1 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3 * tree-inline.c (copy_edges_for_bb): Minor tweak.
4 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
5 debug statement when resetting its value.
6 (expand_call_inline): Copy the locus of the call onto the assignment
7 of the return value, if any. Use local variable in more cases.
9 2018-06-21 Martin Liska <mliska@suse.cz>
11 * ipa-pure-const.c (propagate_nothrow): Use
12 funct_state_summaries->get.
13 (dump_malloc_lattice): Likewise.
14 (propagate_malloc): Likewise.
16 2018-06-21 Richard Biener <rguenther@suse.de>
18 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
19 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
20 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
22 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
23 BLOCK_ABSTRACT_ORIGIN unconditionally.
25 2018-06-21 David Malcolm <dmalcolm@redhat.com>
27 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
29 * ipa-reference.c (ipa_reference_c_finalize): Delete
30 ipa_ref_opt_sum_summaries and set it to NULL.
32 2018-06-21 Tom de Vries <tdevries@suse.de>
34 PR tree-optimization/85859
35 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
36 test with comment from bb_no_side_effects_p.
38 2018-06-21 Richard Biener <rguenther@suse.de>
40 PR tree-optimization/86232
41 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
42 max for constant niter.
44 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
46 * config/aarch64/aarch64-simd.md
47 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
49 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
51 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
52 Make opernads of the unspec commutative.
54 2018-06-21 Richard Biener <rguenther@suse.de>
56 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
57 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
58 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
59 (vect_analyze_data_ref_dependence): Re-order checks to deal with
61 (vect_record_base_alignments): Do not record base alignment
62 for gathers or scatters.
63 (vect_compute_data_ref_alignment): Drop return value that is always
64 true. Bail out early for gathers or scatters.
65 (vect_enhance_data_refs_alignment): Bail out early for gathers
67 (vect_find_same_alignment_drs): Likewise.
68 (vect_analyze_data_refs_alignment): Remove dead code.
69 (vect_slp_analyze_and_verify_node_alignment): Likewise.
70 (vect_analyze_data_refs): For possible gathers or scatters do
71 not create an alternate DR, just check their possible validity
72 and mark them. Adjust DECL_NONALIASED handling to not rely
74 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
75 update inits of gathers or scatters.
76 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
77 Also copy gather/scatter flag to pattern vinfo.
79 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
81 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
82 behavior of vec_packsu (vector unsigned long long, vector unsigned
83 long long) to match behavior of vec_packs with same signature.
85 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
86 Thomas Schwinge <thomas@codesourcery.com>
87 Cesar Philippidis <cesar@codesourcery.com>
89 * gimplify.c (gimplify_scan_omp_clauses): Add support for
90 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
91 (gimplify_adjust_omp_clauses): Likewise.
92 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
93 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
94 (gimplify_omp_target_update): Update handling of acc update and
96 * omp-low.c (install_var_field): Remove unused parameter
97 base_pointers_restrict.
98 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
99 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
101 (omp_target_base_pointers_restrict_p): Delete.
102 (scan_omp_target): Update call to scan_sharing_clauses.
103 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
105 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
106 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
107 (convert_local_omp_clauses): Likewise.
108 * tree-pretty-print.c (dump_omp_clause): Likewise.
109 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
111 (omp_clause_code_name): Likewise.
113 2018-06-20 Jakub Jelinek <jakub@redhat.com>
116 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
119 PR tree-optimization/86231
120 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
121 anti-range don't overwrite *vr0min before using it to compute *vr0max.
123 2018-06-20 Tom de Vries <tdevries@suse.de>
125 PR tree-optimization/86097
126 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
127 iv type if signedness of iv type is not the same as that of *nit.
129 2018-06-20 Jakub Jelinek <jakub@redhat.com>
131 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
132 EDGE_EH edges, verify they are all EDGE_EH.
134 2018-06-20 Maya Rashish <coypu@sdf.org>
136 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
138 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
140 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
141 * config/aarch64/aarch64.c (xgene1_tunings): Add
142 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
143 (aarch64_mode_valid_for_sched_fusion_p):
145 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
146 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
148 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
150 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
151 * config/aarch64/iterators.md (VQ2): New mode iterator.
153 2018-06-20 Martin Liska <mliska@suse.cz>
155 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
156 Change default ratio from 10 to 8.
158 2018-06-20 Martin Liska <mliska@suse.cz>
160 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
162 (bit_test_cluster::find_bit_tests): Likewise.
163 (switch_decision_tree::analyze_switch_statement): Find clusters.
164 * tree-switch-conversion.h (struct jump_table_cluster): Document
167 2018-06-20 Martin Liska <mliska@suse.cz>
169 * tree-switch-conversion.c (switch_conversion::collect):
170 Record m_uniq property.
171 (switch_conversion::expand): Bail out for special conditions.
172 (group_cluster::~group_cluster): New.
173 (group_cluster::group_cluster): Likewise.
174 (group_cluster::dump): Likewise.
175 (jump_table_cluster::emit): New.
176 (switch_decision_tree::fix_phi_operands_for_edges): New.
177 (struct case_node): Remove struct.
178 (jump_table_cluster::can_be_handled): New.
179 (case_values_threshold): Moved to header.
180 (reset_out_edges_aux): Likewise.
181 (jump_table_cluster::is_beneficial): New.
182 (bit_test_cluster::can_be_handled): Likewise.
183 (add_case_node): Remove.
