 * tree-loop-distribution.c (data_ref_segment_size): Return
13339 DR_STEP * (niters - 1).
13340 (compute_alias_check_pairs): Update call to the dr_with_seg_len
13342 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
13343 (vect_preserves_scalar_order_p): New function, split out from...
13344 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
13345 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
13346 (vect_vfa_access_size): New function.
13347 (vect_vfa_align): Likewise.
13348 (vect_compile_time_alias): Take access_size_a and access_b arguments.
13349 (dump_lower_bound): New function.
13350 (vect_check_lower_bound): Likewise.
13351 (vect_small_gap_p): Likewise.
13352 (vectorizable_with_step_bound_p): Likewise.
13353 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
13354 depencies if the vectorization factor is 1. Convert the checks
13355 for nonzero steps into checks on the bounds of DR_STEP. Try using
13356 a bunds check for variable steps if the minimum required step is
13357 relatively small. Update calls to the dr_with_seg_len
13358 constructor and to vect_compile_time_alias.
13359 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
13361 (vect_loop_versioning): Call it.
13362 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
13364 (vect_estimate_min_profitable_iters): Account for any bounds checks.
13366 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13367 Alan Hayward <alan.hayward@arm.com>
13368 David Sherwood <david.sherwood@arm.com>
13370 * doc/sourcebuild.texi (vect_scatter_store): Document.
13371 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
13373 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
13375 * genopinit.c (main): Add supports_vec_scatter_store and
13376 supports_vec_scatter_store_cached to target_optabs.
13377 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
13378 IFN_MASK_SCATTER_STORE.
13379 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
13381 * internal-fn.h (internal_store_fn_p): Declare.
13382 (internal_fn_stored_value_index): Likewise.
13383 * internal-fn.c (scatter_store_direct): New macro.
13384 (expand_scatter_store_optab_fn): New function.
13385 (direct_scatter_store_optab_supported_p): New macro.
13386 (internal_store_fn_p): New function.
13387 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
13388 IFN_MASK_SCATTER_STORE.
13389 (internal_fn_mask_index): Likewise.
13390 (internal_fn_stored_value_index): New function.
13391 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
13392 for scatter stores.
13393 * optabs-query.h (supports_vec_scatter_store_p): Declare.
13394 * optabs-query.c (supports_vec_scatter_store_p): New function.
13395 * tree-vectorizer.h (vect_get_store_rhs): Declare.
13396 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
13397 true for scatter stores.
13398 (vect_gather_scatter_fn_p): Handle scatter stores too.
13399 (vect_check_gather_scatter): Consider using scatter stores if
13400 supports_vec_scatter_store_p.
13401 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
13402 scatter stores too.
13403 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13404 internal_fn_stored_value_index.
13405 (check_load_store_masking): Handle scatter stores too.
13406 (vect_get_store_rhs): Make public.
13407 (vectorizable_call): Use internal_store_fn_p.
13408 (vectorizable_store): Handle scatter store internal functions.
13409 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
13410 when deciding whether the end of the group has been reached.
13411 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
13412 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
13413 (mask_scatter_store<mode>): New insns.
13415 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13416 Alan Hayward <alan.hayward@arm.com>
13417 David Sherwood <david.sherwood@arm.com>
13419 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
13420 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
13421 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
13423 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
13424 Use vect_truncate_gather_scatter_offset if we can't treat the
13425 operation as a normal gather load or scatter store.
13426 (get_group_load_store_type): Take the gather_scatter_info
13427 as argument. Try using a gather load or scatter store for
13428 single-element groups.
13429 (get_load_store_type): Update calls to get_group_load_store_type
13430 and vect_use_strided_gather_scatters_p.
13432 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13433 Alan Hayward <alan.hayward@arm.com>
13434 David Sherwood <david.sherwood@arm.com>
13436 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
13437 optional tree argument.
13438 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
13440 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
13441 but continue to use the current value as a fallback.
13442 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
13443 to compare the updates.
13444 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
13445 (get_load_store_type): Use it when handling a strided access.
13446 (vect_get_strided_load_store_ops): New function.
13447 (vect_get_data_ptr_increment): Likewise.
13448 (vectorizable_load): Handle strided gather loads. Always pass
13449 a step to vect_create_data_ref_ptr and bump_vector_ptr.
13451 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13452 Alan Hayward <alan.hayward@arm.com>
13453 David Sherwood <david.sherwood@arm.com>
13455 * doc/md.texi (gather_load@var{m}): Document.
13456 (mask_gather_load@var{m}): Likewise.
13457 * genopinit.c (main): Add supports_vec_gather_load and
13458 supports_vec_gather_load_cached to target_optabs.
13459 * optabs-tree.c (init_tree_optimization_optabs): Use
13460 ggc_cleared_alloc to allocate target_optabs.
13461 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
13462 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
13464 * internal-fn.h (internal_load_fn_p): Declare.
13465 (internal_gather_scatter_fn_p): Likewise.
13466 (internal_fn_mask_index): Likewise.
13467 (internal_gather_scatter_fn_supported_p): Likewise.
13468 * internal-fn.c (gather_load_direct): New macro.
13469 (expand_gather_load_optab_fn): New function.
13470 (direct_gather_load_optab_supported_p): New macro.
13471 (direct_internal_fn_optab): New function.
13472 (internal_load_fn_p): Likewise.
13473 (internal_gather_scatter_fn_p): Likewise.
13474 (internal_fn_mask_index): Likewise.
13475 (internal_gather_scatter_fn_supported_p): Likewise.
13476 * optabs-query.c (supports_at_least_one_mode_p): New function.
13477 (supports_vec_gather_load_p): Likewise.
13478 * optabs-query.h (supports_vec_gather_load_p): Declare.
13479 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
13480 and memory_type field.
13481 (NUM_PATTERNS): Bump to 15.
13482 * tree-vect-data-refs.c: Include internal-fn.h.
13483 (vect_gather_scatter_fn_p): New function.
13484 (vect_describe_gather_scatter_call): Likewise.
13485 (vect_check_gather_scatter): Try using internal functions for
13486 gather loads. Recognize existing calls to a gather load function.
13487 (vect_analyze_data_refs): Consider using gather loads if
13488 supports_vec_gather_load_p.
13489 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
13490 (vect_get_gather_scatter_offset_type): Likewise.
13491 (vect_convert_mask_for_vectype): Likewise.
13492 (vect_add_conversion_to_patterm): Likewise.
13493 (vect_try_gather_scatter_pattern): Likewise.
13494 (vect_recog_gather_scatter_pattern): New pattern recognizer.
13495 (vect_vect_recog_func_ptrs): Add it.
13496 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13497 internal_fn_mask_index and internal_gather_scatter_fn_p.
13498 (check_load_store_masking): Take the gather_scatter_info as an
13499 argument and handle gather loads.
13500 (vect_get_gather_scatter_ops): New function.
13501 (vectorizable_call): Check internal_load_fn_p.
13502 (vectorizable_load): Likewise. Handle gather load internal
13504 (vectorizable_store): Update call to check_load_store_masking.
13505 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
13506 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
13507 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
13508 (aarch64_gather_scale_operand_d): New predicates.
13509 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
13510 (mask_gather_load<mode>): New insns.
13512 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13513 Alan Hayward <alan.hayward@arm.com>
13514 David Sherwood <david.sherwood@arm.com>
13516 * optabs.def (fold_left_plus_optab): New optab.
13517 * doc/md.texi (fold_left_plus_@var{m}): Document.
13518 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
13519 * internal-fn.c (fold_left_direct): Define.
13520 (expand_fold_left_optab_fn): Likewise.
13521 (direct_fold_left_optab_supported_p): Likewise.
13522 * fold-const-call.c (fold_const_fold_left): New function.
13523 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
13524 * tree-parloops.c (valid_reduction_p): New function.
13525 (gather_scalar_reductions): Use it.
13526 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
13527 (vect_finish_replace_stmt): Declare.
13528 * tree-vect-loop.c (fold_left_reduction_fn): New function.
13529 (needs_fold_left_reduction_p): New function, split out from...
13530 (vect_is_simple_reduction): ...here. Accept reductions that
13531 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
13532 (vect_force_simple_reduction): Also store the reduction type in
13533 the assignment's STMT_VINFO_REDUC_TYPE.
13534 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
13535 (merge_with_identity): New function.
13536 (vect_expand_fold_left): Likewise.
13537 (vectorize_fold_left_reduction): Likewise.
13538 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
13539 scalar phi in place for it. Check for target support and reject
13540 cases that would reassociate the operation. Defer the transform
13541 phase to vectorize_fold_left_reduction.
13542 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
13543 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
13544 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
13546 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13548 * tree-if-conv.c (predicate_mem_writes): Remove redundant
13549 call to ifc_temp_var.
13551 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13552 Alan Hayward <alan.hayward@arm.com>
13553 David Sherwood <david.sherwood@arm.com>
13555 * target.def (legitimize_address_displacement): Take the original
13556 offset as a poly_int.
13557 * targhooks.h (default_legitimize_address_displacement): Update
13559 * targhooks.c (default_legitimize_address_displacement): Likewise.
13560 * doc/tm.texi: Regenerate.
13561 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
13562 as an argument, moving assert of ad->disp == ad->disp_term to...
13563 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
13564 Try calling targetm.legitimize_address_displacement before expanding
13565 the address rather than afterwards, and adjust for the new interface.
13566 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
13567 Match the new hook interface. Handle SVE addresses.
13568 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
13569 new hook interface.
13571 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13573 * Makefile.in (OBJS): Add early-remat.o.
13574 * target.def (select_early_remat_modes): New hook.
13575 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
13576 * doc/tm.texi: Regenerate.
13577 * targhooks.h (default_select_early_remat_modes): Declare.
13578 * targhooks.c (default_select_early_remat_modes): New function.
13579 * timevar.def (TV_EARLY_REMAT): New timevar.
13580 * passes.def (pass_early_remat): New pass.
13581 * tree-pass.h (make_pass_early_remat): Declare.
13582 * early-remat.c: New file.
13583 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
13585 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
13587 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13588 Alan Hayward <alan.hayward@arm.com>
13589 David Sherwood <david.sherwood@arm.com>
13591 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
13592 vfm1 with a bound_epilog parameter.
13593 (vect_do_peeling): Update calls accordingly, and move the prologue
13594 call earlier in the function. Treat the base bound_epilog as 0 for
13595 fully-masked loops and retain vf - 1 for other loops. Add 1 to
13596 this base when peeling for gaps.
13597 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
13598 with fully-masked loops.
13599 (vect_estimate_min_profitable_iters): Handle the single peeled
13600 iteration in that case.
13602 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13603 Alan Hayward <alan.hayward@arm.com>
13604 David Sherwood <david.sherwood@arm.com>
13606 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
13607 single-element interleaving even if the size is not a power of 2.
13608 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
13609 accesses for single-element interleaving if the group size is
13612 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13613 Alan Hayward <alan.hayward@arm.com>
13614 David Sherwood <david.sherwood@arm.com>
13616 * doc/md.texi (fold_extract_last_@var{m}): Document.
13617 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
13618 * optabs.def (fold_extract_last_optab): New optab.
13619 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
13620 * internal-fn.c (fold_extract_direct): New macro.
13621 (expand_fold_extract_optab_fn): Likewise.
13622 (direct_fold_extract_optab_supported_p): Likewise.
13623 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
13624 * tree-vect-loop.c (vect_model_reduction_cost): Handle
13625 EXTRACT_LAST_REDUCTION.
13626 (get_initial_def_for_reduction): Do not create an initial vector
13627 for EXTRACT_LAST_REDUCTION reductions.
13628 (vectorizable_reduction): Leave the scalar phi in place for
13629 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
13630 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
13631 epilogue code for EXTRACT_LAST_REDUCTION and defer the
13632 transform phase to vectorizable_condition.
13633 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
13635 (vect_finish_stmt_generation): ...here.
13636 (vect_finish_replace_stmt): New function.
13637 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
13638 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
13640 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
13642 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13643 Alan Hayward <alan.hayward@arm.com>
13644 David Sherwood <david.sherwood@arm.com>
13646 * doc/md.texi (extract_last_@var{m}): Document.
13647 * optabs.def (extract_last_optab): New optab.
13648 * internal-fn.def (EXTRACT_LAST): New internal function.
13649 * internal-fn.c (cond_unary_direct): New macro.
13650 (expand_cond_unary_optab_fn): Likewise.
13651 (direct_cond_unary_optab_supported_p): Likewise.
13652 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
13653 loops using EXTRACT_LAST.
13654 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
13655 (extract_last_<mode>): ...this optab.
13656 (vec_extract<mode><Vel>): Update accordingly.
13658 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13659 Alan Hayward <alan.hayward@arm.com>
13660 David Sherwood <david.sherwood@arm.com>
13662 * target.def (empty_mask_is_expensive): New hook.
13663 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
13664 * doc/tm.texi: Regenerate.
13665 * targhooks.h (default_empty_mask_is_expensive): Declare.
13666 * targhooks.c (default_empty_mask_is_expensive): New function.
13667 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
13668 if the target says that empty masks are expensive.
13669 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
13671 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
13673 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13674 Alan Hayward <alan.hayward@arm.com>
13675 David Sherwood <david.sherwood@arm.com>
13677 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
13678 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
13679 (vect_use_loop_mask_for_alignment_p): New function.
13680 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
13681 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
13682 niters_skip argument. Make sure that the first niters_skip elements
13683 of the first iteration are inactive.
13684 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
13685 Update call to vect_set_loop_masks_directly.
13686 (get_misalign_in_elems): New function, split out from...
13687 (vect_gen_prolog_loop_niters): ...here.
13688 (vect_update_init_of_dr): Take a code argument that specifies whether
13689 the adjustment should be added or subtracted.
13690 (vect_update_init_of_drs): Likewise.
13691 (vect_prepare_for_masked_peels): New function.
13692 (vect_do_peeling): Skip prologue peeling if we're using a mask
13693 instead. Update call to vect_update_inits_of_drs.
13694 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13696 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
13697 alignment. Do not include the number of peeled iterations in
13698 the minimum threshold in that case.
13699 (vectorizable_induction): Adjust the start value down by
13700 LOOP_VINFO_MASK_SKIP_NITERS iterations.
13701 (vect_transform_loop): Call vect_prepare_for_masked_peels.
13702 Take the number of skipped iterations into account when calculating
13704 * tree-vect-stmts.c (vect_gen_while_not): New function.
13706 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13707 Alan Hayward <alan.hayward@arm.com>
13708 David Sherwood <david.sherwood@arm.com>
13710 * doc/sourcebuild.texi (vect_fully_masked): Document.
13711 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
13712 default value to 0.
13713 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
13715 (vect_analyze_loop_2): ...here. Don't check the vectorization
13716 factor against the number of loop iterations if the loop is
13719 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13720 Alan Hayward <alan.hayward@arm.com>
13721 David Sherwood <david.sherwood@arm.com>
13723 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
13724 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
13725 (dump_groups): Update accordingly.
13726 (iv_use::mem_type): New member variable.
13727 (address_p): New function.
13728 (record_use): Add a mem_type argument and initialize the new
13730 (record_group_use): Add a mem_type argument. Use address_p.
13731 Remove obsolete null checks of base_object. Update call to record_use.
13732 (find_interesting_uses_op): Update call to record_group_use.
13733 (find_interesting_uses_cond): Likewise.
13734 (find_interesting_uses_address): Likewise.
13735 (get_mem_type_for_internal_fn): New function.
13736 (find_address_like_use): Likewise.
13737 (find_interesting_uses_stmt): Try find_address_like_use before
13738 calling find_interesting_uses_op.
13739 (addr_offset_valid_p): Use the iv mem_type field as the type
13740 of the addressed memory.
13741 (add_autoinc_candidates): Likewise.
13742 (get_address_cost): Likewise.
13743 (split_small_address_groups_p): Use address_p.
13744 (split_address_groups): Likewise.
13745 (add_iv_candidate_for_use): Likewise.
13746 (autoinc_possible_for_pair): Likewise.
13747 (rewrite_groups): Likewise.
13748 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
13749 (determine_group_iv_cost): Update after split of USE_ADDRESS.
13750 (get_alias_ptr_type_for_ptr_address): New function.
13751 (rewrite_use_address): Rewrite address uses in calls that were
13752 identified by find_address_like_use.
13754 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13755 Alan Hayward <alan.hayward@arm.com>
13756 David Sherwood <david.sherwood@arm.com>
13758 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
13760 * gimple-expr.h (is_gimple_addressable: Likewise.
13761 * gimple-expr.c (is_gimple_address): Likewise.
13762 * internal-fn.c (expand_call_mem_ref): New function.
13763 (expand_mask_load_optab_fn): Use it.
13764 (expand_mask_store_optab_fn): Likewise.
13766 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13767 Alan Hayward <alan.hayward@arm.com>
13768 David Sherwood <david.sherwood@arm.com>
13770 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
13771 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
13772 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
13773 (cond_umax@var{mode}): Document.
13774 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
13775 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
13776 (cond_umin_optab, cond_umax_optab): New optabs.
13777 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
13778 (COND_IOR, COND_XOR): New internal functions.
13779 * internal-fn.h (get_conditional_internal_fn): Declare.
13780 * internal-fn.c (cond_binary_direct): New macro.
13781 (expand_cond_binary_optab_fn): Likewise.
13782 (direct_cond_binary_optab_supported_p): Likewise.
13783 (get_conditional_internal_fn): New function.
13784 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
13785 Cope with reduction statements that are vectorized as calls rather
13787 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
13788 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
13789 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
13790 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13791 (UNSPEC_COND_EOR): New unspecs.
13792 (optab): Add mappings for them.
13793 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
13794 (sve_int_op, sve_fp_op): New int attributes.
13796 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13797 Alan Hayward <alan.hayward@arm.com>
13798 David Sherwood <david.sherwood@arm.com>
13800 * optabs.def (while_ult_optab): New optab.
13801 * doc/md.texi (while_ult@var{m}@var{n}): Document.
13802 * internal-fn.def (WHILE_ULT): New internal function.
13803 * internal-fn.h (direct_internal_fn_supported_p): New override
13804 that takes two types as argument.
13805 * internal-fn.c (while_direct): New macro.
13806 (expand_while_optab_fn): New function.
13807 (convert_optab_supported_p): Likewise.
13808 (direct_while_optab_supported_p): New macro.
13809 * wide-int.h (wi::udiv_ceil): New function.
13810 * tree-vectorizer.h (rgroup_masks): New structure.
13811 (vec_loop_masks): New typedef.
13812 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
13813 and fully_masked_p.
13814 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
13815 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
13816 (vect_max_vf): New function.
13817 (slpeel_make_loop_iterate_ntimes): Delete.
13818 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
13819 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
13820 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
13821 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
13822 internal-fn.h, stor-layout.h and optabs-query.h.
13823 (vect_set_loop_mask): New function.
13824 (add_preheader_seq): Likewise.
13825 (add_header_seq): Likewise.
13826 (interleave_supported_p): Likewise.
13827 (vect_maybe_permute_loop_masks): Likewise.
13828 (vect_set_loop_masks_directly): Likewise.
13829 (vect_set_loop_condition_masked): Likewise.
13830 (vect_set_loop_condition_unmasked): New function, split out from
13831 slpeel_make_loop_iterate_ntimes.
13832 (slpeel_make_loop_iterate_ntimes): Rename to..
13833 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
13834 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
13835 (vect_do_peeling): Update call accordingly.
13836 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
13838 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13839 mask_compare_type, can_fully_mask_p and fully_masked_p.
13840 (release_vec_loop_masks): New function.
13841 (_loop_vec_info): Use it to free the loop masks.
13842 (can_produce_all_loop_masks_p): New function.
13843 (vect_get_max_nscalars_per_iter): Likewise.
13844 (vect_verify_full_masking): Likewise.
13845 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
13846 retries, and free the mask rgroups before retrying. Check loop-wide
13847 reasons for disallowing fully-masked loops. Make the final decision
13848 about whether use a fully-masked loop or not.
13849 (vect_estimate_min_profitable_iters): Do not assume that peeling
13850 for the number of iterations will be needed for fully-masked loops.
13851 (vectorizable_reduction): Disable fully-masked loops.
13852 (vectorizable_live_operation): Likewise.
13853 (vect_halve_mask_nunits): New function.
13854 (vect_double_mask_nunits): Likewise.
13855 (vect_record_loop_mask): Likewise.
13856 (vect_get_loop_mask): Likewise.
13857 (vect_transform_loop): Handle the case in which the final loop
13858 iteration might handle a partial vector. Call vect_set_loop_condition
13859 instead of slpeel_make_loop_iterate_ntimes.
13860 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
13861 (check_load_store_masking): New function.
13862 (prepare_load_store_mask): Likewise.
13863 (vectorizable_store): Handle fully-masked loops.
13864 (vectorizable_load): Likewise.
13865 (supportable_widening_operation): Use vect_halve_mask_nunits for
13867 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
13868 (vect_gen_while): New function.
13869 * config/aarch64/aarch64.md (umax<mode>3): New expander.
13870 (aarch64_uqdec<mode>): New insn.
13872 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13873 Alan Hayward <alan.hayward@arm.com>
13874 David Sherwood <david.sherwood@arm.com>
13876 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
13877 (reduc_xor_scal_optab): New optabs.
13878 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
13879 (reduc_xor_scal_@var{m}): Document.
13880 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
13881 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
13882 internal functions.
13883 * fold-const-call.c (fold_const_call): Handle them.
13884 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
13885 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
13886 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
13887 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
13888 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
13889 (UNSPEC_XORV): New unspecs.
13890 (optab): Add entries for them.
13891 (BITWISEV): New int iterator.
13892 (bit_reduc_op): New int attributes.
13894 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13895 Alan Hayward <alan.hayward@arm.com>
13896 David Sherwood <david.sherwood@arm.com>
13898 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
13899 * internal-fn.def (VEC_SHL_INSERT): New internal function.
13900 * optabs.def (vec_shl_insert_optab): New optab.
13901 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
13902 (duplicate_and_interleave): Likewise.
13903 * tree-vect-loop.c: Include internal-fn.h.
13904 (neutral_op_for_slp_reduction): New function, split out from
13905 get_initial_defs_for_reduction.
13906 (get_initial_def_for_reduction): Handle option 2 for variable-length
13907 vectors by loading the neutral value into a vector and then shifting
13908 the initial value into element 0.
13909 (get_initial_defs_for_reduction): Replace the code argument with
13910 the neutral value calculated by neutral_op_for_slp_reduction.
13911 Use gimple_build_vector for constant-length vectors.
13912 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
13913 but the first group_size elements have a neutral value.
13914 Use duplicate_and_interleave otherwise.
13915 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
13916 Update call to get_initial_defs_for_reduction. Handle SLP
13917 reductions for variable-length vectors by creating one vector
13918 result for each scalar result, with the elements associated
13919 with other scalar results stubbed out with the neutral value.
13920 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
13921 Require IFN_VEC_SHL_INSERT for double reductions on
13922 variable-length vectors, or SLP reductions that have
13923 a neutral value. Require can_duplicate_and_interleave_p
13924 support for variable-length unchained SLP reductions if there
13925 is no neutral value, such as for MIN/MAX reductions. Also require
13926 the number of vector elements to be a multiple of the number of
13927 SLP statements when doing variable-length unchained SLP reductions.
13928 Update call to vect_create_epilog_for_reduction.
13929 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
13930 and remove initial values.
13931 (duplicate_and_interleave): Make public.
13932 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
13933 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
13935 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13936 Alan Hayward <alan.hayward@arm.com>
13937 David Sherwood <david.sherwood@arm.com>
13939 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
13940 (can_duplicate_and_interleave_p): New function.
13941 (vect_get_and_check_slp_defs): Take the vector of statements
13942 rather than just the current one. Remove excess parentheses.
13943 Restriction rejectinon of vect_constant_def and vect_external_def
13944 for variable-length vectors to boolean types, or types for which
13945 can_duplicate_and_interleave_p is false.
13946 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
13947 (duplicate_and_interleave): New function.
13948 (vect_get_constant_vectors): Use gimple_build_vector for
13949 constant-length vectors and suitable variable-length constant
13950 vectors. Use duplicate_and_interleave for other variable-length
13951 vectors. Don't defer the update when inserting new statements.
13953 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13954 Alan Hayward <alan.hayward@arm.com>
13955 David Sherwood <david.sherwood@arm.com>
13957 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
13958 min_profitable_iters doesn't go negative.
13960 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13961 Alan Hayward <alan.hayward@arm.com>
13962 David Sherwood <david.sherwood@arm.com>
13964 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
13965 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
13966 * optabs.def (vec_mask_load_lanes_optab): New optab.
13967 (vec_mask_store_lanes_optab): Likewise.
13968 * internal-fn.def (MASK_LOAD_LANES): New internal function.
13969 (MASK_STORE_LANES): Likewise.
13970 * internal-fn.c (mask_load_lanes_direct): New macro.
13971 (mask_store_lanes_direct): Likewise.
13972 (expand_mask_load_optab_fn): Handle masked operations.
13973 (expand_mask_load_lanes_optab_fn): New macro.
13974 (expand_mask_store_optab_fn): Handle masked operations.
13975 (expand_mask_store_lanes_optab_fn): New macro.
13976 (direct_mask_load_lanes_optab_supported_p): Likewise.
13977 (direct_mask_store_lanes_optab_supported_p): Likewise.
13978 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
13980 (vect_load_lanes_supported): Likewise.
13981 * tree-vect-data-refs.c (strip_conversion): New function.
13982 (can_group_stmts_p): Likewise.
13983 (vect_analyze_data_ref_accesses): Use it instead of checking
13984 for a pair of assignments.
13985 (vect_store_lanes_supported): Take a masked_p parameter.
13986 (vect_load_lanes_supported): Likewise.
13987 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
13988 vect_store_lanes_supported and vect_load_lanes_supported.
13989 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13990 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
13991 parameter. Don't allow gaps for masked accesses.
13992 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
13993 and vect_load_lanes_supported.
13994 (get_load_store_type): Take a masked_p parameter and update
13995 call to get_group_load_store_type.
13996 (vectorizable_store): Update call to get_load_store_type.
13997 Handle IFN_MASK_STORE_LANES.
13998 (vectorizable_load): Update call to get_load_store_type.
13999 Handle IFN_MASK_LOAD_LANES.
14001 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14002 Alan Hayward <alan.hayward@arm.com>
14003 David Sherwood <david.sherwood@arm.com>
14005 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
14007 * config/aarch64/aarch64-protos.h
14008 (aarch64_sve_struct_memory_operand_p): Declare.
14009 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
14010 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
14011 (VPRED, vpred): Handle SVE structure modes.
14012 * config/aarch64/constraints.md (Utx): New constraint.
14013 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
14014 (aarch64_sve_struct_nonimmediate_operand): New predicates.
14015 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
14016 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
14017 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
14018 structure modes. Split into pieces after RA.
14019 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
14020 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
14022 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
14023 SVE structure modes.
14024 (aarch64_classify_address): Likewise.
14025 (sizetochar): Move earlier in file.
14026 (aarch64_print_operand): Handle SVE register lists.
14027 (aarch64_array_mode): New function.
14028 (aarch64_sve_struct_memory_operand_p): Likewise.
14029 (TARGET_ARRAY_MODE): Redefine.
14031 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14032 Alan Hayward <alan.hayward@arm.com>
14033 David Sherwood <david.sherwood@arm.com>
14035 * target.def (array_mode): New target hook.
14036 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
14037 * doc/tm.texi: Regenerate.
14038 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
14039 * hooks.c (hook_optmode_mode_uhwi_none): New function.
14040 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
14041 targetm.array_mode.
14042 * stor-layout.c (mode_for_array): Likewise. Support polynomial
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 * fold-const.c (fold_binary_loc): Check the argument types
14050 rather than the result type when testing for a vector operation.
14052 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14054 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
14055 * doc/tm.texi: Regenerate.
14057 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14058 Alan Hayward <alan.hayward@arm.com>
14059 David Sherwood <david.sherwood@arm.com>
14061 * doc/invoke.texi (-msve-vector-bits=): Document new option.
14062 (sve): Document new AArch64 extension.
14063 * doc/md.texi (w): Extend the description of the AArch64
14064 constraint to include SVE vectors.
14065 (Upl, Upa): Document new AArch64 predicate constraints.
14066 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
14068 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
14069 (msve-vector-bits=): New option.
14070 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
14071 SVE when these are disabled.
14072 (sve): New extension.
14073 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
14074 modes. Adjust their number of units based on aarch64_sve_vg.
14075 (MAX_BITSIZE_MODE_ANY_MODE): Define.
14076 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
14077 aarch64_addr_query_type.
14078 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
14079 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
14080 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
14081 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
14082 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
14083 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
14084 (aarch64_simd_imm_zero_p): Delete.
14085 (aarch64_check_zero_based_sve_index_immediate): Declare.
14086 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14087 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14088 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14089 (aarch64_sve_float_mul_immediate_p): Likewise.
14090 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14091 rather than an rtx.
14092 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
14093 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
14094 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
14095 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
14096 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
14097 (aarch64_regmode_natural_size): Likewise.
14098 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
14099 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
14101 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
14102 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
14103 for VG and the SVE predicate registers.
14104 (V_ALIASES): Add a "z"-prefixed alias.
14105 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
14106 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
14107 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
14108 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
14109 (REG_CLASS_NAMES): Add entries for them.
14110 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
14111 and the predicate registers.
14112 (aarch64_sve_vg): Declare.
14113 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
14114 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
14115 (REGMODE_NATURAL_SIZE): Define.
14116 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
14118 * config/aarch64/aarch64.c: Include cfgrtl.h.
14119 (simd_immediate_info): Add a constructor for series vectors,
14120 and an associated step field.
14121 (aarch64_sve_vg): New variable.
14122 (aarch64_dbx_register_number): Handle VG and the predicate registers.
14123 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
14124 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
14125 (VEC_ANY_DATA, VEC_STRUCT): New constants.
14126 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
14127 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
14128 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
14129 (aarch64_get_mask_mode): New functions.
14130 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
14131 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
14132 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
14133 predicate modes and predicate registers. Explicitly restrict
14134 GPRs to modes of 16 bytes or smaller. Only allow FP registers
14135 to store a vector mode if it is recognized by
14136 aarch64_classify_vector_mode.
14137 (aarch64_regmode_natural_size): New function.
14138 (aarch64_hard_regno_caller_save_mode): Return the original mode
14140 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
14141 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
14142 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
14143 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
14145 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
14146 does not overlap dest if the function is frame-related. Handle
14148 (aarch64_split_add_offset): New function.
14149 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
14150 them aarch64_add_offset.
14151 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
14152 and update call to aarch64_sub_sp.
14153 (aarch64_add_cfa_expression): New function.
14154 (aarch64_expand_prologue): Pass extra temporary registers to the
14155 functions above. Handle the case in which we need to emit new
14156 DW_CFA_expressions for registers that were originally saved
14157 relative to the stack pointer, but now have to be expressed
14158 relative to the frame pointer.
14159 (aarch64_output_mi_thunk): Pass extra temporary registers to the
14161 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
14162 IP0 and IP1 values for SVE frames.
14163 (aarch64_expand_vec_series): New function.
14164 (aarch64_expand_sve_widened_duplicate): Likewise.
14165 (aarch64_expand_sve_const_vector): Likewise.
14166 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
14167 Handle SVE constants. Use emit_move_insn to move a force_const_mem
14168 into the register, rather than emitting a SET directly.
14169 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
14170 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
14171 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
14172 (offset_9bit_signed_scaled_p): New functions.
14173 (aarch64_replicate_bitmask_imm): New function.
14174 (aarch64_bitmask_imm): Use it.
14175 (aarch64_cannot_force_const_mem): Reject expressions involving
14176 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
14177 (aarch64_classify_index): Handle SVE indices, by requiring
14178 a plain register index with a scale that matches the element size.
14179 (aarch64_classify_address): Handle SVE addresses. Assert that
14180 the mode of the address is VOIDmode or an integer mode.
14181 Update call to aarch64_classify_symbol.
14182 (aarch64_classify_symbolic_expression): Update call to
14183 aarch64_classify_symbol.
14184 (aarch64_const_vec_all_in_range_p): New function.
14185 (aarch64_print_vector_float_operand): Likewise.
14186 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
14187 "vN" for FP registers with SVE modes. Handle (const ...) vectors
14188 and the FP immediates 1.0 and 0.5.
14189 (aarch64_print_address_internal): Handle SVE addresses.
14190 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
14191 (aarch64_regno_regclass): Handle predicate registers.
14192 (aarch64_secondary_reload): Handle big-endian reloads of SVE
14194 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
14195 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
14196 (aarch64_convert_sve_vector_bits): New function.
14197 (aarch64_override_options): Use it to handle -msve-vector-bits=.
14198 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14199 rather than an rtx.
14200 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
14201 Handle SVE vector and predicate modes. Accept VL-based constants
14202 that need only one temporary register, and VL offsets that require
14203 no temporary registers.
14204 (aarch64_conditional_register_usage): Mark the predicate registers
14205 as fixed if SVE isn't available.
14206 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
14207 Return true for SVE vector and predicate modes.
14208 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
14209 rather than an unsigned int. Handle SVE modes.
14210 (aarch64_preferred_simd_mode): Update call accordingly. Handle
14212 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
14214 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14215 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14216 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14217 (aarch64_sve_float_mul_immediate_p): New functions.
14218 (aarch64_sve_valid_immediate): New function.
14219 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
14220 Explicitly reject structure modes. Check for INDEX constants.
14221 Handle PTRUE and PFALSE constants.
14222 (aarch64_check_zero_based_sve_index_immediate): New function.
14223 (aarch64_simd_imm_zero_p): Delete.