184 (bit_test_cluster::is_beneficial): New.
185 (case_bit_test::cmp): New.
186 (bit_test_cluster::emit): New.
187 (expand_switch_as_decision_tree_p): Remove.
188 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
189 (fix_phi_operands_for_edge): Likewise.
190 (switch_decision_tree::analyze_switch_statement): New.
191 (compute_cases_per_edge): Move ...
192 (switch_decision_tree::compute_cases_per_edge): ... here.
193 (try_switch_expansion): Likewise.
194 (switch_decision_tree::try_switch_expansion): Likewise.
195 (record_phi_operand_mapping): Likewise.
196 (switch_decision_tree::record_phi_operand_mapping): Likewise.
197 (emit_case_decision_tree): Likewise.
198 (switch_decision_tree::emit): Likewise.
199 (balance_case_nodes): Likewise.
200 (switch_decision_tree::balance_case_nodes): Likewise.
201 (dump_case_nodes): Likewise.
202 (switch_decision_tree::dump_case_nodes): Likewise.
203 (emit_jump): Likewise.
204 (switch_decision_tree::emit_jump): Likewise.
205 (emit_cmp_and_jump_insns): Likewise.
206 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
207 (emit_case_nodes): Likewise.
208 (switch_decision_tree::emit_case_nodes): Likewise.
209 (conditional_probability): Remove.
210 * tree-switch-conversion.h (enum cluster_type): New.
212 (struct cluster): Likewise.
213 (cluster::cluster): Likewise.
214 (struct simple_cluster): Likewise.
215 (simple_cluster::simple_cluster): Likewise.
216 (struct group_cluster): Likewise.
217 (struct jump_table_cluster): Likewise.
218 (struct bit_test_cluster): Likewise.
219 (struct min_cluster_item): Likewise.
220 (struct case_tree_node): Likewise.
221 (case_tree_node::case_tree_node): Likewise.
222 (jump_table_cluster::case_values_threshold): Likewise.
223 (struct case_bit_test): Likewise.
224 (struct switch_decision_tree): Likewise.
225 (struct switch_conversion): Likewise.
226 (switch_decision_tree::reset_out_edges_aux): Likewise.
228 2018-06-20 Martin Liska <mliska@suse.cz>
230 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
231 (hoist_edge_and_branch_if_true): Likewise.
232 (expand_switch_using_bit_tests_p): Likewise.
233 (struct case_bit_test): Likewise.
234 (case_bit_test_cmp): Likewise.
235 (emit_case_bit_tests): Likewise.
236 (switch_conversion::switch_conversion): New class.
237 (struct switch_conv_info): Remove old struct.
238 (collect_switch_conv_info): More to ...
239 (switch_conversion::collect): ... this.
240 (check_range): Likewise.
241 (switch_conversion::check_range): Likewise.
242 (check_all_empty_except_final): Likewise.
243 (switch_conversion::check_all_empty_except_final): Likewise.
244 (check_final_bb): Likewise.
245 (switch_conversion::check_final_bb): Likewise.
246 (create_temp_arrays): Likewise.
247 (switch_conversion::create_temp_arrays): Likewise.
248 (free_temp_arrays): Likewise.
249 (gather_default_values): Likewise.
250 (switch_conversion::gather_default_values): Likewise.
251 (build_constructors): Likewise.
252 (switch_conversion::build_constructors): Likewise.
253 (constructor_contains_same_values_p): Likewise.
254 (switch_conversion::contains_same_values_p): Likewise.
255 (array_value_type): Likewise.
256 (switch_conversion::array_value_type): Likewise.
257 (build_one_array): Likewise.
258 (switch_conversion::build_one_array): Likewise.
259 (build_arrays): Likewise.
260 (switch_conversion::build_arrays): Likewise.
261 (gen_def_assigns): Likewise.
262 (switch_conversion::gen_def_assigns): Likewise.
263 (prune_bbs): Likewise.
264 (switch_conversion::prune_bbs): Likewise.
265 (fix_phi_nodes): Likewise.
266 (switch_conversion::fix_phi_nodes): Likewise.
267 (gen_inbound_check): Likewise.
268 (switch_conversion::gen_inbound_check): Likewise.
269 (process_switch): Use the newly created class.
270 (switch_conversion::expand): New.
271 (switch_conversion::~switch_conversion): New.
272 * tree-switch-conversion.h: New file.
274 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
276 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
277 tree-vect-patterns.c.
278 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
279 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
280 (vect_recog_sad_pattern): Likewise.
281 (vect_recog_widen_sum_pattern): Likewise.
282 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
283 (vect_recog_widen_shift_pattern): Remove the type_in argument.
284 (vect_recog_rotate_pattern): Likewise.
285 (vect_recog_mult_pattern): Likewise.
286 (vect_recog_vector_vector_shift_pattern): Likewise.
287 (vect_recog_divmod_pattern): Likewise.
288 (vect_recog_mixed_size_cond_pattern): Likewise.
289 (vect_recog_bool_pattern): Likewise.
290 (vect_recog_mask_conversion_pattern): Likewise.
291 (vect_try_gather_scatter_pattern): Likewise.
292 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
293 (vect_recog_over_widening_pattern): Likewise.
294 (vect_recog_gather_scatter_pattern): Likewise.
295 (vect_recog_func_ptr): Move from tree-vectorizer.h
296 (vect_vect_recog_func_ptrs): Move further down the file.
297 (vect_recog_func): Likewise. Remove the third argument.
298 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
299 (vect_pattern_recog_1): Expect the pattern function to do any
300 necessary target tests. Also expect it to provide a vector type.
301 Remove the type_in handling.
303 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
305 * tree-vect-patterns.c (vect_pattern_detected): New function.
306 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
307 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
308 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
309 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
310 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
311 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
312 (vect_recog_mask_conversion_pattern)
313 (vect_try_gather_scatter_pattern): Likewise.
315 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
317 * tree-vect-patterns.c (vect_get_internal_def): New function.
318 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
319 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
320 (search_type_for_mask_1): Use it.
322 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
324 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
325 redundant WIDEN_SUM_EXPR handling.
326 (vect_recog_sad_pattern): Likewise.
328 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
330 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
331 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
332 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
333 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
334 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
336 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
338 * tree-vect-stmts.c (vectorizable_call): Make sure that we
339 use the stmt_vec_info of the original bb statement for the
340 new zero assignment, even if the call is part of a pattern.
342 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
344 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
345 that the sequence is attached to the original statement rather
346 than the pattern statement.
347 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
348 PATTERN_DEF_SEQ from the original statement rather than
349 the main pattern statement.
350 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
351 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
352 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
354 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
356 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
357 definition statements before the early exit for statements that aren't
359 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
361 (vect_transform_loop): ...here. Process pattern definition
362 statements without first checking whether the main pattern
363 statement is live or relevant.
365 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
367 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
368 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
370 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
372 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
373 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
374 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
375 (expand_block_compare): Change select_block_compare_mode call.
376 (expand_strncmp_align_check): Use new functions, fix comment.
377 (emit_final_str_compare_gpr): New function.
378 (expand_strn_compare): Refactor and clean up code.
379 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
381 2018-06-19 Tony Reix <tony.reix@atos.com>
382 Damien Bergamini <damien.bergamini@atos.com>
383 David Edelsohn <dje.gcc@gmail.com>
385 * collect2.c (static_obj): New variable.
386 (static_libs): New variable.
387 (is_in_list): Uncomment declaration.
388 (main): Track AIX libraries linked statically.
389 (is_in_list): Uncomment definition.
390 (scan_prog_file): Don't add AIX shared libraries initializer
391 to constructor list if linking statically.
393 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
395 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
397 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
399 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
401 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
404 2018-06-19 Martin Liska <mliska@suse.cz>
406 * config/i386/i386.c (ix86_can_inline_p): Do not use
407 ipa_fn_summaries::get_create.
408 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
410 (devirtualization_time_bonus): Likewise.
411 (ipcp_propagate_stage): Likewise.
412 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
413 (edge_set_predicate): Likewise.
414 (evaluate_conditions_for_known_args): Likewise.
415 (evaluate_properties_for_edge): Likewise.
416 (ipa_call_summary::reset): Tranform to ...
417 (ipa_call_summary::~ipa_call_summary): ... this.
418 (ipa_fn_summary::reset): Transform to ...
419 (ipa_fn_summary::~ipa_fn_summary): ... this.
420 (ipa_fn_summary_t::remove): Rename to ...
421 (ipa_fn_summary_t::remove_callees): ... this.
422 (ipa_fn_summary_t::duplicate): Use placement new
423 instead of memory copy.
424 (ipa_call_summary_t::duplicate): Likewise.
425 (ipa_call_summary_t::remove): Remove.
426 (dump_ipa_call_summary): Change get_create to get.
427 (ipa_dump_fn_summary): Dump only when summary exists.
428 (analyze_function_body): Use symbol_summary::get instead
430 (compute_fn_summary): Likewise.
431 (estimate_edge_devirt_benefit): Likewise.
432 (estimate_edge_size_and_time): Likewise.
433 (inline_update_callee_summaries): Likewise.
434 (remap_edge_change_prob): Likewise.
435 (remap_edge_summaries): Likewise.
436 (ipa_merge_fn_summary_after_inlining): Likewise.
437 (write_ipa_call_summary): Likewise.
438 (ipa_fn_summary_write): Likewise.
439 (ipa_free_fn_summary): Likewise.
440 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
441 (struct ipa_call_summary): Likewise.
442 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
444 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
445 (estimate_size_after_inlining): Likewise.
446 (estimate_growth): Likewise.
447 (growth_likely_positive): Likewise.
448 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
449 (inline_call): Likewise.
450 * ipa-inline.c (caller_growth_limits): Likewise.
451 (can_inline_edge_p): Likewise.
452 (can_inline_edge_by_limits_p): Likewise.
453 (compute_uninlined_call_time): Likewise.
454 (compute_inlined_call_time): Likewise.
455 (want_inline_small_function_p): Likewise.
456 (edge_badness): Likewise.
457 (update_caller_keys): Likewise.
458 (update_callee_keys): Likewise.
459 (inline_small_functions): Likewise.
460 (inline_to_all_callers_1): Likewise.
461 (dump_overall_stats): Likewise.
462 (early_inline_small_functions): Likewise.
463 (early_inliner): Likewise.
464 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
465 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
466 * ipa-pure-const.c (malloc_candidate_p): Likewise.
467 * ipa-split.c (execute_split_functions): Likewise.
468 * symbol-summary.h: Likewise.
469 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
471 2018-06-19 Richard Biener <rguenther@suse.de>
473 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
474 (vectorize_loops): ... here. Fix dbgcnt handling.