14224 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
14225 vector modes. Accept constants in the range of CNT[BHWD].
14226 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
14227 ask for an Advanced SIMD mode.
14228 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
14229 (aarch64_simd_vector_alignment): Handle SVE predicates.
14230 (aarch64_vectorize_preferred_vector_alignment): New function.
14231 (aarch64_simd_vector_alignment_reachable): Use it instead of
14233 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
14234 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
14236 (MAX_VECT_LEN): Delete.
14237 (expand_vec_perm_d): Add a vec_flags field.
14238 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
14239 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14240 (aarch64_evpc_ext): Don't apply a big-endian lane correction
14242 (aarch64_evpc_rev): Rename to...
14243 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
14244 (aarch64_evpc_rev_global): New function.
14245 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
14246 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
14248 (aarch64_evpc_sve_tbl): New function.
14249 (aarch64_expand_vec_perm_const_1): Update after rename of
14250 aarch64_evpc_rev. Handle SVE permutes too, trying
14251 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
14252 than aarch64_evpc_tbl.
14253 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
14254 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
14255 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
14256 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
14257 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
14258 (aarch64_expand_sve_vcond): New functions.
14259 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
14260 of aarch64_vector_mode_p.
14261 (aarch64_dwarf_poly_indeterminate_value): New function.
14262 (aarch64_compute_pressure_classes): Likewise.
14263 (aarch64_can_change_mode_class): Likewise.
14264 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
14265 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
14266 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
14267 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
14268 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
14269 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
14270 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
14271 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
14273 (Dn, Dl, Dr): Accept const as well as const_vector.
14274 (Dz): Likewise. Compare against CONST0_RTX.
14275 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
14276 of "vector" where appropriate.
14277 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
14278 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
14279 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
14280 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
14281 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
14282 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
14283 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
14284 (v_int_equiv): Extend to SVE modes.
14285 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
14287 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
14288 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
14289 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
14290 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
14291 (SVE_COND_FP_CMP): New int iterators.
14292 (perm_hilo): Handle the new unpack unspecs.
14293 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
14295 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
14296 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
14297 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
14298 (aarch64_equality_operator, aarch64_constant_vector_operand)
14299 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
14300 (aarch64_sve_nonimmediate_operand): Likewise.
14301 (aarch64_sve_general_operand): Likewise.
14302 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
14303 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
14304 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
14305 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
14306 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
14307 (aarch64_sve_float_arith_immediate): Likewise.
14308 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
14309 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
14310 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
14311 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
14312 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
14313 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
14314 (aarch64_sve_float_arith_operand): Likewise.
14315 (aarch64_sve_float_arith_with_sub_operand): Likewise.
14316 (aarch64_sve_float_mul_operand): Likewise.
14317 (aarch64_sve_vec_perm_operand): Likewise.
14318 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
14319 (aarch64_mov_operand): Accept const_poly_int and const_vector.
14320 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
14321 as well as const_vector.
14322 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
14323 in file. Use CONST0_RTX and CONSTM1_RTX.
14324 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
14325 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
14326 Use aarch64_simd_imm_zero.
14327 * config/aarch64/aarch64-sve.md: New file.
14328 * config/aarch64/aarch64.md: Include it.
14329 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
14330 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
14331 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
14332 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
14333 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
14334 (sve): New attribute.
14335 (enabled): Disable instructions with the sve attribute unless
14337 (movqi, movhi): Pass CONST_POLY_INT operaneds through
14338 aarch64_expand_mov_immediate.
14339 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
14340 CNT[BHSD] immediates.
14341 (movti): Split CONST_POLY_INT moves into two halves.
14342 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
14343 Split additions that need a temporary here if the destination
14344 is the stack pointer.
14345 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
14346 (*add<mode>3_poly_1): New instruction.
14347 (set_clobber_cc): New expander.
14349 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14351 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
14352 parameter and use it instead of GET_MODE_SIZE (innermode). Use
14353 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
14354 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
14355 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
14356 Change innermode from fixed_mode_size to machine_mode.
14357 (simplify_subreg): Update call accordingly. Handle a constant-sized
14358 subreg of a variable-length CONST_VECTOR.
14360 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14361 Alan Hayward <alan.hayward@arm.com>
14362 David Sherwood <david.sherwood@arm.com>
14364 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
14365 (add_offset_to_base): New function, split out from...
14366 (create_mem_ref): ...here. When handling a scale other than 1,
14367 check first whether the address is valid without the offset.
14368 Add it into the base if so, leaving the index and scale as-is.
14370 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14373 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
14374 fold_for_warn before checking if arg2 is INTEGER_CST.
14376 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
14378 * config/rs6000/predicates.md (load_multiple_operation): Delete.
14379 (store_multiple_operation): Delete.
14380 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
14381 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
14382 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
14383 guarded by TARGET_STRING.
14384 (rs6000_output_load_multiple): Delete.
14385 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
14386 OPTION_MASK_STRING / TARGET_STRING handling.
14387 (print_operand) <'N', 'O'>: Add comment that these are unused now.
14388 (const rs6000_opt_masks) <"string">: Change mask to 0.
14389 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
14390 (MASK_STRING): Delete.
14391 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
14393 (load_multiple): Delete.
14400 (store_multiple): Delete.
14407 (movmemsi_8reg): Delete.
14408 (corresponding unnamed define_insn): Delete.
14409 (movmemsi_6reg): Delete.
14410 (corresponding unnamed define_insn): Delete.
14411 (movmemsi_4reg): Delete.
14412 (corresponding unnamed define_insn): Delete.
14413 (movmemsi_2reg): Delete.
14414 (corresponding unnamed define_insn): Delete.
14415 (movmemsi_1reg): Delete.
14416 (corresponding unnamed define_insn): Delete.
14417 * config/rs6000/rs6000.opt (mno-string): New.
14418 (mstring): Replace by deprecation warning stub.
14419 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
14421 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14423 * regrename.c (regrename_do_replace): If replacing the same
14424 reg multiple times, try to reuse last created gen_raw_REG.
14427 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
14428 main to workaround a bug in GDB.
14430 2018-01-12 Tom de Vries <tom@codesourcery.com>
14433 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
14435 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
14437 PR rtl-optimization/80481
14438 * ira-color.c (get_cap_member): New function.
14439 (allocnos_conflict_by_live_ranges_p): Use it.
14440 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
14441 (setup_slot_coalesced_allocno_live_ranges): Ditto.
14443 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
14446 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
14447 (*saddl_se_1): Ditto.
14448 (*ssubsi_1): Ditto.
14449 (*ssubl_se_1): Ditto.
14451 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14453 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
14454 rather than wi::to_widest for DR_INITs.
14455 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
14456 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
14457 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
14459 (vect_analyze_group_access_1): Note that here.
14461 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14463 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
14464 polynomial type sizes.
14466 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14468 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
14469 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
14470 (gimple_add_tmp_var): Likewise.
14472 2018-01-12 Martin Liska <mliska@suse.cz>
14474 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
14475 (gimple_alloc_sizes): Likewise.
14476 (dump_gimple_statistics): Use PRIu64 in printf format.
14477 * gimple.h: Change uint64_t to int.
14479 2018-01-12 Martin Liska <mliska@suse.cz>
14481 * tree-core.h: Use uint64_t instead of int.
14482 * tree.c (tree_node_counts): Likewise.
14483 (tree_node_sizes): Likewise.
14484 (dump_tree_statistics): Use PRIu64 in printf format.
14486 2018-01-12 Martin Liska <mliska@suse.cz>
14488 * Makefile.in: As qsort_chk is implemented in vec.c, add
14489 vec.o to linkage of gencfn-macros.
14490 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
14491 passing the info to record_node_allocation_statistics.
14492 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
14494 * ggc-common.c (struct ggc_usage): Add operator== and use
14495 it in operator< and compare function.
14496 * mem-stats.h (struct mem_usage): Likewise.
14497 * vec.c (struct vec_usage): Remove operator< and compare
14498 function. Can be simply inherited.
14500 2018-01-12 Martin Jambor <mjambor@suse.cz>
14503 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
14504 * tree-ssa-math-opts.c: Include domwalk.h.
14505 (convert_mult_to_fma_1): New function.
14506 (fma_transformation_info): New type.
14507 (fma_deferring_state): Likewise.
14508 (cancel_fma_deferring): New function.
14509 (result_of_phi): Likewise.
14510 (last_fma_candidate_feeds_initial_phi): Likewise.
14511 (convert_mult_to_fma): Added deferring logic, split actual
14512 transformation to convert_mult_to_fma_1.
14513 (math_opts_dom_walker): New type.
14514 (math_opts_dom_walker::after_dom_children): New method, body moved
14515 here from pass_optimize_widening_mul::execute, added deferring logic
14517 (pass_optimize_widening_mul::execute): Moved most of code to
14518 math_opts_dom_walker::after_dom_children.
14519 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
14520 * config/i386/i386.c (ix86_option_override_internal): Added
14521 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
14523 2018-01-12 Richard Biener <rguenther@suse.de>
14526 * dwarf2out.c (gen_variable_die): Do not reset old_die for
14527 inline instance vars.
14529 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
14532 * config/rx/rx.c (rx_is_restricted_memory_address):
14533 Handle SUBREG case.
14535 2018-01-12 Richard Biener <rguenther@suse.de>
14537 PR tree-optimization/80846
14538 * target.def (split_reduction): New target hook.
14539 * targhooks.c (default_split_reduction): New function.
14540 * targhooks.h (default_split_reduction): Declare.
14541 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
14542 target requests first reduce vectors by combining low and high
14544 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
14545 (get_vectype_for_scalar_type_and_size): Export.
14546 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
14547 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
14548 * doc/tm.texi: Regenerate.
14549 * config/i386/i386.c (ix86_split_reduction): Implement
14550 TARGET_VECTORIZE_SPLIT_REDUCTION.
14552 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14555 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
14556 in PIC mode except for TARGET_VXWORKS_RTP.
14557 * config/sparc/sparc.c: Include cfgrtl.h.
14558 (TARGET_INIT_PIC_REG): Define.
14559 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
14560 (sparc_pic_register_p): New predicate.
14561 (sparc_legitimate_address_p): Use it.
14562 (sparc_legitimize_pic_address): Likewise.
14563 (sparc_delegitimize_address): Likewise.
14564 (sparc_mode_dependent_address_p): Likewise.
14565 (gen_load_pcrel_sym): Remove 4th parameter.
14566 (load_got_register): Adjust call to above. Remove obsolete stuff.
14567 (sparc_expand_prologue): Do not call load_got_register here.
14568 (sparc_flat_expand_prologue): Likewise.
14569 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
14570 (sparc_use_pseudo_pic_reg): New function.
14571 (sparc_init_pic_reg): Likewise.
14572 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14573 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
14575 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
14577 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
14578 Add item for branch_cost.
14580 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14582 PR rtl-optimization/83565
14583 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
14584 not extend the result to a larger mode for rotate operations.
14585 (num_sign_bit_copies1): Likewise.
14587 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14590 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
14592 Use values-Xc.o for -pedantic.
14593 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
14595 2018-01-12 Martin Liska <mliska@suse.cz>
14598 * ipa-devirt.c (final_warning_record::grow_type_warnings):
14600 (possible_polymorphic_call_targets): Use it.
14601 (ipa_devirt): Likewise.
14603 2018-01-12 Martin Liska <mliska@suse.cz>
14605 * profile-count.h (enum profile_quality): Use 0 as invalid
14606 enum value of profile_quality.
14608 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
14610 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
14611 -mext-string options.
14613 2018-01-12 Richard Biener <rguenther@suse.de>
14615 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
14616 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
14617 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
14619 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
14621 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
14623 * configure.ac (--with-long-double-format): Add support for the
14624 configuration option to change the default long double format on
14626 * config.gcc (powerpc*-linux*-*): Likewise.
14627 * configure: Regenerate.
14628 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
14629 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
14630 used without modification.
14632 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14634 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
14635 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
14636 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
14637 MISC_BUILTIN_SPEC_BARRIER.
14638 (rs6000_init_builtins): Likewise.
14639 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
14641 (speculation_barrier): New define_insn.
14642 * doc/extend.texi: Document __builtin_speculation_barrier.
14644 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14647 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
14648 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
14649 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
14651 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
14652 integral modes instead of "ss" and "sd".
14653 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
14654 vectors with 32-bit and 64-bit elements.
14655 (vecdupssescalarmodesuffix): New mode attribute.
14656 (vec_dup<mode>): Use it.
14658 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
14661 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
14662 frame if argument is passed on stack.
14664 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14667 * ree.c (combine_reaching_defs): Optimize also
14668 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
14669 reg2=any_extend(exp); reg1=reg2;, formatting fix.
14671 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14673 PR middle-end/83189
14674 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
14676 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14678 PR middle-end/83718
14679 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
14680 after they are computed.
14682 2018-01-11 Bin Cheng <bin.cheng@arm.com>
14684 PR tree-optimization/83695
14685 * gimple-loop-linterchange.cc
14686 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
14687 reset cached scev information after interchange.
14688 (pass_linterchange::execute): Remove call to scev_reset_htab.
14690 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14692 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
14693 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
14694 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
14695 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
14696 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
14697 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
14698 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
14699 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
14700 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
14701 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
14702 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
14703 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
14704 (V_lane_reg): Likewise.
14705 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
14707 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
14708 (vfmal_lane_low<mode>_intrinsic,
14709 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
14710 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
14711 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
14712 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
14713 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
14714 vfmsl_lane_high<mode>_intrinsic): New define_insns.
14716 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14718 * config/arm/arm-cpus.in (fp16fml): New feature.
14719 (ALL_SIMD): Add fp16fml.
14720 (armv8.2-a): Add fp16fml as an option.
14721 (armv8.3-a): Likewise.
14722 (armv8.4-a): Add fp16fml as part of fp16.
14723 * config/arm/arm.h (TARGET_FP16FML): Define.
14724 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
14726 * config/arm/arm-modes.def (V2HF): Define.
14727 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
14728 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
14729 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
14730 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
14731 vfmsl_low, vfmsl_high): New set of builtins.
14732 * config/arm/iterators.md (PLUSMINUS): New code iterator.
14733 (vfml_op): New code attribute.
14734 (VFMLHALVES): New int iterator.
14735 (VFML, VFMLSEL): New mode attributes.
14736 (V_reg): Define mapping for V2HF.
14737 (V_hi, V_lo): New mode attributes.
14738 (VF_constraint): Likewise.
14739 (vfml_half, vfml_half_selector): New int attributes.
14740 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
14742 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
14743 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
14745 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
14746 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
14747 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
14748 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
14750 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
14751 Document new effective target and option set.
14753 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14755 * config/arm/arm-cpus.in (armv8_4): New feature.
14756 (ARMv8_4a): New fgroup.
14757 (armv8.4-a): New arch.
14758 * config/arm/arm-tables.opt: Regenerate.
14759 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
14760 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
14761 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
14762 Add matching rules for -march=armv8.4-a and extensions.
14763 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
14765 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
14768 * config/rx/rx.md (BW): New mode attribute.
14769 (sync_lock_test_and_setsi): Add mode suffix to insn output.
14771 2018-01-11 Richard Biener <rguenther@suse.de>
14773 PR tree-optimization/83435
14774 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
14775 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
14776 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
14778 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14779 Alan Hayward <alan.hayward@arm.com>
14780 David Sherwood <david.sherwood@arm.com>
14782 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
14784 (aarch64_classify_address): Initialize it. Track polynomial offsets.
14785 (aarch64_print_address_internal): Use it to check for a zero offset.
14787 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14788 Alan Hayward <alan.hayward@arm.com>
14789 David Sherwood <david.sherwood@arm.com>
14791 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
14792 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
14793 Return a poly_int64 rather than a HOST_WIDE_INT.
14794 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
14795 rather than a HOST_WIDE_INT.
14796 * config/aarch64/aarch64.h (aarch64_frame): Protect with
14797 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
14798 hard_fp_offset, frame_size, initial_adjust, callee_offset and
14799 final_offset from HOST_WIDE_INT to poly_int64.
14800 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
14801 to_constant when getting the number of units in an Advanced SIMD
14803 (aarch64_builtin_vectorized_function): Check for a constant number
14805 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
14807 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
14808 attribute instead of GET_MODE_NUNITS.
14809 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
14810 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
14811 GET_MODE_SIZE for fixed-size registers.
14812 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
14813 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
14814 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
14815 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
14816 (aarch64_print_operand, aarch64_print_address_internal)
14817 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
14818 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
14819 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
14820 Handle polynomial GET_MODE_SIZE.
14821 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
14822 wider than SImode without modification.
14823 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
14824 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
14825 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
14826 passing and returning SVE modes.
14827 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
14828 rather than GEN_INT.
14829 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
14830 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
14831 (aarch64_allocate_and_probe_stack_space): Likewise.
14832 (aarch64_layout_frame): Cope with polynomial offsets.
14833 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
14834 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
14835 polynomial offsets.
14836 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
14837 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
14838 poly_int64 rather than a HOST_WIDE_INT.