475 (try_vectorize_loop): Wrap try_vectorize_loop_1.
477 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
480 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
481 ieee128 argument takes up only one (vector) register, not two (floating
484 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
486 * gimplify.c (gimplify_init_constructor): Really never clear for an
487 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
489 2018-06-19 Richard Biener <rguenther@suse.de>
491 PR tree-optimization/86179
492 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
493 after failed recognition.
495 2018-06-18 Martin Sebor <msebor@redhat.com>
498 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
499 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
500 Handle integer subtraction.
501 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
502 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
504 2018-06-18 David Malcolm <dmalcolm@redhat.com>
506 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
507 param from rtx to rtx_insn *.
508 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
510 (frv_ifcvt_modify_insn): Likwise.
511 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
512 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
513 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
514 as_a <rtx_insn *> cast to local "unprotected_region" once
515 it's been established that it's not NULL or pc_rtx.
516 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
517 param "sethi" from rtx to rtx_insn *.
518 (nds32_group_float_insns): Likewise for param "insn".
519 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
521 (vax_output_int_subtract): Likewise.
522 * config/vax/vax.c (vax_output_int_add): Likewise for param
524 (vax_output_int_subtract): Likewise.
525 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
526 (emit_pattern_after): Likewise for param "after".
527 (emit_insn_after): Likewise.
528 (emit_jump_insn_after): Likewise.
529 (emit_call_insn_after): Likewise.
530 (emit_debug_insn_after): Likewise.
531 (emit_pattern_before): Likewise for param "before".
532 (emit_insn_before): Likewise.
533 (emit_jump_insn_before): Likewise.
534 * final.c (get_insn_template): Likewise for param "insn", removing
536 * output.h (get_insn_template): Likewise for 2nd param.
537 * rtl.h (emit_insn_before): Likewise.
538 (emit_jump_insn_before): Likewise.
539 (emit_debug_insn_before_noloc): Likewise.
540 (emit_insn_after): Likewise.
541 (emit_jump_insn_after): Likewise.
542 (emit_call_insn_after): Likewise.
543 (emit_debug_insn_after): Likewise.
544 (set_insn_deleted): Likewise for param.
546 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
549 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
550 floating point modes, so that IFmode is numerically greater than
551 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
552 to declare the ordering. This prevents IFmode from being
553 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
554 machine. Include rs6000-modes.h to share the fractional values
555 between genmodes* and the rest of the compiler.
559 * config/rs6000/rs6000-modes.h: New file.
560 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
561 meaning of rs6000_long_double_size so that 126..128 selects an
562 appropriate 128-bit floating point type.
563 (rs6000_option_override_internal): Likewise.
564 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
565 (TARGET_LONG_DOUBLE_128): Change the meaning of
566 rs6000_long_double_size so that 126..128 selects an appropriate
567 128-bit floating point type.
568 (LONG_DOUBLE_TYPE_SIZE): Update comment.
569 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
570 source and destination to match the standard usage.
571 (truncifkf2): Likewise.
572 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
573 ISA 2.07 to use an explicit clobber, instead of passing in a
575 (copysign<mode>3_soft): Likewise.
577 2018-06-18 David Malcolm <dmalcolm@redhat.com>
579 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
580 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
581 (vect_slp_analyze_instance_dependence): Likewise.
582 (vect_enhance_data_refs_alignment): Likewise.
583 (vect_analyze_data_refs_alignment): Likewise.
584 (vect_slp_analyze_and_verify_instance_alignment
585 (vect_analyze_data_ref_accesses): Likewise.
586 (vect_prune_runtime_alias_test_list): Likewise.
587 (vect_analyze_data_refs): Likewise.
588 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
589 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
590 (vect_analyze_scalar_cycles_1): Likewise.
591 (vect_get_loop_niters): Likewise.
592 (vect_analyze_loop_form_1): Likewise.
593 (vect_update_vf_for_slp): Likewise.
594 (vect_analyze_loop_operations): Likewise.
595 (vect_analyze_loop): Likewise.
596 (vectorizable_induction): Likewise.
597 (vect_transform_loop): Likewise.
598 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
599 * tree-vect-slp.c (vect_analyze_slp): Likewise.
600 (vect_make_slp_decision): Likewise.
601 (vect_detect_hybrid_slp): Likewise.
602 (vect_slp_analyze_operations): Likewise.
603 (vect_slp_bb): Likewise.
604 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
605 (vectorizable_bswap): Likewise.
606 (vectorizable_call): Likewise.
607 (vectorizable_simd_clone_call): Likewise.
608 (vectorizable_conversion): Likewise.
609 (vectorizable_assignment): Likewise.
610 (vectorizable_shift): Likewise.
611 (vectorizable_operation): Likewise.
612 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
614 2018-06-18 Martin Sebor <msebor@redhat.com>
616 PR tree-optimization/81384
617 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
618 * builtins.c (expand_builtin_strnlen): New function.
619 (expand_builtin): Call it.
620 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
621 * builtins.def (BUILT_IN_STRNLEN): New.
622 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
623 Warn for bounds in excess of maximum object size.
624 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
625 single-value ranges. Handle strnlen.
626 (handle_builtin_strlen): Handle strnlen.
627 (strlen_check_and_optimize_stmt): Same.
628 * doc/extend.texi (Other Builtins): Document strnlen.