14839 (aarch64_get_separate_components, aarch64_process_components)
14840 (aarch64_expand_prologue, aarch64_expand_epilogue)
14841 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
14842 (aarch64_anchor_offset): New function, split out from...
14843 (aarch64_legitimize_address): ...here.
14844 (aarch64_builtin_vectorization_cost): Handle polynomial
14845 TYPE_VECTOR_SUBPARTS.
14846 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
14848 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
14849 number of elements from the PARALLEL rather than the mode.
14850 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
14851 rather than GET_MODE_BITSIZE.
14852 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
14853 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
14854 (aarch64_expand_vec_perm_const_1): Handle polynomial
14855 d->perm.length () and d->perm elements.
14856 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
14857 Apply to_constant to d->perm elements.
14858 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
14859 polynomial CONST_VECTOR_NUNITS.
14860 (aarch64_move_pointer): Take amount as a poly_int64 rather
14862 (aarch64_progress_pointer): Avoid temporary variable.
14863 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
14864 the mode attribute instead of GET_MODE.
14866 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14867 Alan Hayward <alan.hayward@arm.com>
14868 David Sherwood <david.sherwood@arm.com>
14870 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
14871 x exists before using it.
14872 (aarch64_add_constant_internal): Rename to...
14873 (aarch64_add_offset_1): ...this. Replace regnum with separate
14874 src and dest rtxes. Handle the case in which they're different,
14875 including when the offset is zero. Replace scratchreg with an rtx.
14876 Use 2 additions if there is no spare register into which we can
14877 move a 16-bit constant.
14878 (aarch64_add_constant): Delete.
14879 (aarch64_add_offset): Replace reg with separate src and dest
14880 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
14881 Use aarch64_add_offset_1.
14882 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
14883 an rtx rather than an int. Take the delta as a poly_int64
14884 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
14885 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
14886 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
14887 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
14888 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
14889 and aarch64_add_sp.
14890 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
14891 aarch64_add_constant.
14893 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14895 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
14896 Use scalar_float_mode.
14898 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14900 * config/aarch64/aarch64-simd.md
14901 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
14902 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
14903 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
14904 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
14905 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
14906 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
14907 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
14908 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
14909 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
14910 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
14912 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14915 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
14916 targ_options->x_arm_arch_string is non NULL.
14918 2018-01-11 Tamar Christina <tamar.christina@arm.com>
14920 * config/aarch64/aarch64.h
14921 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
14923 2018-01-11 Sudakshina Das <sudi.das@arm.com>
14926 * expmed.c (emit_store_flag_force): Swap if const op0
14927 and change VOIDmode to mode of op0.
14929 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14931 PR rtl-optimization/83761
14932 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
14933 than bytes to mode_for_size.
14935 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14937 PR middle-end/83189
14938 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
14939 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
14942 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14944 PR middle-end/83575
14945 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
14946 when in layout mode.
14947 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
14948 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
14951 2018-01-10 Michael Collison <michael.collison@arm.com>
14953 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
14954 * config/aarch64/aarch64-option-extension.def: Add
14955 AARCH64_OPT_EXTENSION of 'fp16fml'.
14956 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14957 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
14958 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
14959 * config/aarch64/constraints.md (Ui7): New constraint.
14960 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
14961 (VFMLA_SEL_W): Ditto.
14964 (VFMLA16_LOW): New int iterator.
14965 (VFMLA16_HIGH): Ditto.
14966 (UNSPEC_FMLAL): New unspec.
14967 (UNSPEC_FMLSL): Ditto.
14968 (UNSPEC_FMLAL2): Ditto.
14969 (UNSPEC_FMLSL2): Ditto.
14970 (f16mac): New code attribute.
14971 * config/aarch64/aarch64-simd-builtins.def
14972 (aarch64_fmlal_lowv2sf): Ditto.
14973 (aarch64_fmlsl_lowv2sf): Ditto.
14974 (aarch64_fmlalq_lowv4sf): Ditto.
14975 (aarch64_fmlslq_lowv4sf): Ditto.
14976 (aarch64_fmlal_highv2sf): Ditto.
14977 (aarch64_fmlsl_highv2sf): Ditto.
14978 (aarch64_fmlalq_highv4sf): Ditto.
14979 (aarch64_fmlslq_highv4sf): Ditto.
14980 (aarch64_fmlal_lane_lowv2sf): Ditto.
14981 (aarch64_fmlsl_lane_lowv2sf): Ditto.
14982 (aarch64_fmlal_laneq_lowv2sf): Ditto.
14983 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
14984 (aarch64_fmlalq_lane_lowv4sf): Ditto.
14985 (aarch64_fmlsl_lane_lowv4sf): Ditto.
14986 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
14987 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
14988 (aarch64_fmlal_lane_highv2sf): Ditto.
14989 (aarch64_fmlsl_lane_highv2sf): Ditto.
14990 (aarch64_fmlal_laneq_highv2sf): Ditto.
14991 (aarch64_fmlsl_laneq_highv2sf): Ditto.
14992 (aarch64_fmlalq_lane_highv4sf): Ditto.
14993 (aarch64_fmlsl_lane_highv4sf): Ditto.
14994 (aarch64_fmlalq_laneq_highv4sf): Ditto.
14995 (aarch64_fmlsl_laneq_highv4sf): Ditto.
14996 * config/aarch64/aarch64-simd.md:
14997 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
14998 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14999 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
15000 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15001 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
15002 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
15003 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
15004 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
15005 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
15006 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
15007 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
15008 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
15009 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
15010 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
15011 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
15012 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
15013 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
15014 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
15015 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
15016 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
15017 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
15018 (vfmlsl_low_u32): Ditto.
15019 (vfmlalq_low_u32): Ditto.
15020 (vfmlslq_low_u32): Ditto.
15021 (vfmlal_high_u32): Ditto.
15022 (vfmlsl_high_u32): Ditto.
15023 (vfmlalq_high_u32): Ditto.
15024 (vfmlslq_high_u32): Ditto.
15025 (vfmlal_lane_low_u32): Ditto.
15026 (vfmlsl_lane_low_u32): Ditto.
15027 (vfmlal_laneq_low_u32): Ditto.
15028 (vfmlsl_laneq_low_u32): Ditto.
15029 (vfmlalq_lane_low_u32): Ditto.
15030 (vfmlslq_lane_low_u32): Ditto.
15031 (vfmlalq_laneq_low_u32): Ditto.
15032 (vfmlslq_laneq_low_u32): Ditto.
15033 (vfmlal_lane_high_u32): Ditto.
15034 (vfmlsl_lane_high_u32): Ditto.
15035 (vfmlal_laneq_high_u32): Ditto.
15036 (vfmlsl_laneq_high_u32): Ditto.
15037 (vfmlalq_lane_high_u32): Ditto.
15038 (vfmlslq_lane_high_u32): Ditto.
15039 (vfmlalq_laneq_high_u32): Ditto.
15040 (vfmlslq_laneq_high_u32): Ditto.
15041 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
15042 (AARCH64_FL_FOR_ARCH8_4): New.
15043 (AARCH64_ISA_F16FML): New ISA flag.
15044 (TARGET_F16FML): New feature flag for fp16fml.
15045 (doc/invoke.texi): Document new fp16fml option.
15047 2018-01-10 Michael Collison <michael.collison@arm.com>
15049 * config/aarch64/aarch64-builtins.c:
15050 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
15051 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15052 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
15053 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
15054 (AARCH64_ISA_SHA3): New ISA flag.
15055 (TARGET_SHA3): New feature flag for sha3.
15056 * config/aarch64/iterators.md (sha512_op): New int attribute.
15057 (CRYPTO_SHA512): New int iterator.
15058 (UNSPEC_SHA512H): New unspec.
15059 (UNSPEC_SHA512H2): Ditto.
15060 (UNSPEC_SHA512SU0): Ditto.
15061 (UNSPEC_SHA512SU1): Ditto.
15062 * config/aarch64/aarch64-simd-builtins.def
15063 (aarch64_crypto_sha512hqv2di): New builtin.
15064 (aarch64_crypto_sha512h2qv2di): Ditto.
15065 (aarch64_crypto_sha512su0qv2di): Ditto.
15066 (aarch64_crypto_sha512su1qv2di): Ditto.
15067 (aarch64_eor3qv8hi): Ditto.
15068 (aarch64_rax1qv2di): Ditto.
15069 (aarch64_xarqv2di): Ditto.
15070 (aarch64_bcaxqv8hi): Ditto.
15071 * config/aarch64/aarch64-simd.md:
15072 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
15073 (aarch64_crypto_sha512su0qv2di): Ditto.
15074 (aarch64_crypto_sha512su1qv2di): Ditto.
15075 (aarch64_eor3qv8hi): Ditto.
15076 (aarch64_rax1qv2di): Ditto.
15077 (aarch64_xarqv2di): Ditto.
15078 (aarch64_bcaxqv8hi): Ditto.
15079 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
15080 (vsha512h2q_u64): Ditto.
15081 (vsha512su0q_u64): Ditto.
15082 (vsha512su1q_u64): Ditto.
15083 (veor3q_u16): Ditto.
15084 (vrax1q_u64): Ditto.
15085 (vxarq_u64): Ditto.
15086 (vbcaxq_u16): Ditto.
15087 * config/arm/types.md (crypto_sha512): New type attribute.
15088 (crypto_sha3): Ditto.
15089 (doc/invoke.texi): Document new sha3 option.
15091 2018-01-10 Michael Collison <michael.collison@arm.com>
15093 * config/aarch64/aarch64-builtins.c:
15094 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
15095 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15096 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
15097 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
15098 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
15099 (AARCH64_ISA_SM4): New ISA flag.
15100 (TARGET_SM4): New feature flag for sm4.
15101 * config/aarch64/aarch64-simd-builtins.def
15102 (aarch64_sm3ss1qv4si): Ditto.
15103 (aarch64_sm3tt1aq4si): Ditto.
15104 (aarch64_sm3tt1bq4si): Ditto.
15105 (aarch64_sm3tt2aq4si): Ditto.
15106 (aarch64_sm3tt2bq4si): Ditto.
15107 (aarch64_sm3partw1qv4si): Ditto.
15108 (aarch64_sm3partw2qv4si): Ditto.
15109 (aarch64_sm4eqv4si): Ditto.
15110 (aarch64_sm4ekeyqv4si): Ditto.
15111 * config/aarch64/aarch64-simd.md:
15112 (aarch64_sm3ss1qv4si): Ditto.
15113 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
15114 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
15115 (aarch64_sm4eqv4si): Ditto.
15116 (aarch64_sm4ekeyqv4si): Ditto.
15117 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
15118 (sm3part_op): Ditto.
15119 (CRYPTO_SM3TT): Ditto.
15120 (CRYPTO_SM3PART): Ditto.
15121 (UNSPEC_SM3SS1): New unspec.
15122 (UNSPEC_SM3TT1A): Ditto.
15123 (UNSPEC_SM3TT1B): Ditto.
15124 (UNSPEC_SM3TT2A): Ditto.
15125 (UNSPEC_SM3TT2B): Ditto.
15126 (UNSPEC_SM3PARTW1): Ditto.
15127 (UNSPEC_SM3PARTW2): Ditto.
15128 (UNSPEC_SM4E): Ditto.
15129 (UNSPEC_SM4EKEY): Ditto.
15130 * config/aarch64/constraints.md (Ui2): New constraint.
15131 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
15132 * config/arm/types.md (crypto_sm3): New type attribute.
15133 (crypto_sm4): Ditto.
15134 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
15135 (vsm3tt1aq_u32): Ditto.
15136 (vsm3tt1bq_u32): Ditto.
15137 (vsm3tt2aq_u32): Ditto.
15138 (vsm3tt2bq_u32): Ditto.
15139 (vsm3partw1q_u32): Ditto.
15140 (vsm3partw2q_u32): Ditto.
15141 (vsm4eq_u32): Ditto.
15142 (vsm4ekeyq_u32): Ditto.
15143 (doc/invoke.texi): Document new sm4 option.
15145 2018-01-10 Michael Collison <michael.collison@arm.com>
15147 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
15148 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
15149 (AARCH64_FL_FOR_ARCH8_4): New.
15150 (AARCH64_FL_V8_4): New flag.
15151 (doc/invoke.texi): Document new armv8.4-a option.
15153 2018-01-10 Michael Collison <michael.collison@arm.com>
15155 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15156 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
15157 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
15158 * config/aarch64/aarch64-option-extension.def: Add
15159 AARCH64_OPT_EXTENSION of 'sha2'.
15160 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
15161 (crypto): Disable sha2 and aes if crypto disabled.
15162 (crypto): Enable aes and sha2 if enabled.
15163 (simd): Disable sha2 and aes if simd disabled.
15164 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
15166 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
15167 (TARGET_SHA2): New feature flag for sha2.
15168 (TARGET_AES): New feature flag for aes.
15169 * config/aarch64/aarch64-simd.md:
15170 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
15171 conditional on TARGET_AES.
15172 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
15173 (aarch64_crypto_sha1hsi): Make pattern conditional
15175 (aarch64_crypto_sha1hv4si): Ditto.
15176 (aarch64_be_crypto_sha1hv4si): Ditto.
15177 (aarch64_crypto_sha1su1v4si): Ditto.
15178 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
15179 (aarch64_crypto_sha1su0v4si): Ditto.
15180 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
15181 (aarch64_crypto_sha256su0v4si): Ditto.
15182 (aarch64_crypto_sha256su1v4si): Ditto.
15183 (doc/invoke.texi): Document new aes and sha2 options.
15185 2018-01-10 Martin Sebor <msebor@redhat.com>
15187 PR tree-optimization/83781
15188 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
15191 2018-01-11 Martin Sebor <msebor@gmail.com>
15192 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15194 PR tree-optimization/83501
15195 PR tree-optimization/81703
15197 * tree-ssa-strlen.c (get_string_cst): Rename...
15198 (get_string_len): ...to this. Handle global constants.
15199 (handle_char_store): Adjust.
15201 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
15202 Jim Wilson <jimw@sifive.com>
15204 * config/riscv/riscv-protos.h (riscv_output_return): New.
15205 * config/riscv/riscv.c (struct machine_function): New naked_p field.
15206 (riscv_attribute_table, riscv_output_return),
15207 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
15208 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
15209 (riscv_compute_frame_info): Only compute frame->mask if not a naked
15211 (riscv_expand_prologue): Add early return for naked function.
15212 (riscv_expand_epilogue): Likewise.
15213 (riscv_function_ok_for_sibcall): Return false for naked function.
15214 (riscv_set_current_function): New.
15215 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
15216 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
15217 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
15218 * doc/extend.texi (RISC-V Function Attributes): New.
15220 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
15222 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
15223 check for 128-bit long double before checking TCmode.
15224 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
15225 128-bit long doubles before checking TFmode or TCmode.
15226 (FLOAT128_IBM_P): Likewise.
15228 2018-01-10 Martin Sebor <msebor@redhat.com>
15230 PR tree-optimization/83671
15231 * builtins.c (c_strlen): Unconditionally return zero for the empty
15233 Use -Warray-bounds for warnings.
15234 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
15235 for non-constant array indices with COMPONENT_REF, arrays of
15236 arrays, and pointers to arrays.
15237 (gimple_fold_builtin_strlen): Determine and set length range for
15238 non-constant character arrays.
15240 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
15242 PR middle-end/81897
15243 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
15246 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
15248 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
15250 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15253 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
15254 VECTOR_MEM_ALTIVEC_OR_VSX_P.
15255 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
15256 indexed_or_indirect_operand predicate.
15257 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
15258 (*vsx_le_perm_load_v8hi): Likewise.
15259 (*vsx_le_perm_load_v16qi): Likewise.
15260 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
15261 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
15262 (*vsx_le_perm_store_v8hi): Likewise.
15263 (*vsx_le_perm_store_v16qi): Likewise.
15264 (eight unnamed splitters): Likewise.
15266 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15268 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
15269 * config/rs6000/emmintrin.h: Likewise.
15270 * config/rs6000/mmintrin.h: Likewise.
15271 * config/rs6000/xmmintrin.h: Likewise.
15273 2018-01-10 David Malcolm <dmalcolm@redhat.com>
15276 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
15278 * tree.c (tree_nop_conversion): Return true for location wrapper
15280 (maybe_wrap_with_location): New function.
15281 (selftest::check_strip_nops): New function.
15282 (selftest::test_location_wrappers): New function.
15283 (selftest::tree_c_tests): Call it.
15284 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
15285 (maybe_wrap_with_location): New decl.
15286 (EXPR_LOCATION_WRAPPER_P): New macro.
15287 (location_wrapper_p): New inline function.
15288 (tree_strip_any_location_wrapper): New inline function.
15290 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
15293 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
15294 stack_realign_offset for the largest alignment of stack slot
15296 (ix86_find_max_used_stack_alignment): New function.
15297 (ix86_finalize_stack_frame_flags): Use it. Set
15298 max_used_stack_alignment if we don't realign stack.