630 2018-06-18 Maya Rashish <coypu@sdf.org>
632 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
633 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
634 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
636 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
638 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
640 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
642 * tree.c (escaped_string::escape): Replace cast to char * by
643 const_cast<char *> (unescaped).
645 2018-06-18 Nick Clifton <nickc@redhat.com>
648 * tree.c (escaped_string): New class. Converts an unescaped
649 string into its escaped equivalent.
650 (warn_deprecated_use): Use the new class to convert the
651 deprecation message, if present.
652 (test_escaped_strings): New self test.
653 (test_c_tests): Add test_escaped_strings.
654 * doc/extend.texi (deprecated): Add a note that the
655 deprecation message is affected by the -fmessage-length
656 option, and that control characters will be escaped.
657 (#pragma GCC error): Document this pragma.
658 (#pragma GCC warning): Likewise.
659 * doc/invoke.texi (-fmessage-length): Document this option's
660 effect on the #warning and #error preprocessor directives and
661 the deprecated attribute.
663 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
665 * tree.c (decl_value_expr_lookup): Revert latest change.
666 (decl_value_expr_insert): Likewise.
668 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
670 * gimplify.c (nonlocal_vlas): Delete.
671 (nonlocal_vla_vars): Likewise.
672 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
674 (gimplify_body): Do not create and destroy nonlocal_vlas.
675 * tree-nested.c: Include diagnostic.h.
676 (use_pointer_in_frame): Tweak.
677 (lookup_field_for_decl): Add assertion and declare the transformation.
678 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
679 internal error when the reference is in a wrong context. Do not
680 create a debug decl by default.
681 (note_nonlocal_block_vlas): Delete.
682 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
683 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
684 create a debug decl by default.
685 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
686 call is in a wrong context.
687 (fixup_vla_decls): New function.
688 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
689 debug variables were created.
690 * tree.c (decl_value_expr_lookup): Add checking assertion.
691 (decl_value_expr_insert): Likewise.
693 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
696 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
697 * tree-scalar-evolution.c (interpret_expr): Likewise.
698 (expression_expensive_p): Likewise.
699 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
700 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
701 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
702 (ssa_defined_by_minus_one_stmt_p): New.
704 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
707 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
708 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
709 * dojump.c (do_jump): Likewise.
710 * expr.c (expand_expr_real_2): Check operand type's sign.
711 * fold-const.c (const_unop): Handle ABSU_EXPR.
712 (fold_abs_const): Likewise.
713 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
714 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
715 (strip_sign_op_1): Likesise.
716 * match.pd: Add new pattern to generate ABSU_EXPR.
717 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
718 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
719 * tree-eh.c (operation_could_trap_helper_p): Likewise.
720 * tree-inline.c (estimate_operator_cost): Likewise.
721 * tree-pretty-print.c (dump_generic_node): Likewise.
722 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
723 * tree.def (ABSU_EXPR): New.
725 2018-06-16 Jakub Jelinek <jakub@redhat.com>
728 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
729 documented as preserved for backward compatibility only.
730 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
732 PR rtl-optimization/86108
733 * bb-reorder.c (create_forwarder_block): Renamed to ...
734 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
735 jump from new landing pad to the second part.
736 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
739 2018-06-15 Jakub Jelinek <jakub@redhat.com>
742 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
743 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
744 Only call store_expr for halves if the mode is the same.
747 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
748 Fix up comment formatting.
750 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
752 * typed-splay-tree.h (typed_splay_tree::remove): New function.
753 (typed_splay_tree::closure,
754 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
755 (typed_splay_tree::typed_splay_tree,
756 typed_splay_tree::operator =): Declared private.
757 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
758 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
759 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
760 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
761 typed_splay_tree::splay_tree_splay,
762 typed_splay_tree::splay_tree_foreach_helper,
763 typed_splay_tree::splay_tree_insert,
764 typed_splay_tree::splay_tree_remove,
765 typed_splay_tree::splay_tree_lookup,
766 typed_splay_tree::splay_tree_predecessor,
767 typed_splay_tree::splay_tree_successor,
768 typed_splay_tree::splay_tree_min,
769 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
770 (typed_splay_tree::root, typed_splay_tree::comp,
771 typed_splay_tree::delete_key,
772 typed_splay_tree::delete_value): New data members.
773 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
774 typed_splay_tree::remove.
776 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
778 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
779 -mginv and -mno-ginv to the assembler.
780 * config/mips/mips.opt (-mcrc): New option.
782 * doc/invoke.text (-mcrc): Document.
785 2018-06-15 Nick Clifton <nickc@redhat.com>
788 * tree.c (escaped_string): New class. Converts an unescaped
789 string into its escaped equivalent.
790 (warn_deprecated_use): Use the new class to convert the
791 deprecation message, if present.
792 (test_escaped_strings): New self test.
793 (test_c_tests): Add test_escaped_strings.
794 * doc/extend.texi (deprecated): Add a note that the
795 deprecation message is affected by the -fmessage-length
796 option, and that control characters will be escaped.
797 (#pragma GCC error): Document this pragma.
798 (#pragma GCC warning): Likewise.
799 * doc/invoke.texi (-fmessage-length): Document this option's
800 effect on the #warning and #error preprocessor directives and
801 the deprecated attribute.
803 2018-06-15 Richard Biener <rguenther@suse.de>
805 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
806 here, also noting vector size used.
807 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
808 size used in MSG_OPTIMIZED_LOCATIONS dump.