15299 * config/i386/i386.h (machine_function): Add
15300 max_used_stack_alignment.
15302 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
15304 * config/arm/arm.opt (-mbranch-cost): New option.
15305 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
15308 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
15311 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
15312 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
15314 2018-01-10 Richard Biener <rguenther@suse.de>
15317 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
15318 early out so it also covers the case where we have a non-NULL
15321 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15323 PR tree-optimization/83753
15324 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
15325 for non-strided grouped accesses if the number of elements is 1.
15327 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15330 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
15331 * i386.h (TARGET_USE_GATHER): Define.
15332 * x86-tune.def (X86_TUNE_USE_GATHER): New.
15334 2018-01-10 Martin Liska <mliska@suse.cz>
15337 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
15338 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
15340 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
15341 CLEANUP_NO_PARTITIONING is not set.
15343 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15345 * doc/rtl.texi: Remove documentation of (const ...) wrappers
15346 for vectors, as a partial revert of r254296.
15347 * rtl.h (const_vec_p): Delete.
15348 (const_vec_duplicate_p): Don't test for vector CONSTs.
15349 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
15350 * expmed.c (make_tree): Likewise.
15353 * common.md (E, F): Use CONSTANT_P instead of checking for
15355 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
15356 checking for CONST_VECTOR.
15358 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15360 PR middle-end/83575
15361 * predict.c (force_edge_cold): Handle in more sane way edges
15362 with no prediction.
15364 2018-01-09 Carl Love <cel@us.ibm.com>
15366 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
15368 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
15369 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
15370 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
15371 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
15372 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
15373 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
15374 * config/rs6000/rs6000-protos.h: Add extern defition for
15375 rs6000_generate_float2_double_code.
15376 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
15378 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
15379 (float2_v2df): Add define_expand.
15381 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
15384 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
15385 op_mode in the force_to_mode call.
15387 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15389 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
15390 instead of checking each element individually.
15391 (aarch64_evpc_uzp): Likewise.
15392 (aarch64_evpc_zip): Likewise.
15393 (aarch64_evpc_ext): Likewise.
15394 (aarch64_evpc_rev): Likewise.
15395 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
15396 instead of checking each element individually. Return true without
15398 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
15399 whether all selected elements come from the same input, instead of
15400 checking each element individually. Remove calls to gen_rtx_REG,
15401 start_sequence and end_sequence and instead assert that no rtl is
15404 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15406 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
15407 order of HIGH and CONST checks.
15409 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15411 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
15412 if the destination isn't an SSA_NAME.
15414 2018-01-09 Richard Biener <rguenther@suse.de>
15416 PR tree-optimization/83668
15417 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
15419 (canonicalize_loop_form): ... here, renamed from ...
15420 (canonicalize_loop_closed_ssa_form): ... this and amended to
15421 swap successor edges for loop exit blocks to make us use
15422 the RPO order we need for initial schedule generation.
15424 2018-01-09 Joseph Myers <joseph@codesourcery.com>
15426 PR tree-optimization/64811
15427 * match.pd: When optimizing comparisons with Inf, avoid
15428 introducing or losing exceptions from comparisons with NaN.
15430 2018-01-09 Martin Liska <mliska@suse.cz>
15433 * asan.c (shadow_mem_size): Add gcc_assert.
15435 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
15437 Don't save registers in main().
15440 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
15441 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
15442 * config/avr/avr.c (avr_set_current_function): Don't error if
15443 naked, OS_task or OS_main are specified at the same time.
15444 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
15446 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
15448 * common/config/avr/avr-common.c (avr_option_optimization_table):
15449 Switch on -mmain-is-OS_task for optimizing compilations.
15451 2018-01-09 Richard Biener <rguenther@suse.de>
15453 PR tree-optimization/83572
15454 * graphite.c: Include cfganal.h.
15455 (graphite_transform_loops): Connect infinite loops to exit
15456 and remove fake edges at the end.
15458 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15460 * ipa-inline.c (edge_badness): Revert accidental checkin.
15462 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15465 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
15466 symbols; not inline clones.
15468 2018-01-09 Jakub Jelinek <jakub@redhat.com>
15471 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
15472 hard registers. Formatting fixes.
15474 PR preprocessor/83722
15475 * gcc.c (try_generate_repro): Pass
15476 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
15477 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
15480 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
15481 Kito Cheng <kito.cheng@gmail.com>
15483 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
15484 (riscv_leaf_function_p): Delete.
15485 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
15487 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15489 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
15491 (do_ifelse): New function.
15492 (do_isel): New function.
15493 (do_sub3): New function.
15494 (do_add3): New function.
15495 (do_load_mask_compare): New function.
15496 (do_overlap_load_compare): New function.
15497 (expand_compare_loop): New function.
15498 (expand_block_compare): Call expand_compare_loop() when appropriate.
15499 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
15500 option description.
15501 (-mblock-compare-inline-loop-limit): New option.
15503 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15506 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
15507 Reverse order of second and third operands in first alternative.
15508 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
15509 of first and second elements in UNSPEC_VPERMR vector.
15510 (altivec_expand_vec_perm_le): Likewise.
15512 2018-01-08 Jeff Law <law@redhat.com>
15514 PR rtl-optimizatin/81308
15515 * tree-switch-conversion.c (cfg_altered): New file scoped static.
15516 (process_switch): If group_case_labels makes a change, then set
15518 (pass_convert_switch::execute): If a switch is converted, then
15519 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
15521 PR rtl-optimization/81308
15522 * recog.c (split_all_insns): Conditionally cleanup the CFG after
15525 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
15527 PR target/83663 - Revert r255946
15528 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
15529 generation for cases where splatting a value is not useful.
15530 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
15531 across a vec_duplicate and a paradoxical subreg forming a vector
15532 mode to a vec_concat.
15534 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15536 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
15537 -march=armv8.3-a variants.
15538 * config/arm/t-multilib: Likewise.
15539 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
15541 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15543 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
15545 (cceq_ior_compare_complement): Give it a name so I can use it, and
15546 change boolean_or_operator predicate to boolean_operator so it can
15547 be used to generate a crand.
15548 (eqne): New code iterator.
15549 (bd/bd_neg): New code_attrs.
15550 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
15551 a single define_insn.
15552 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
15553 decrement (bdnzt/bdnzf/bdzt/bdzf).
15554 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
15555 with the new names of the branch decrement patterns, and added the
15556 names of the branch decrement conditional patterns.
15558 2018-01-08 Richard Biener <rguenther@suse.de>
15560 PR tree-optimization/83563
15561 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
15564 2018-01-08 Richard Biener <rguenther@suse.de>
15566 PR middle-end/83713
15567 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
15569 2018-01-08 Richard Biener <rguenther@suse.de>
15571 PR tree-optimization/83685
15572 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
15573 references to abnormals.
15575 2018-01-08 Richard Biener <rguenther@suse.de>
15578 * dwarf2out.c (output_indirect_strings): Handle empty
15579 skeleton_debug_str_hash.
15580 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
15582 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15584 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
15585 (emit_store_direct): Likewise.
15586 (arc_trampoline_adjust_address): Likewise.
15587 (arc_asm_trampoline_template): New function.
15588 (arc_initialize_trampoline): Use asm_trampoline_template.
15589 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
15590 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
15591 * config/arc/arc.md (flush_icache): Delete pattern.
15593 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15595 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
15596 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
15599 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15602 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
15603 by not USED_FOR_TARGET.
15604 (make_pass_resolve_sw_modes): Likewise.
15606 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15608 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
15611 2018-01-08 Richard Biener <rguenther@suse.de>
15613 PR middle-end/83580
15614 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
15616 2018-01-08 Richard Biener <rguenther@suse.de>
15618 PR middle-end/83517
15619 * match.pd ((t * 2) / 2) -> t): Add missing :c.
15621 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
15623 PR middle-end/81897
15624 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
15625 basic blocks with a small number of successors.
15626 (convert_control_dep_chain_into_preds): Improve handling of
15628 (dump_predicates): Split apart into...
15629 (dump_pred_chain): ...here...
15630 (dump_pred_info): ...and here.
15631 (can_one_predicate_be_invalidated_p): Add debugging printfs.
15632 (can_chain_union_be_invalidated_p): Improve check for invalidation
15634 (uninit_uses_cannot_happen): Avoid unnecessary if
15635 convert_control_dep_chain_into_preds yielded nothing.
15637 2018-01-06 Martin Sebor <msebor@redhat.com>
15639 PR tree-optimization/83640
15640 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
15641 subtracting negative offset from size.
15642 (builtin_access::overlap): Adjust offset bounds of the access to fall
15643 within the size of the object if possible.
15645 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
15647 PR rtl-optimization/83699
15648 * expmed.c (extract_bit_field_1): Restrict the vector usage of
15649 extract_bit_field_as_subreg to cases in which the extracted
15650 value is also a vector.
15652 * lra-constraints.c (process_alt_operands): Test for the equivalence
15653 substitutions when detecting a possible reload cycle.
15655 2018-01-06 Jakub Jelinek <jakub@redhat.com>
15658 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
15659 by default if flag_selective_schedling{,2}. Formatting fixes.
15661 PR rtl-optimization/83682
15662 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
15663 if it has non-VECTOR_MODE element mode.
15664 (vec_duplicate_p): Likewise.
15666 PR middle-end/83694
15667 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
15668 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
15670 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15673 * config/i386/i386-builtin.def
15674 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
15675 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
15676 Require also OPTION_MASK_ISA_AVX512F in addition to
15677 OPTION_MASK_ISA_GFNI.
15678 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
15679 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
15680 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
15681 to OPTION_MASK_ISA_GFNI.
15682 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
15683 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
15684 OPTION_MASK_ISA_AVX512BW.
15685 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
15686 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
15687 addition to OPTION_MASK_ISA_GFNI.
15688 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
15689 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
15690 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
15691 to OPTION_MASK_ISA_GFNI.
15692 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
15693 a requirement for all ISAs rather than any of them with a few
15695 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
15697 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
15698 bitmasks to be enabled with 3 exceptions, instead of requiring any
15699 enabled ISA with lots of exceptions.
15700 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
15701 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
15702 Change avx512bw in isa attribute to avx512f.
15703 * config/i386/sgxintrin.h: Add license boilerplate.
15704 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
15705 to __AVX512F__ and __AVX512VL to __AVX512VL__.
15706 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
15707 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
15709 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
15710 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
15711 temporarily sse2 rather than sse if not enabled already.
15714 * config/i386/sse.md (VI248_VLBW): Rename to ...
15715 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
15716 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
15717 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
15718 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
15719 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
15720 mode iterator instead of VI248_VLBW.
15722 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
15724 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
15725 (record_modified): Skip clobbers; add debug output.
15726 (param_change_prob): Use sreal frequencies.
15728 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15730 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
15731 punt for user-aligned variables.
15733 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15735 * tree-chrec.c (chrec_contains_symbols): Return true for
15738 2018-01-05 Sudakshina Das <sudi.das@arm.com>
15741 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
15742 of (x|y) == x for BICS pattern.
15744 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15746 PR tree-optimization/83605
15747 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
15748 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
15751 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
15753 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
15754 * config/epiphany/rtems.h: New file.
15756 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15757 Uros Bizjak <ubizjak@gmail.com>
15760 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
15761 QIreg_operand instead of register_operand predicate.
15762 * config/i386/i386.c (ix86_rop_should_change_byte_p,
15763 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
15764 comments instead of -fmitigate[-_]rop.
15766 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15769 * cgraphunit.c (symbol_table::compile): Switch to text_section
15770 before calling assembly_start debug hook.
15771 * run-rtl-passes.c (run_rtl_passes): Likewise.
15774 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15776 * tree-vrp.c (extract_range_from_binary_expr_1): Check
15777 range_int_cst_p rather than !symbolic_range_p before calling
15778 extract_range_from_multiplicative_op_1.
15780 2018-01-04 Jeff Law <law@redhat.com>
15782 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
15783 redundant test in assertion.
15785 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15787 * doc/rtl.texi: Document machine_mode wrapper classes.
15789 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15791 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
15792 using tree_to_uhwi.
15794 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15796 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
15797 the VEC_PERM_EXPR fold to fail.
15799 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15802 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
15803 to switched_sections.
15805 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15808 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
15809 test for d.testing.
15811 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
15814 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
15815 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
15817 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15820 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
15821 is BLKmode and bitpos not zero or mode change is needed.
15823 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15826 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
15829 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
15832 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
15833 instead of MULT rtx. Update all corresponding splitters.
15834 (*saddl_se): Ditto.
15835 (*ssub<modesuffix>): Ditto.
15836 (*ssubl_se): Ditto.
15837 (*cmp_sadd_di): Update split patterns.
15838 (*cmp_sadd_si): Ditto.
15839 (*cmp_sadd_sidi): Ditto.
15840 (*cmp_ssub_di): Ditto.
15841 (*cmp_ssub_si): Ditto.
15842 (*cmp_ssub_sidi): Ditto.
15843 * config/alpha/predicates.md (const23_operand): New predicate.
15844 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
15845 Look for ASHIFT, not MULT inner operand.
15846 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
15848 2018-01-04 Martin Liska <mliska@suse.cz>
15850 PR gcov-profile/83669
15851 * gcov.c (output_intermediate_file): Add version to intermediate
15853 * doc/gcov.texi: Document new field 'version' in intermediate
15854 file format. Fix location of '-k' option of gcov command.
15856 2018-01-04 Martin Liska <mliska@suse.cz>
15859 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
15861 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15863 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
15865 2018-01-03 Martin Sebor <msebor@redhat.com>
15867 PR tree-optimization/83655
15868 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
15869 checking calls with invalid arguments.
15871 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15873 * tree-vect-stmts.c (vect_get_store_rhs): New function.
15874 (vectorizable_mask_load_store): Delete.
15875 (vectorizable_call): Return false for masked loads and stores.
15876 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
15877 instead of gimple_assign_rhs1.
15878 (vectorizable_load): Handle IFN_MASK_LOAD.
15879 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
15881 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15883 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
15885 (vectorizable_mask_load_store): ...here.
15886 (vectorizable_load): ...and here.
15888 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15890 * tree-vect-stmts.c (vect_build_all_ones_mask)
15891 (vect_build_zero_merge_argument): New functions, split out from...
15892 (vectorizable_load): ...here.
15894 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15896 * tree-vect-stmts.c (vect_check_store_rhs): New function,
15898 (vectorizable_mask_load_store): ...here.
15899 (vectorizable_store): ...and here.
15901 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15903 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
15905 (vectorizable_mask_load_store): ...here.
15907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15909 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
15910 (vect_model_store_cost): Take a vec_load_store_type instead of a
15912 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
15913 (vect_model_store_cost): Take a vec_load_store_type instead of a
15915 (vectorizable_mask_load_store): Update accordingly.
15916 (vectorizable_store): Likewise.
15917 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
15919 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15921 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
15922 IFN_MASK_LOAD calls here rather than...
15923 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
15925 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15926 Alan Hayward <alan.hayward@arm.com>
15927 David Sherwood <david.sherwood@arm.com>
15929 * expmed.c (extract_bit_field_1): For vector extracts,
15930 fall back to extract_bit_field_as_subreg if vec_extract
15933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15934 Alan Hayward <alan.hayward@arm.com>
15935 David Sherwood <david.sherwood@arm.com>
15937 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
15938 they are variable or constant sized.
15939 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
15940 slots for constant-sized data.
15942 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15943 Alan Hayward <alan.hayward@arm.com>
15944 David Sherwood <david.sherwood@arm.com>
15946 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
15947 handling COND_EXPRs with boolean comparisons, try to find a better
15948 basis for the mask type than the boolean itself.
15950 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15952 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
15953 is calculated and how it can be overridden.
15954 * genmodes.c (max_bitsize_mode_any_mode): New variable.
15955 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
15957 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
15960 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15961 Alan Hayward <alan.hayward@arm.com>
15962 David Sherwood <david.sherwood@arm.com>
15964 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
15965 Remove the mode argument.
15966 (aarch64_simd_valid_immediate): Remove the mode and inverse
15968 * config/aarch64/iterators.md (bitsize): New iterator.
15969 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
15970 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
15971 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
15972 aarch64_simd_valid_immediate.
15973 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
15974 (aarch64_reg_or_bic_imm): Likewise.
15975 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
15976 with an insn_type enum and msl with a modifier_type enum.
15977 Replace element_width with a scalar_mode. Change the shift
15978 to unsigned int. Add constructors for scalar_float_mode and
15979 scalar_int_mode elements.
15980 (aarch64_vect_float_const_representable_p): Delete.
15981 (aarch64_can_const_movi_rtx_p)
15982 (aarch64_simd_scalar_immediate_valid_for_move)
15983 (aarch64_simd_make_constant): Update call to
15984 aarch64_simd_valid_immediate.
15985 (aarch64_advsimd_valid_immediate_hs): New function.
15986 (aarch64_advsimd_valid_immediate): Likewise.
15987 (aarch64_simd_valid_immediate): Remove mode and inverse
15988 arguments. Rewrite to use the above. Use const_vec_duplicate_p
15989 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
15990 and aarch64_float_const_representable_p on the result.
15991 (aarch64_output_simd_mov_immediate): Remove mode argument.
15992 Update call to aarch64_simd_valid_immediate and use of
15993 simd_immediate_info.
15994 (aarch64_output_scalar_simd_mov_immediate): Update call
15997 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15998 Alan Hayward <alan.hayward@arm.com>
15999 David Sherwood <david.sherwood@arm.com>
16001 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
16002 (mode_nunits): Likewise CONST_MODE_NUNITS.
16003 * machmode.def (ADJUST_NUNITS): Document.
16004 * genmodes.c (mode_data::need_nunits_adj): New field.
16005 (blank_mode): Update accordingly.
16006 (adj_nunits): New variable.
16007 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
16009 (emit_mode_size_inline): Set need_bytesize_adj for all modes
16010 listed in adj_nunits.
16011 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
16012 listed in adj_nunits. Don't emit case statements for such modes.
16013 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
16014 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
16015 nothing if adj_nunits is nonnull.
16016 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
16017 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
16018 (emit_mode_fbit): Update use of print_maybe_const_decl.
16019 (emit_move_size): Likewise. Treat the array as non-const
16021 (emit_mode_adjustments): Handle adj_nunits.
16023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16025 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
16026 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
16027 (VECTOR_MODES): Use it.
16028 (make_vector_modes): Take the prefix as an argument.
16030 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16031 Alan Hayward <alan.hayward@arm.com>
16032 David Sherwood <david.sherwood@arm.com>
16034 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
16035 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
16036 for MODE_VECTOR_BOOL.
16037 * machmode.def (VECTOR_BOOL_MODE): Document.
16038 * genmodes.c (VECTOR_BOOL_MODE): New macro.
16039 (make_vector_bool_mode): New function.
16040 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
16042 * lto-streamer-in.c (lto_input_mode_table): Likewise.
16043 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
16045 * stor-layout.c (int_mode_for_mode): Likewise.
16046 * tree.c (build_vector_type_for_mode): Likewise.
16047 * varasm.c (output_constant_pool_2): Likewise.
16048 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
16049 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
16050 for MODE_VECTOR_BOOL.
16051 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
16052 of mode class checks.
16053 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
16054 instead of a list of mode class checks.
16055 (expand_vector_scalar_condition): Likewise.
16056 (type_for_widest_vector_mode): Handle BImode as an inner mode.
16058 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16059 Alan Hayward <alan.hayward@arm.com>
16060 David Sherwood <david.sherwood@arm.com>
16062 * machmode.h (mode_size): Change from unsigned short to
16064 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
16065 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16066 or if measurement_type is not polynomial.
16067 (fixed_size_mode::includes_p): Check for constant-sized modes.
16068 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
16069 return a poly_uint16 rather than an unsigned short.
16070 (emit_mode_size): Change the type of mode_size from unsigned short
16071 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
16072 (emit_mode_adjustments): Cope with polynomial vector sizes.
16073 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16075 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16077 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
16078 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
16079 * caller-save.c (setup_save_areas): Likewise.
16080 (replace_reg_with_saved_mem): Likewise.
16081 * calls.c (emit_library_call_value_1): Likewise.
16082 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
16083 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
16084 (gen_lowpart_for_combine): Likewise.
16085 * convert.c (convert_to_integer_1): Likewise.
16086 * cse.c (equiv_constant, cse_insn): Likewise.
16087 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
16088 (cselib_subst_to_values): Likewise.
16089 * dce.c (word_dce_process_block): Likewise.
16090 * df-problems.c (df_word_lr_mark_ref): Likewise.
16091 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
16092 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
16093 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
16094 (rtl_for_decl_location): Likewise.
16095 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
16096 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
16097 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
16098 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
16099 (expand_expr_real_1): Likewise.
16100 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
16101 (pad_below): Likewise.
16102 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16103 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
16104 * ira.c (get_subreg_tracking_sizes): Likewise.
16105 * ira-build.c (ira_create_allocno_objects): Likewise.
16106 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
16107 (ira_sort_regnos_for_alter_reg): Likewise.
16108 * ira-costs.c (record_operand_costs): Likewise.
16109 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
16110 (resolve_simple_move): Likewise.
16111 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
16112 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
16113 (lra_constraints): Likewise.
16114 (CONST_POOL_OK_P): Reject variable-sized modes.
16115 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
16116 (add_pseudo_to_slot, lra_spill): Likewise.
16117 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16118 * optabs-query.c (get_best_extraction_insn): Likewise.
16119 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16120 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
16121 (expand_mult_highpart, valid_multiword_target_p): Likewise.
16122 * recog.c (offsettable_address_addr_space_p): Likewise.
16123 * regcprop.c (maybe_mode_change): Likewise.
16124 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
16125 * regrename.c (build_def_use): Likewise.
16126 * regstat.c (dump_reg_info): Likewise.
16127 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
16128 (find_reloads, find_reloads_subreg_address): Likewise.
16129 * reload1.c (eliminate_regs_1): Likewise.
16130 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
16131 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
16132 (simplify_binary_operation_1, simplify_subreg): Likewise.
16133 * targhooks.c (default_function_arg_padding): Likewise.
16134 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
16135 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
16136 (verify_gimple_assign_ternary): Likewise.
16137 * tree-inline.c (estimate_move_cost): Likewise.
16138 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16139 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
16140 (get_address_cost_ainc): Likewise.
16141 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
16142 (vect_supportable_dr_alignment): Likewise.
16143 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16144 (vectorizable_reduction): Likewise.
16145 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
16146 (vectorizable_operation, vectorizable_load): Likewise.
16147 * tree.c (build_same_sized_truth_vector_type): Likewise.
16148 * valtrack.c (cleanup_auto_inc_dec): Likewise.
16149 * var-tracking.c (emit_note_insn_var_location): Likewise.
16150 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
16151 (ADDR_VEC_ALIGN): Likewise.
16153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16154 Alan Hayward <alan.hayward@arm.com>
16155 David Sherwood <david.sherwood@arm.com>
16157 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
16159 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16160 or if measurement_type is polynomial.
16161 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
16162 * combine.c (make_extraction): Likewise.
16163 * dse.c (find_shift_sequence): Likewise.
16164 * dwarf2out.c (mem_loc_descriptor): Likewise.
16165 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
16166 (extract_bit_field, extract_low_bits): Likewise.
16167 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
16168 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
16169 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
16170 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
16171 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16172 * reload.c (find_reloads): Likewise.
16173 * reload1.c (alter_reg): Likewise.
16174 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
16175 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
16176 * tree-if-conv.c (predicate_mem_writes): Likewise.
16177 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
16178 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
16179 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
16180 * valtrack.c (dead_debug_insert_temp): Likewise.
16181 * varasm.c (mergeable_constant_section): Likewise.
16182 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
16184 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16185 Alan Hayward <alan.hayward@arm.com>
16186 David Sherwood <david.sherwood@arm.com>
16188 * expr.c (expand_assignment): Cope with polynomial mode sizes
16189 when assigning to a CONCAT.
16191 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16192 Alan Hayward <alan.hayward@arm.com>
16193 David Sherwood <david.sherwood@arm.com>
16195 * machmode.h (mode_precision): Change from unsigned short to
16197 (mode_to_precision): Return a poly_uint16 rather than an unsigned
16199 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
16200 or if measurement_type is not polynomial.
16201 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
16202 in which the mode is already known to be a scalar_int_mode.
16203 * genmodes.c (emit_mode_precision): Change the type of mode_precision
16204 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
16206 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16207 for GET_MODE_PRECISION.
16208 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16209 for GET_MODE_PRECISION.
16210 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
16212 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
16213 (expand_field_assignment, make_extraction): Likewise.
16214 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
16215 (get_last_value): Likewise.
16216 * convert.c (convert_to_integer_1): Likewise.
16217 * cse.c (cse_insn): Likewise.
16218 * expr.c (expand_expr_real_1): Likewise.
16219 * lra-constraints.c (simplify_operand_subreg): Likewise.
16220 * optabs-query.c (can_atomic_load_p): Likewise.
16221 * optabs.c (expand_atomic_load): Likewise.
16222 (expand_atomic_store): Likewise.
16223 * ree.c (combine_reaching_defs): Likewise.
16224 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
16225 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
16226 * tree.h (type_has_mode_precision_p): Likewise.
16227 * ubsan.c (instrument_si_overflow): Likewise.
16229 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16230 Alan Hayward <alan.hayward@arm.com>
16231 David Sherwood <david.sherwood@arm.com>
16233 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
16234 polynomial numbers of units.
16235 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
16236 (valid_vector_subparts_p): New function.
16237 (build_vector_type): Remove temporary shim and take the number
16238 of units as a poly_uint64 rather than an int.
16239 (build_opaque_vector_type): Take the number of units as a
16240 poly_uint64 rather than an int.
16241 * tree.c (build_vector_from_ctor): Handle polynomial
16242 TYPE_VECTOR_SUBPARTS.
16243 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
16244 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
16245 (build_vector_from_val): If the number of units is variable,
16246 use build_vec_duplicate_cst for constant operands and
16247 VEC_DUPLICATE_EXPR otherwise.
16248 (make_vector_type): Remove temporary is_constant ().
16249 (build_vector_type, build_opaque_vector_type): Take the number of
16250 units as a poly_uint64 rather than an int.
16251 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
16253 * cfgexpand.c (expand_debug_expr): Likewise.
16254 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
16255 (store_constructor, expand_expr_real_1): Likewise.
16256 (const_scalar_mask_from_tree): Likewise.
16257 * fold-const-call.c (fold_const_reduction): Likewise.
16258 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
16259 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
16260 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
16261 (fold_relational_const): Likewise.
16262 (native_interpret_vector): Likewise. Change the size from an
16263 int to an unsigned int.
16264 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
16265 TYPE_VECTOR_SUBPARTS.
16266 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
16267 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
16268 duplicating a non-constant operand into a variable-length vector.
16269 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
16270 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
16271 * ipa-icf.c (sem_variable::equals): Likewise.
16272 * match.pd: Likewise.
16273 * omp-simd-clone.c (simd_clone_subparts): Likewise.
16274 * print-tree.c (print_node): Likewise.
16275 * stor-layout.c (layout_type): Likewise.
16276 * targhooks.c (default_builtin_vectorization_cost): Likewise.
16277 * tree-cfg.c (verify_gimple_comparison): Likewise.
16278 (verify_gimple_assign_binary): Likewise.
16279 (verify_gimple_assign_ternary): Likewise.
16280 (verify_gimple_assign_single): Likewise.
16281 * tree-pretty-print.c (dump_generic_node): Likewise.
16282 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16283 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
16284 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
16285 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
16286 (vect_shift_permute_load_chain): Likewise.
16287 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
16288 (expand_vector_condition, optimize_vector_constructor): Likewise.
16289 (lower_vec_perm, get_compute_type): Likewise.
16290 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16291 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
16292 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
16293 (vect_recog_mask_conversion_pattern): Likewise.
16294 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
16295 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
16296 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16297 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
16298 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
16299 (vectorizable_shift, vectorizable_operation, vectorizable_store)
16300 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
16301 (supportable_widening_operation): Likewise.
16302 (supportable_narrowing_operation): Likewise.
16303 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
16305 * varasm.c (output_constant): Likewise.
16307 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16308 Alan Hayward <alan.hayward@arm.com>
16309 David Sherwood <david.sherwood@arm.com>
16311 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
16312 so that both the length == 3 and length != 3 cases set up their
16313 own permute vectors. Add comments explaining why we know the
16314 number of elements is constant.
16315 (vect_permute_load_chain): Likewise.
16317 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16318 Alan Hayward <alan.hayward@arm.com>
16319 David Sherwood <david.sherwood@arm.com>
16321 * machmode.h (mode_nunits): Change from unsigned char to
16323 (ONLY_FIXED_SIZE_MODES): New macro.
16324 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
16325 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
16326 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
16328 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
16329 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
16330 or if measurement_type is not polynomial.
16331 * genmodes.c (ZERO_COEFFS): New macro.
16332 (emit_mode_nunits_inline): Make mode_nunits_inline return a
16334 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
16335 Use ZERO_COEFFS when emitting initializers.
16336 * data-streamer.h (bp_pack_poly_value): New function.
16337 (bp_unpack_poly_value): Likewise.
16338 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16339 for GET_MODE_NUNITS.
16340 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16341 for GET_MODE_NUNITS.
16342 * tree.c (make_vector_type): Remove temporary shim and make
16343 the real function take the number of units as a poly_uint64
16344 rather than an int.
16345 (build_vector_type_for_mode): Handle polynomial nunits.
16346 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
16347 * emit-rtl.c (const_vec_series_p_1): Likewise.
16348 (gen_rtx_CONST_VECTOR): Likewise.
16349 * fold-const.c (test_vec_duplicate_folding): Likewise.
16350 * genrecog.c (validate_pattern): Likewise.
16351 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
16352 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16353 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
16354 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
16355 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
16356 * rtlanal.c (subreg_get_info): Likewise.
16357 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16358 (vect_grouped_load_supported): Likewise.
16359 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
16360 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
16361 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
16362 (simplify_const_unary_operation, simplify_binary_operation_1)
16363 (simplify_const_binary_operation, simplify_ternary_operation)
16364 (test_vector_ops_duplicate, test_vector_ops): Likewise.
16365 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
16366 instead of CONST_VECTOR_NUNITS.
16367 * varasm.c (output_constant_pool_2): Likewise.
16368 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
16369 explicit-encoded elements in the XVEC for variable-length vectors.
16371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16373 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
16375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16376 Alan Hayward <alan.hayward@arm.com>
16377 David Sherwood <david.sherwood@arm.com>
16379 * coretypes.h (fixed_size_mode): Declare.
16380 (fixed_size_mode_pod): New typedef.
16381 * builtins.h (target_builtins::x_apply_args_mode)
16382 (target_builtins::x_apply_result_mode): Change type to
16383 fixed_size_mode_pod.
16384 * builtins.c (apply_args_size, apply_result_size, result_vector)
16385 (expand_builtin_apply_args_1, expand_builtin_apply)
16386 (expand_builtin_return): Update accordingly.
16388 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16390 * cse.c (hash_rtx_cb): Hash only the encoded elements.
16391 * cselib.c (cselib_hash_rtx): Likewise.
16392 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
16393 CONST_VECTOR encoding.
16395 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16396 Jeff Law <law@redhat.com>
16399 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
16400 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
16401 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
16402 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
16405 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
16406 explicitly probe *sp in a noreturn function if there were any callee
16407 register saves or frame pointer is needed.
16409 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16412 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
16413 BLKmode for ternary, binary or unary expressions.
16416 * var-tracking.c (delete_vta_debug_insn): New inline function.
16417 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
16418 insns from get_insns () to NULL instead of each bb separately.
16419 Use delete_vta_debug_insn. No longer static.
16420 (vt_debug_insns_local, variable_tracking_main_1): Adjust
16421 delete_vta_debug_insns callers.
16422 * rtl.h (delete_vta_debug_insns): Declare.
16423 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
16424 instead of variable_tracking_main.
16426 2018-01-03 Martin Sebor <msebor@redhat.com>
16428 PR tree-optimization/83603
16429 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
16430 arguments past the endof the argument list in functions declared
16431 without a prototype.
16432 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
16433 Avoid checking when arguments are null.
16435 2018-01-03 Martin Sebor <msebor@redhat.com>
16438 * doc/extend.texi (attribute const): Fix a typo.
16439 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
16440 issuing -Wsuggest-attribute for void functions.
16442 2018-01-03 Martin Sebor <msebor@redhat.com>
16444 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
16445 offset_int::from instead of wide_int::to_shwi.
16446 (maybe_diag_overlap): Remove assertion.
16447 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
16448 * gimple-ssa-sprintf.c (format_directive): Same.
16449 (parse_directive): Same.
16450 (sprintf_dom_walker::compute_format_length): Same.
16451 (try_substitute_return_value): Same.
16453 2018-01-03 Jeff Law <law@redhat.com>
16455 PR middle-end/83654
16456 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
16457 non-constant residual for zero at runtime and avoid probing in
16458 that case. Reorganize code for trailing problem to mirror handling
16461 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16463 PR tree-optimization/83501
16464 * tree-ssa-strlen.c (get_string_cst): New.
16465 (handle_char_store): Call get_string_cst.
16467 2018-01-03 Martin Liska <mliska@suse.cz>
16469 PR tree-optimization/83593
16470 * tree-ssa-strlen.c: Include tree-cfg.h.
16471 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
16472 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
16473 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
16475 (strlen_dom_walker::before_dom_children): Call
16476 gimple_purge_dead_eh_edges. Dump tranformation with details
16478 (strlen_dom_walker::before_dom_children): Update call by adding
16479 new argument cleanup_eh.
16480 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
16482 2018-01-03 Martin Liska <mliska@suse.cz>
16485 * cif-code.def (VARIADIC_THUNK): New enum value.