809 (pass_slp_vectorize::execute): Adjust.
811 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
814 * config/arc/arc.c (arc_return_address_register): Fix
817 2018-06-15 Richard Biener <rguenther@suse.de>
820 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
821 leave useless conversion stripping to force_gimple_operand_gsi.
822 (gimplify_build2): Likewise.
823 (gimplify_build1): Likewise.
825 2018-06-15 Richard Biener <rguenther@suse.de>
828 * tree-cfg.c (move_stmt_op): unshare invariant addresses
829 before adjusting their block.
831 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
833 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
834 multilibs for *-*-rtems*.
835 * config/riscv/t-rtems: New file.
837 2018-06-14 Jakub Jelinek <jakub@redhat.com>
840 * match.pd ((A +- CST1) +- CST2): Punt if last resort
841 unsigned_type_for returns NULL.
844 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
845 subregs of multi-word pseudos unless the float mode has word size.
847 2018-06-14 Richard Biener <rguenther@suse.de>
850 * tree-vect-generic.c (build_word_mode_vector_type): Remove
851 duplicate and harmful type_hash_canon.
852 * tree.c (type_hash_canon): Assert we didn't find ourselves.
854 2018-06-14 Richard Biener <rguenther@suse.de>
857 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
860 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
862 * config/rtems.h (STDINT_LONG32): Define.
864 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
865 Prachi Godbole <prachi.godbole@imgtec.com>
867 * config/mips/mips-cpus.def: Define P6600.
868 * config/mips/mips-tables.opt: Regenerate.
869 * config/mips/mips.c (mips_ucbranch_type): New enum.
870 (mips_rtx_cost_data): Add support for P6600.
871 (mips_issue_rate): Likewise.
872 (mips_multipass_dfa_lookahead): Likewise.
873 (mips_avoid_hazard): Likewise.
874 (mips_reorg_process_insns): Likewise.
875 (mips_classify_branch_p6600): New function.
876 * config/mips/mips.h (TUNE_P6600): New define.
877 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
878 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
879 * config/mips/mips.md: Include p6600.md.
880 (processor): Add p6600.
881 * config/mips/p6600.md: New file.
882 * doc/invoke.texi: Add p6600 to supported architectures.
884 2018-06-13 Martin Sebor <msebor@redhat.com>
886 PR tree-optimization/86114
887 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
889 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
891 2018-06-13 Richard Biener <rguenther@suse.de>
893 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
894 Properly set vector type of the intermediate stmt.
895 * tree-vect-stmts.c (vectorizable_operation): The destination
896 var always has vectype_out type.
898 2018-06-13 Jeff Law <law@redhat.com>
900 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
901 integer 0 for argument to print_rtl_with_bb.
902 (rl78_reorg): Likewise.
904 2018-06-13 David Malcolm <dmalcolm@redhat.com>
906 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
907 from rtx to rtx_insn *.
908 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
910 (add_sched_insns_for_speculation): Likewise for local "target",
911 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
912 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
913 from rtx_insn ** to rtx_code_label **.
914 (reorg_emit_nops): Likewise.
915 (c6x_reorg): Likewise for local "call_labels".
916 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
918 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
919 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
920 the loops over LABEL_REFS.
921 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
923 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
924 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
925 (split_branches): Strengthen local "olabel" from rtx to
926 rtx_insn *, adding a safe_as_a cast.
927 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
929 (add_insn_after): Likewise for first two params.
930 (add_insn_before): Likewise.
931 (remove_insn): Likewise for param.
932 (emit_pattern_before_noloc): Likewise for second and third params.
933 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
934 (emit_call_insn_before_noloc): Likewise.
935 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
937 (emit_barrier_before): Likewise.
938 (emit_label_before): Strengthen "label" param from "rtx" to
939 "rtx_code_label *". Strengthen "before" param from "rtx" to
941 (emit_insn_after_1): Strengthen "after" param from "rtx" to
943 (emit_pattern_after_noloc): Likewise.
944 (emit_insn_after_noloc): Likewise.
945 (emit_jump_insn_after_noloc): Likewise.
946 (emit_call_insn_after_noloc): Likewise.
947 (emit_debug_insn_after_noloc): Likewise.
948 (emit_barrier_after): Likewise.
949 (emit_label_after): Likewise for both params.
950 (emit_pattern_after_setloc): Likewise for "after" param. Convert
951 "loc" param from "int" to "location_t".
952 (emit_insn_after_setloc): Likewise.
953 (emit_jump_insn_after_setloc): Likewise.
954 (emit_call_insn_after_setloc): Likewise.
955 (emit_debug_insn_after_setloc): Likewise.
956 (emit_pattern_before_setloc): Likewise for "before" param. Convert
957 "loc" param from "int" to "location_t".
958 (emit_pattern_before): Convert NULL_RTX to NULL.
959 (emit_insn_before_setloc): Convert "loc" param from "int" to
961 (emit_jump_insn_before_setloc): Likewise.
962 (emit_call_insn_before_setloc): Likewise.
963 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
964 rtx_insn *. Convert "loc" param from "int" to "location_t".
965 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
966 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
967 Convert 3rd param from "int" to "location_t".
968 (emit_barrier_before, emit_barrier_after, next_real_insn):
969 Strengthen param from rtx to rtx_insn *.
970 (emit_label_before): Strengthen 1st param from "rtx" to
971 "rtx_code_label *". Strengthen 2nd param from "rtx" to
973 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
974 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
975 Strengthen 2nd param from "rtx" to "rtx_insn *".
976 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
977 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
978 Likewise. Convert 3rd param from "int" to "location_t".
979 (emit_label_after): Strengthen 1st param from "rtx" to
981 (next_real_insn, remove_insn): Strengthen param from "rtx" to
983 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
984 from "rtx" to "rtx_insn *".
986 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
988 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
989 bodies streamed in with -Q.
990 * dumpfile.c (dump_files): Add lto-stream-out dump file.
991 * dumpfile.h (tree_dump_index): Add lto_stream_out.
992 * gimple-streamer-out.c: Include gimple-pretty-print.h
993 (output_bb): Dump stmts streamed.
994 * lto-section-out.c: Include print-tree.h
995 (lto_begin_section): Dump sections created.
996 (lto_output_decl_index): Dump decl encoded.
997 * lto-streamer-out.c: Include print-tree.h
998 (create_output_block): Dump output block created.
999 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
1000 (output_function): Dump function output.
1001 (output_constructor): Dump constructor streamed.
1002 (write_global_stream): Output indexes encoded.
1003 (produce_asm_for_decls): Dump streams encoded.
1004 * lto-streamer.c (streamer_dump_file): New global var.
1005 * lto-streamer.h (streamer_dump_file): Declare.
1006 * passes.c (ipa_write_summaries): Initialize streamer dump.
1007 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
1010 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
1013 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
1014 offsets for register save directives. Emit a second batch of save
1015 directives, if need be, when the function accesses prior frames.
1017 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1019 * config/arc/fpu.md (fmasf4): Force operand to register.
1020 (fnmasf4): Likewise.
1022 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1024 * config/arc/arc-protos.h (arc_pad_return): Remove.
1025 * config/arc/arc.c (machine_function): Remove force_short_suffix
1027 (arc_print_operand): Adjust printing of '&'.
1028 (arc_verify_short): Remove conditional printing of short suffix.
1029 (arc_final_prescan_insn): Remove reference to size_reason.
1030 (pad_return): New function.
1031 (arc_reorg): Call pad_return.
1032 (arc_pad_return): Remove.
1033 (arc_init_machine_status): Remove reference to force_short_suffix.
1034 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
1035 (attr length): When attribute iscompact is true force to 2
1036 regardless; in the case of maybe check if we want to force the
1037 instruction to have 4 bytes length.
1038 (nopv): Change it to generate 4 byte long nop as well.
1039 (blockage): New pattern.
1040 (simple_return): Remove call to arc_pad_return.
1041 (p_return_i): Likewise.
1043 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1045 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
1047 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1049 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
1052 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1054 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
1055 for ARC700 and ARCv2.
1057 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
1060 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
1061 operands[2] instead of operands[1].
1064 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1066 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
1067 case, check whether the outer register overlaps an unallocatable
1068 register, not just whether it fits the required class.
1070 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1072 * poly-int.h (can_div_trunc_p): Add new overload in which all values
1074 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
1075 (memrefs_conflict_p): Likewise.
1076 (init_alias_analysis): Likewise.
1077 * cfgexpand.c (expand_debug_expr): Likewise.
1078 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
1079 * cse.c (fold_rtx): Likewise.
1080 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
1081 * expr.c (emit_block_move_hints): Likewise.
1082 (clear_storage_hints, push_block, emit_push_insn): Likewise.
1083 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
1084 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
1085 (emit_group_store): Likewise.
1086 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
1087 to read the PRE/POST_MODIFY increment.
1088 * calls.c (store_one_arg): Use strip_offset.
1089 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
1091 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
1093 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
1094 (simplify_binary_operation_1): Extend CONST_INT handling to
1096 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
1097 than a HOST_WIDE_INT.
1098 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
1100 (adjust_mems, add_stores): Update accodingly.
1101 (vt_canonicalize_addr): Track polynomial offsets.
1102 (emit_note_insn_var_location): Likewise.
1103 (vt_add_function_parameter): Likewise.
1104 (vt_initialize): Likewise.
1106 2018-06-12 Jeff Law <law@redhat.com>
1108 * config.gcc (alpha*-*-freebsd*): Remove.
1109 * config/alpha/freebsd.h: Remove.
1111 2018-06-12 David Malcolm <dmalcolm@redhat.com>
1114 * spellcheck-tree.c (levenshtein_distance): Rename to...
1115 (get_edit_distance): ...this, and update for underlying renaming.
1116 * spellcheck-tree.h (levenshtein_distance): Rename to...
1117 (get_edit_distance): ...this.
1118 * spellcheck.c (levenshtein_distance): Rename to...
1119 (get_edit_distance): ...this. Convert from Levenshtein distance
1120 to Damerau-Levenshtein distance by supporting transpositions of
1121 adjacent characters. Rename "v1" to "v_next" and "v0" to
1123 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
1124 (selftest::test_edit_distance_unit_test_oneway): ...this, and
1125 update for underlying renaming.
1126 (selftest::levenshtein_distance_unit_test): Rename to...
1127 (selftest::test_get_edit_distance_unit): ...this, and update for
1128 underlying renaming.
1129 (selftest::test_find_closest_string): Add example from PR 69968
1130 where transposition helps
1131 (selftest::test_metric_conditions): Update for renaming.
1132 (selftest::test_metric_conditions): Likewise.