16486 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
16489 2018-01-03 Jan Beulich <jbeulich@suse.com>
16491 * sse.md (mov<mode>_internal): Tighten condition for when to use
16492 vmovdqu<ssescalarsize> for TI and OI modes.
16494 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16496 Update copyright years.
16498 2018-01-03 Martin Liska <mliska@suse.cz>
16501 * ipa-visibility.c (function_and_variable_visibility): Skip
16502 functions with noipa attribure.
16504 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16506 * gcc.c (process_command): Update copyright notice dates.
16507 * gcov-dump.c (print_version): Ditto.
16508 * gcov.c (print_version): Ditto.
16509 * gcov-tool.c (print_version): Ditto.
16510 * gengtype.c (create_file): Ditto.
16511 * doc/cpp.texi: Bump @copying's copyright year.
16512 * doc/cppinternals.texi: Ditto.
16513 * doc/gcc.texi: Ditto.
16514 * doc/gccint.texi: Ditto.
16515 * doc/gcov.texi: Ditto.
16516 * doc/install.texi: Ditto.
16517 * doc/invoke.texi: Ditto.
16519 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16521 * vector-builder.h (vector_builder::m_full_nelts): Change from
16522 unsigned int to poly_uint64.
16523 (vector_builder::full_nelts): Update prototype accordingly.
16524 (vector_builder::new_vector): Likewise.
16525 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
16526 (vector_builder::operator ==): Likewise.
16527 (vector_builder::finalize): Likewise.
16528 * int-vector-builder.h (int_vector_builder::int_vector_builder):
16529 Take the number of elements as a poly_uint64 rather than an
16531 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
16532 from unsigned int to poly_uint64.
16533 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
16534 (vec_perm_indices::new_vector): Likewise.
16535 (vec_perm_indices::length): Likewise.
16536 (vec_perm_indices::nelts_per_input): Likewise.
16537 (vec_perm_indices::input_nelts): Likewise.
16538 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
16539 number of elements per input as a poly_uint64 rather than an
16540 unsigned int. Use the original encoding for variable-length
16541 vectors, rather than clamping each individual element.
16542 For the second and subsequent elements in each pattern,
16543 clamp the step and base before clamping their sum.
16544 (vec_perm_indices::series_p): Handle polynomial element counts.
16545 (vec_perm_indices::all_in_range_p): Likewise.
16546 (vec_perm_indices_to_tree): Likewise.
16547 (vec_perm_indices_to_rtx): Likewise.
16548 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
16549 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
16550 (tree_vector_builder::new_binary_operation): Handle polynomial
16551 element counts. Return false if we need to know the number
16552 of elements at compile time.
16553 * fold-const.c (fold_vec_perm): Punt if the number of elements
16554 isn't known at compile time.
16556 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16558 * vec-perm-indices.h (vec_perm_builder): Change element type
16559 from HOST_WIDE_INT to poly_int64.
16560 (vec_perm_indices::element_type): Update accordingly.
16561 (vec_perm_indices::clamp): Handle polynomial element_types.
16562 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16563 (vec_perm_indices::all_in_range_p): Likewise.
16564 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
16566 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
16567 polynomial vec_perm_indices element types.
16568 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
16569 * fold-const.c (fold_vec_perm): Likewise.
16570 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
16571 * tree-vect-generic.c (lower_vec_perm): Likewise.
16572 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16573 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
16574 element type to HOST_WIDE_INT.
16576 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16577 Alan Hayward <alan.hayward@arm.com>
16578 David Sherwood <david.sherwood@arm.com>
16580 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
16581 rather than an int. Use plus_constant.
16582 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
16583 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
16585 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16586 Alan Hayward <alan.hayward@arm.com>
16587 David Sherwood <david.sherwood@arm.com>
16589 * calls.c (emit_call_1, expand_call): Change struct_value_size from
16590 a HOST_WIDE_INT to a poly_int64.
16592 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16593 Alan Hayward <alan.hayward@arm.com>
16594 David Sherwood <david.sherwood@arm.com>
16596 * calls.c (load_register_parameters): Cope with polynomial
16597 mode sizes. Require a constant size for BLKmode parameters
16598 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
16599 forces a parameter to be padded at the lsb end in order to
16600 fill a complete number of words, require the parameter size
16601 to be ordered wrt UNITS_PER_WORD.
16603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16604 Alan Hayward <alan.hayward@arm.com>
16605 David Sherwood <david.sherwood@arm.com>
16607 * reload1.c (spill_stack_slot_width): Change element type
16608 from unsigned int to poly_uint64_pod.
16609 (alter_reg): Treat mode sizes as polynomial.
16611 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16612 Alan Hayward <alan.hayward@arm.com>
16613 David Sherwood <david.sherwood@arm.com>
16615 * reload.c (complex_word_subreg_p): New function.
16616 (reload_inner_reg_of_subreg, push_reload): Use it.
16618 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16619 Alan Hayward <alan.hayward@arm.com>
16620 David Sherwood <david.sherwood@arm.com>
16622 * lra-constraints.c (process_alt_operands): Reject matched
16623 operands whose sizes aren't ordered.
16624 (match_reload): Refer to this check here.
16626 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16627 Alan Hayward <alan.hayward@arm.com>
16628 David Sherwood <david.sherwood@arm.com>
16630 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
16631 that the mode size is in the set {1, 2, 4, 8, 16}.
16633 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16634 Alan Hayward <alan.hayward@arm.com>
16635 David Sherwood <david.sherwood@arm.com>
16637 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
16638 Use plus_constant instead of gen_rtx_PLUS.
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 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
16645 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
16646 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
16647 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
16648 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
16649 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
16650 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
16651 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
16652 * config/i386/i386.c (ix86_push_rounding): ...this new function.
16653 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
16655 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
16656 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
16657 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
16658 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
16659 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
16660 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
16661 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
16662 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
16663 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
16664 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
16666 * expr.c (emit_move_resolve_push): Treat the input and result
16667 of PUSH_ROUNDING as a poly_int64.
16668 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
16669 (emit_push_insn): Likewise.
16670 * lra-eliminations.c (mark_not_eliminable): Likewise.
16671 * recog.c (push_operand): Likewise.
16672 * reload1.c (elimination_effects): Likewise.
16673 * rtlanal.c (nonzero_bits1): Likewise.
16674 * calls.c (store_one_arg): Likewise. Require the padding to be
16675 known at compile time.
16677 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16678 Alan Hayward <alan.hayward@arm.com>
16679 David Sherwood <david.sherwood@arm.com>
16681 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
16682 Use plus_constant instead of gen_rtx_PLUS.
16684 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16685 Alan Hayward <alan.hayward@arm.com>
16686 David Sherwood <david.sherwood@arm.com>
16688 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
16689 rather than an int.
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 (expand_expr_real_1): Use tree_to_poly_uint64
16696 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
16697 via stack temporaries. Treat the mode size as polynomial too.
16699 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16700 Alan Hayward <alan.hayward@arm.com>
16701 David Sherwood <david.sherwood@arm.com>
16703 * expr.c (expand_expr_real_2): When handling conversions involving
16704 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
16705 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
16706 as a poly_uint64 too.
16708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16709 Alan Hayward <alan.hayward@arm.com>
16710 David Sherwood <david.sherwood@arm.com>
16712 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
16714 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16715 Alan Hayward <alan.hayward@arm.com>
16716 David Sherwood <david.sherwood@arm.com>
16718 * combine.c (can_change_dest_mode): Handle polynomial
16719 REGMODE_NATURAL_SIZE.
16720 * expmed.c (store_bit_field_1): Likewise.
16721 * expr.c (store_constructor): Likewise.
16722 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
16723 and polynomial REGMODE_NATURAL_SIZE.
16724 (gen_lowpart_common): Likewise.
16725 * reginfo.c (record_subregs_of_mode): Likewise.
16726 * rtlanal.c (read_modify_subreg_p): Likewise.
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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
16733 numbers of elements.
16735 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16736 Alan Hayward <alan.hayward@arm.com>
16737 David Sherwood <david.sherwood@arm.com>
16739 * match.pd: Cope with polynomial numbers of vector elements.
16741 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16742 Alan Hayward <alan.hayward@arm.com>
16743 David Sherwood <david.sherwood@arm.com>
16745 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
16746 in a POINTER_PLUS_EXPR.
16748 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16749 Alan Hayward <alan.hayward@arm.com>
16750 David Sherwood <david.sherwood@arm.com>
16752 * omp-simd-clone.c (simd_clone_subparts): New function.
16753 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
16754 (ipa_simd_modify_function_body): Likewise.
16756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16757 Alan Hayward <alan.hayward@arm.com>
16758 David Sherwood <david.sherwood@arm.com>
16760 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
16761 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
16762 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
16763 (expand_vector_condition, vector_element): Likewise.
16764 (subparts_gt): New function.
16765 (get_compute_type): Use subparts_gt.
16766 (count_type_subparts): Delete.
16767 (expand_vector_operations_1): Use subparts_gt instead of
16768 count_type_subparts.
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-data-refs.c (vect_no_alias_p): Replace with...
16775 (vect_compile_time_alias): ...this new function. Do the calculation
16776 on poly_ints rather than trees.
16777 (vect_prune_runtime_alias_test_list): Update call accordingly.
16779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16780 Alan Hayward <alan.hayward@arm.com>
16781 David Sherwood <david.sherwood@arm.com>
16783 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
16785 (vect_schedule_slp_instance): Likewise.
16787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16788 Alan Hayward <alan.hayward@arm.com>
16789 David Sherwood <david.sherwood@arm.com>
16791 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
16792 constant and extern definitions for variable-length vectors.
16793 (vect_get_constant_vectors): Note that the number of units
16794 is known to be constant.
16796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16797 Alan Hayward <alan.hayward@arm.com>
16798 David Sherwood <david.sherwood@arm.com>
16800 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
16801 of units as polynomial. Choose between WIDE and NARROW based
16804 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16805 Alan Hayward <alan.hayward@arm.com>
16806 David Sherwood <david.sherwood@arm.com>
16808 * tree-vect-stmts.c (simd_clone_subparts): New function.
16809 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
16811 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16812 Alan Hayward <alan.hayward@arm.com>
16813 David Sherwood <david.sherwood@arm.com>
16815 * tree-vect-stmts.c (vectorizable_call): Treat the number of
16816 vectors as polynomial. Use build_index_vector for
16817 IFN_GOMP_SIMD_LANE.
16819 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16820 Alan Hayward <alan.hayward@arm.com>
16821 David Sherwood <david.sherwood@arm.com>
16823 * tree-vect-stmts.c (get_load_store_type): Treat the number of
16824 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
16825 for variable-length vectors.
16826 (vectorizable_mask_load_store): Treat the number of units as
16827 polynomial, asserting that it is constant if the condition has
16828 already been enforced.
16829 (vectorizable_store, vectorizable_load): Likewise.
16831 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16832 Alan Hayward <alan.hayward@arm.com>
16833 David Sherwood <david.sherwood@arm.com>
16835 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
16836 of units as polynomial. Punt if we can't tell at compile time
16837 which vector contains the final result.
16839 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16840 Alan Hayward <alan.hayward@arm.com>
16841 David Sherwood <david.sherwood@arm.com>
16843 * tree-vect-loop.c (vectorizable_induction): Treat the number
16844 of units as polynomial. Punt on SLP inductions. Use an integer
16845 VEC_SERIES_EXPR for variable-length integer reductions. Use a
16846 cast of such a series for variable-length floating-point
16849 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16850 Alan Hayward <alan.hayward@arm.com>
16851 David Sherwood <david.sherwood@arm.com>
16853 * tree.h (build_index_vector): Declare.
16854 * tree.c (build_index_vector): New function.
16855 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
16856 of units as polynomial, forcibly converting it to a constant if
16857 vectorizable_reduction has already enforced the condition.
16858 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
16859 to create a {1,2,3,...} vector.
16860 (vectorizable_reduction): Treat the number of units as polynomial.
16861 Choose vectype_in based on the largest scalar element size rather
16862 than the smallest number of units. Enforce the restrictions
16865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16866 Alan Hayward <alan.hayward@arm.com>
16867 David Sherwood <david.sherwood@arm.com>
16869 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
16870 number of units as polynomial.
16872 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16873 Alan Hayward <alan.hayward@arm.com>
16874 David Sherwood <david.sherwood@arm.com>
16876 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
16877 * target.def (autovectorize_vector_sizes): Return the vector sizes
16878 by pointer, using vector_sizes rather than a bitmask.
16879 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
16880 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
16881 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
16883 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
16884 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
16885 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
16886 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
16887 * omp-general.c (omp_max_vf): Likewise.
16888 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16889 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
16890 * tree-vect-loop.c (vect_analyze_loop): Likewise.
16891 * tree-vect-slp.c (vect_slp_bb): Likewise.
16892 * doc/tm.texi: Regenerate.
16893 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
16895 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
16896 the vector size as a poly_uint64 rather than an unsigned int.
16897 (current_vector_size): Change from an unsigned int to a poly_uint64.
16898 (get_vectype_for_scalar_type): Update accordingly.
16899 * tree.h (build_truth_vector_type): Take the size and number of
16900 units as a poly_uint64 rather than an unsigned int.
16901 (build_vector_type): Add a temporary overload that takes
16902 the number of units as a poly_uint64 rather than an unsigned int.
16903 * tree.c (make_vector_type): Likewise.
16904 (build_truth_vector_type): Take the number of units as a poly_uint64
16905 rather than an unsigned int.
16907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16908 Alan Hayward <alan.hayward@arm.com>
16909 David Sherwood <david.sherwood@arm.com>
16911 * target.def (get_mask_mode): Take the number of units and length
16912 as poly_uint64s rather than unsigned ints.
16913 * targhooks.h (default_get_mask_mode): Update accordingly.
16914 * targhooks.c (default_get_mask_mode): Likewise.
16915 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
16916 * doc/tm.texi: Regenerate.
16918 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16919 Alan Hayward <alan.hayward@arm.com>
16920 David Sherwood <david.sherwood@arm.com>
16922 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
16923 * omp-general.c (omp_max_vf): Likewise.
16924 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
16925 (expand_omp_simd): Handle polynomial safelen.
16926 * omp-low.c (omplow_simd_context): Add a default constructor.
16927 (omplow_simd_context::max_vf): Change from int to poly_uint64.
16928 (lower_rec_simd_input_clauses): Update accordingly.
16929 (lower_rec_input_clauses): Likewise.
16931 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16932 Alan Hayward <alan.hayward@arm.com>
16933 David Sherwood <david.sherwood@arm.com>
16935 * tree-vectorizer.h (vect_nunits_for_cost): New function.
16936 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
16937 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
16938 (vect_analyze_slp_cost): Likewise.
16939 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
16940 (vect_model_load_cost): Likewise.
16942 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16943 Alan Hayward <alan.hayward@arm.com>
16944 David Sherwood <david.sherwood@arm.com>
16946 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
16947 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
16948 from an unsigned int * to a poly_uint64_pod *.
16949 (calculate_unrolling_factor): New function.
16950 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
16952 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16953 Alan Hayward <alan.hayward@arm.com>
16954 David Sherwood <david.sherwood@arm.com>
16956 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
16957 from an unsigned int to a poly_uint64.
16958 (_loop_vec_info::slp_unrolling_factor): Likewise.
16959 (_loop_vec_info::vectorization_factor): Change from an int
16961 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
16962 (vect_get_num_vectors): New function.
16963 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
16964 (vect_get_num_copies): Use vect_get_num_vectors.
16965 (vect_analyze_data_ref_dependences): Change max_vf from an int *
16966 to an unsigned int *.
16967 (vect_analyze_data_refs): Change min_vf from an int * to a
16969 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16970 than an unsigned HOST_WIDE_INT.
16971 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
16972 (vect_analyze_data_ref_dependence): Change max_vf from an int *
16973 to an unsigned int *.
16974 (vect_analyze_data_ref_dependences): Likewise.
16975 (vect_compute_data_ref_alignment): Handle polynomial vf.
16976 (vect_enhance_data_refs_alignment): Likewise.
16977 (vect_prune_runtime_alias_test_list): Likewise.
16978 (vect_shift_permute_load_chain): Likewise.
16979 (vect_supportable_dr_alignment): Likewise.
16980 (dependence_distance_ge_vf): Take the vectorization factor as a
16981 poly_uint64 rather than an unsigned HOST_WIDE_INT.
16982 (vect_analyze_data_refs): Change min_vf from an int * to a
16984 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
16985 vfm1 as a poly_uint64 rather than an int. Make the same change
16986 for the returned bound_scalar.
16987 (vect_gen_vector_loop_niters): Handle polynomial vf.
16988 (vect_do_peeling): Likewise. Update call to
16989 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
16990 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
16992 * tree-vect-loop.c (vect_determine_vectorization_factor)
16993 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
16994 (vect_get_known_peeling_cost): Likewise.
16995 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
16996 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
16997 (vect_transform_loop): Likewise. Use the lowest possible VF when
16998 updating the upper bounds of the loop.
16999 (vect_min_worthwhile_factor): Make static. Return an unsigned int
17000 rather than an int.
17001 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
17002 polynomial unroll factors.
17003 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
17004 (vect_make_slp_decision): Likewise.
17005 (vect_supported_load_permutation_p): Likewise, and polynomial
17007 (vect_analyze_slp_cost): Handle polynomial vf.
17008 (vect_slp_analyze_node_operations): Likewise.
17009 (vect_slp_analyze_bb_1): Likewise.
17010 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17011 than an unsigned HOST_WIDE_INT.
17012 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
17013 (vectorizable_load): Handle polynomial vf.
17014 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
17016 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
17018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17019 Alan Hayward <alan.hayward@arm.com>
17020 David Sherwood <david.sherwood@arm.com>
17022 * match.pd: Handle bit operations involving three constants
17023 and try to fold one pair.
17025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17027 * tree-vect-loop-manip.c: Include gimple-fold.h.
17028 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
17029 niters_maybe_zero parameters. Handle other cases besides a step of 1.
17030 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
17031 Add a path that uses a step of VF instead of 1, but disable it
17033 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
17034 and niters_no_overflow parameters. Update calls to
17035 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
17036 Create a new SSA name if the latter choses to use a ste other
17037 than zero, and return it via niters_vector_mult_vf_var.
17038 * tree-vect-loop.c (vect_transform_loop): Update calls to
17039 vect_do_peeling, vect_gen_vector_loop_niters and
17040 slpeel_make_loop_iterate_ntimes.
17041 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
17042 (vect_gen_vector_loop_niters): Update declarations after above changes.
17044 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
17046 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
17047 128-bit round to integer instructions.
17048 (ceil<mode>2): Likewise.
17049 (btrunc<mode>2): Likewise.
17050 (round<mode>2): Likewise.
17052 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17054 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
17055 unaligned VSX load/store on P8/P9.
17056 (expand_block_clear): Allow the use of unaligned VSX
17057 load/store on P8/P9.
17059 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17061 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
17063 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
17064 swap associated with both a load and a store.
17066 2018-01-02 Andrew Waterman <andrew@sifive.com>
17068 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
17069 * config/riscv/riscv.md (clear_cache): Use it.
17071 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
17073 * web.c: Remove out-of-date comment.
17075 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17077 * expr.c (fixup_args_size_notes): Check that any existing
17078 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
17079 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
17080 (emit_single_push_insn): ...here.
17082 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17084 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
17085 (const_vector_encoded_nelts): New function.
17086 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
17087 (const_vector_int_elt, const_vector_elt): Declare.
17088 * emit-rtl.c (const_vector_int_elt_1): New function.
17089 (const_vector_elt): Likewise.
17090 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
17091 of CONST_VECTOR_ELT.
17093 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17095 * expr.c: Include rtx-vector-builder.h.
17096 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
17097 directly on the tree encoding.
17098 (const_vector_from_tree): Likewise.
17099 * optabs.c: Include rtx-vector-builder.h.
17100 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
17101 sequence of "u" values.
17102 * vec-perm-indices.c: Include rtx-vector-builder.h.
17103 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
17104 directly on the vec_perm_indices encoding.
17106 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17108 * doc/rtl.texi (const_vector): Describe new encoding scheme.
17109 * Makefile.in (OBJS): Add rtx-vector-builder.o.
17110 * rtx-vector-builder.h: New file.
17111 * rtx-vector-builder.c: Likewise.
17112 * rtl.h (rtx_def::u2): Add a const_vector field.
17113 (CONST_VECTOR_NPATTERNS): New macro.
17114 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
17115 (CONST_VECTOR_DUPLICATE_P): Likewise.
17116 (CONST_VECTOR_STEPPED_P): Likewise.
17117 (CONST_VECTOR_ENCODED_ELT): Likewise.
17118 (const_vec_duplicate_p): Check for a duplicated vector encoding.
17119 (unwrap_const_vec_duplicate): Likewise.
17120 (const_vec_series_p): Check for a non-duplicated vector encoding.
17121 Say that the function only returns true for integer vectors.
17122 * emit-rtl.c: Include rtx-vector-builder.h.
17123 (gen_const_vec_duplicate_1): Delete.
17124 (gen_const_vector): Call gen_const_vec_duplicate instead of
17125 gen_const_vec_duplicate_1.
17126 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
17127 (gen_const_vec_duplicate): Use rtx_vector_builder.
17128 (gen_const_vec_series): Likewise.
17129 (gen_rtx_CONST_VECTOR): Likewise.
17130 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
17131 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17132 Build a new vector rather than modifying a CONST_VECTOR in-place.
17133 (handle_special_swappables): Update call accordingly.
17134 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
17135 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17136 Build a new vector rather than modifying a CONST_VECTOR in-place.
17137 (handle_special_swappables): Update call accordingly.
17139 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17141 * simplify-rtx.c (simplify_const_binary_operation): Use
17142 CONST_VECTOR_ELT instead of XVECEXP.
17144 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17146 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
17147 the selector elements to be different from the data elements
17148 if the selector is a VECTOR_CST.
17149 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
17150 ssizetype for the selector.
17152 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17154 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
17155 before testing each element individually.
17156 * tree-vect-generic.c (lower_vec_perm): Likewise.
17158 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17160 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
17161 * selftest-run-tests.c (selftest::run_tests): Call it.
17162 * vector-builder.h (vector_builder::operator ==): New function.
17163 (vector_builder::operator !=): Likewise.
17164 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
17165 (vec_perm_indices::all_from_input_p): New function.
17166 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17167 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
17168 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
17169 instead of reading the VECTOR_CST directly. Detect whether both
17170 vector inputs are the same before constructing the vec_perm_indices,
17171 and update the number of inputs argument accordingly. Use the
17172 utility functions added above. Only construct sel2 if we need to.
17174 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17176 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
17177 the broadcast of the low byte.
17178 (expand_mult_highpart): Use an explicit encoding for the permutes.
17179 * optabs-query.c (can_mult_highpart_p): Likewise.
17180 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
17181 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17182 (vectorizable_bswap): Likewise.
17183 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
17184 explicit encoding for the power-of-2 permutes.
17185 (vect_permute_store_chain): Likewise.
17186 (vect_grouped_load_supported): Likewise.
17187 (vect_permute_load_chain): Likewise.
17189 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17191 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
17192 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
17193 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
17194 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17195 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
17196 (vect_gen_perm_mask_any): Likewise.
17198 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17200 * int-vector-builder.h: New file.
17201 * vec-perm-indices.h: Include int-vector-builder.h.
17202 (vec_perm_indices): Redefine as an int_vector_builder.
17203 (auto_vec_perm_indices): Delete.
17204 (vec_perm_builder): Redefine as a stand-alone class.
17205 (vec_perm_indices::vec_perm_indices): New function.
17206 (vec_perm_indices::clamp): Likewise.
17207 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
17208 (vec_perm_indices::new_vector): New function.
17209 (vec_perm_indices::new_expanded_vector): Update for new
17210 vec_perm_indices class.
17211 (vec_perm_indices::rotate_inputs): New function.
17212 (vec_perm_indices::all_in_range_p): Operate directly on the
17213 encoded form, without computing elided elements.
17214 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
17215 encoding. Update for new vec_perm_indices class.
17216 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
17217 the given vec_perm_builder.
17218 (expand_vec_perm_var): Update vec_perm_builder constructor.
17219 (expand_mult_highpart): Use vec_perm_builder instead of
17220 auto_vec_perm_indices.
17221 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
17222 vec_perm_indices instead of auto_vec_perm_indices. Use a single
17223 or double series encoding as appropriate.
17224 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
17225 vec_perm_indices instead of auto_vec_perm_indices.
17226 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17227 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17228 (vect_permute_store_chain): Likewise.
17229 (vect_grouped_load_supported): Likewise.
17230 (vect_permute_load_chain): Likewise.
17231 (vect_shift_permute_load_chain): Likewise.
17232 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17233 (vect_transform_slp_perm_load): Likewise.
17234 (vect_schedule_slp_instance): Likewise.
17235 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17236 (vectorizable_mask_load_store): Likewise.
17237 (vectorizable_bswap): Likewise.
17238 (vectorizable_store): Likewise.
17239 (vectorizable_load): Likewise.
17240 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
17241 vec_perm_indices instead of auto_vec_perm_indices. Use
17242 tree_to_vec_perm_builder to read the vector from a tree.
17243 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
17244 vec_perm_builder instead of a vec_perm_indices.
17245 (have_whole_vector_shift): Use vec_perm_builder and
17246 vec_perm_indices instead of auto_vec_perm_indices. Leave the
17247 truncation to calc_vec_perm_mask_for_shift.
17248 (vect_create_epilog_for_reduction): Likewise.
17249 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
17250 from auto_vec_perm_indices to vec_perm_indices.
17251 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17252 instead of changing individual elements.
17253 (aarch64_vectorize_vec_perm_const): Use new_vector to install
17254 the vector in d.perm.
17255 * config/arm/arm.c (expand_vec_perm_d::perm): Change
17256 from auto_vec_perm_indices to vec_perm_indices.
17257 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17258 instead of changing individual elements.
17259 (arm_vectorize_vec_perm_const): Use new_vector to install
17260 the vector in d.perm.
17261 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
17262 Update vec_perm_builder constructor.
17263 (rs6000_expand_interleave): Likewise.
17264 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
17265 (rs6000_expand_interleave): Likewise.
17267 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17269 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
17270 to qimode could truncate the indices.
17271 * optabs.c (expand_vec_perm_var): Likewise.
17273 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17275 * Makefile.in (OBJS): Add vec-perm-indices.o.
17276 * vec-perm-indices.h: New file.
17277 * vec-perm-indices.c: Likewise.
17278 * target.h (vec_perm_indices): Replace with a forward class
17280 (auto_vec_perm_indices): Move to vec-perm-indices.h.
17281 * optabs.h: Include vec-perm-indices.h.
17282 (expand_vec_perm): Delete.
17283 (selector_fits_mode_p, expand_vec_perm_var): Declare.
17284 (expand_vec_perm_const): Declare.
17285 * target.def (vec_perm_const_ok): Replace with...
17286 (vec_perm_const): ...this new hook.
17287 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
17288 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
17289 * doc/tm.texi: Regenerate.
17290 * optabs.def (vec_perm_const): Delete.
17291 * doc/md.texi (vec_perm_const): Likewise.
17292 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
17293 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
17294 expand_vec_perm for constant permutation vectors. Assert that
17295 the mode of variable permutation vectors is the integer equivalent
17296 of the mode that is being permuted.
17297 * optabs-query.h (selector_fits_mode_p): Declare.
17298 * optabs-query.c: Include vec-perm-indices.h.
17299 (selector_fits_mode_p): New function.
17300 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
17301 is defined, instead of checking whether the vec_perm_const_optab
17302 exists. Use targetm.vectorize.vec_perm_const instead of
17303 targetm.vectorize.vec_perm_const_ok. Check whether the indices
17304 fit in the vector mode before using a variable permute.
17305 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
17306 vec_perm_indices instead of an rtx.
17307 (expand_vec_perm): Replace with...
17308 (expand_vec_perm_const): ...this new function. Take the selector
17309 as a vec_perm_indices rather than an rtx. Also take the mode of
17310 the selector. Update call to shift_amt_for_vec_perm_mask.
17311 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
17312 Use vec_perm_indices::new_expanded_vector to expand the original
17313 selector into bytes. Check whether the indices fit in the vector
17314 mode before using a variable permute.
17315 (expand_vec_perm_var): Make global.
17316 (expand_mult_highpart): Use expand_vec_perm_const.
17317 * fold-const.c: Includes vec-perm-indices.h.
17318 * tree-ssa-forwprop.c: Likewise.
17319 * tree-vect-data-refs.c: Likewise.
17320 * tree-vect-generic.c: Likewise.
17321 * tree-vect-loop.c: Likewise.
17322 * tree-vect-slp.c: Likewise.
17323 * tree-vect-stmts.c: Likewise.
17324 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
17326 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
17327 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
17328 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
17329 (aarch64_vectorize_vec_perm_const): ...this new function.
17330 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17331 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17332 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
17333 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
17334 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17335 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17336 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
17338 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
17339 check for NEON modes.
17340 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
17341 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
17342 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
17343 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
17345 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
17346 the old VEC_PERM_CONST conditions.
17347 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
17348 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
17349 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
17350 (ia64_vectorize_vec_perm_const_ok): Merge into...
17351 (ia64_vectorize_vec_perm_const): ...this new function.
17352 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
17353 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
17354 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
17355 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
17356 * config/mips/mips.c (mips_expand_vec_perm_const)
17357 (mips_vectorize_vec_perm_const_ok): Merge into...
17358 (mips_vectorize_vec_perm_const): ...this new function.
17359 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
17360 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
17361 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
17362 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
17363 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
17364 (rs6000_expand_vec_perm_const): Delete.
17365 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
17367 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17368 (altivec_expand_vec_perm_const_le): Take each operand individually.
17369 Operate on constant selectors rather than rtxes.
17370 (altivec_expand_vec_perm_const): Likewise. Update call to
17371 altivec_expand_vec_perm_const_le.
17372 (rs6000_expand_vec_perm_const): Delete.
17373 (rs6000_vectorize_vec_perm_const_ok): Delete.
17374 (rs6000_vectorize_vec_perm_const): New function.
17375 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17376 an element count and rtx array.
17377 (rs6000_expand_extract_even): Update call accordingly.
17378 (rs6000_expand_interleave): Likewise.
17379 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
17380 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
17381 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
17382 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
17383 (rs6000_expand_vec_perm_const): Delete.
17384 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17385 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17386 (altivec_expand_vec_perm_const_le): Take each operand individually.
17387 Operate on constant selectors rather than rtxes.
17388 (altivec_expand_vec_perm_const): Likewise. Update call to
17389 altivec_expand_vec_perm_const_le.
17390 (rs6000_expand_vec_perm_const): Delete.
17391 (rs6000_vectorize_vec_perm_const_ok): Delete.
17392 (rs6000_vectorize_vec_perm_const): New function. Remove stray
17393 reference to the SPE evmerge intructions.
17394 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17395 an element count and rtx array.
17396 (rs6000_expand_extract_even): Update call accordingly.
17397 (rs6000_expand_interleave): Likewise.
17398 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
17399 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
17401 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17403 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17405 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
17406 vector mode and that that mode matches the mode of the data
17408 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
17409 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
17410 directly using expand_vec_perm_1 when forcing selectors into
17412 (expand_vec_perm_var): New function, split out from expand_vec_perm.
17414 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17416 * optabs-query.h (can_vec_perm_p): Delete.
17417 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
17418 * optabs-query.c (can_vec_perm_p): Split into...
17419 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
17420 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
17421 particular selector is valid.
17422 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17423 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17424 (vect_grouped_load_supported): Likewise.
17425 (vect_shift_permute_load_chain): Likewise.
17426 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17427 (vect_transform_slp_perm_load): Likewise.
17428 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17429 (vectorizable_bswap): Likewise.
17430 (vect_gen_perm_mask_checked): Likewise.
17431 * fold-const.c (fold_ternary_loc): Likewise. Don't take
17432 implementations of variable permutation vectors into account
17433 when deciding which selector to use.
17434 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
17435 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
17436 with a false third argument.
17437 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
17438 to test whether the constant selector is valid and can_vec_perm_var_p
17439 to test whether a variable selector is valid.
17441 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17443 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
17444 * optabs-query.c (can_vec_perm_p): Likewise.
17445 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
17446 instead of vec_perm_indices.
17447 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
17448 (vect_gen_perm_mask_checked): Likewise,
17449 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
17450 (vect_gen_perm_mask_checked): Likewise,
17452 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17454 * optabs-query.h (qimode_for_vec_perm): Declare.
17455 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
17456 (qimode_for_vec_perm): ...this new function.
17457 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
17459 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17461 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
17462 does not have a conditional at the top.
17464 2018-01-02 Richard Biener <rguenther@suse.de>
17466 * ipa-inline.c (big_speedup_p): Fix expression.
17468 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17471 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
17474 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17478 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
17479 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
17480 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
17481 cond_taken_branch_cost 3->4.
17483 2018-01-01 Jakub Jelinek <jakub@redhat.com>
17485 PR tree-optimization/83581
17486 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
17487 TODO_cleanup_cfg if any changes have been made.
17489 PR middle-end/83608
17490 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
17491 convert_modes if target mode has the right side, but different mode
17494 PR middle-end/83609
17495 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
17496 last argument when extracting from CONCAT. If either from_real or
17497 from_imag is NULL, use expansion through memory. If result is not
17498 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
17499 the parts directly to inner mode, if even that fails, use expansion
17502 PR middle-end/83623
17503 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
17504 check for bswap in mode rather than HImode and use that in expand_unop
17507 Copyright (C) 2018 Free Software Foundation, Inc.
17509 Copying and distribution of this file, with or without modification,
17510 are permitted in any medium without royalty provided the copyright
17511 notice and this notice are preserved.