1133 (selftest::spellcheck_c_tests): Likewise.
1134 * spellcheck.h (levenshtein_distance): Rename both overloads to...
1135 (get_edit_distance): ...this.
1136 (best_match::consider): Update for renaming.
1138 2018-06-12 Martin Sebor <msebor@redhat.com>
1140 PR tree-optimization/85259
1141 * builtins.c (compute_objsize): Handle constant offsets.
1142 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
1143 true iff a warning has been issued.
1144 * gimple.h (gimple_nonartificial_location): New function.
1145 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
1146 gimple_nonartificial_location and handle -Wno-system-headers.
1147 (handle_builtin_stxncpy): Same.
1149 2018-06-12 Martin Sebor <msebor@redhat.com>
1152 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
1154 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
1156 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
1157 BUILTIN_VEC_XST entries for pointer to double and long long.
1159 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
1162 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1164 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1167 2018-06-12 Martin Liska <mliska@suse.cz>
1169 * doc/options.texi: Document IntegerRange.
1171 2018-06-12 Martin Liska <mliska@suse.cz>
1173 * config/i386/i386.opt: Make MPX-related options as Deprecated.
1174 * opt-functions.awk: Handle Deprecated flag.
1175 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
1177 (read_cmdline_option): Report warning for a deprecated option.
1178 * opts.h (struct cl_option): Add new field cl_deprecated.
1179 (CL_ERR_DEPRECATED): New.
1181 2018-06-12 Martin Liska <mliska@suse.cz>
1183 * doc/options.texi: Document Deprecated option flag.
1185 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1187 * config/arc/arc-arch.h (arc_extras): New enum.
1188 (arc_cpu_t):Add field extra.
1189 (arc_cpu_types): Consider the extras.
1190 * config/arc/arc-cpus.def: Add extras info.
1191 * config/arc/arc-opts.h (processor_type): Consider extra field.
1192 * config/arc/arc.c (arc_override_options): Handle extra field.
1194 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1196 * config/arc/arc-arch.h: Update ARC_OPTX macro.
1197 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
1199 * config/arc/arc.c (arc_init): Update pic warning.
1200 (irq_range): Update irq range parsing warnings.
1201 (arc_override_options): Update various warning messages.
1202 (arc_handle_aux_attribute): Likewise.
1204 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
1206 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
1208 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1210 * doc/sourcebuild.texi: Document usage of line number 0 in verify
1211 compiler messages directives.
1213 2018-06-12 Matthew Fortune <mfortune@gmail.com>
1215 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
1216 * config/mips/mips-tables.opt: Regenerate.
1217 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
1219 * doc/invoke.texi: Document -march=i6500.
1221 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
1223 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
1224 (i6400_gpmul): Add cpu_unit.
1225 (i6400_gpdiv): Likewise.
1226 (i6400_msa_add_d): Update reservations.
1227 (i6400_msa_int_add) Likewise.
1228 (i6400_msa_short_logic3) Likewise.
1229 (i6400_msa_short_logic2) Likewise.
1230 (i6400_msa_short_logic) Likewise.
1231 (i6400_msa_move) Likewise.
1232 (i6400_msa_cmp) Likewise.
1233 (i6400_msa_short_float2) Likewise.
1234 (i6400_msa_div_d) Likewise.
1235 (i6400_msa_long_logic1) Likewise.
1236 (i6400_msa_long_logic2) Likewise.
1237 (i6400_msa_mult) Likewise.
1238 (i6400_msa_long_float2) Likewise.
1239 (i6400_msa_long_float4) Likewise.
1240 (i6400_msa_long_float5) Likewise.
1241 (i6400_msa_long_float8) Likewise.
1242 (i6400_fpu_fadd): Include frint type.
1243 (i6400_fpu_store): New define_insn_reservation.
1244 (i6400_fpu_load): Likewise.
1245 (i6400_fpu_move): Likewise.
1246 (i6400_fpu_fcmp): Likewise.
1247 (i6400_fpu_fmadd): Likewise.
1248 (i6400_int_mult): Include imul3nc type and update reservation.
1249 (i6400_int_div): Include idiv3 type and update reservation.
1250 (i6400_int_load): Update to check type not move_type.
1251 (i6400_int_store): Likewise.
1252 (i6400_int_prefetch): Set zero latency.
1254 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
1256 * gcc.c: Document new %@{...} sequence.
1257 (LINK_COMMAND_SPEC): Use it for the -L switches.
1258 (cpp_unique_options): Use it for the -I switches.
1259 (at_file_argbuf): New global variable.
1260 (in_at_file): Likewise.
1261 (alloc_args): Create at_file_argbuf.
1262 (clear_args): Truncate at_file_argbuf.
1263 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
1264 (open_at_file): New function.
1265 (close_at_file): Likewise.
1266 (create_at_file): Delete.
1267 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
1270 (validate_switches_from_spec): Deal with %@{...} sequence.
1271 (validate_switches): Likewise.
1272 (driver::finalize): Call clear_args.
1274 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
1276 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
1278 2018-06-11 Martin Sebor <msebor@redhat.com>
1280 * doc/invoke.texi (-Wall): List -Wc++17-compat.
1281 (Wno-class-memaccess): Add @opindex.
1282 (Wno-templates, Wno-multiple-inheritance): Same.
1283 (Wno-virtual-inheritance, Wno-namespaces): Same.
1284 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
1285 (Wno-format-overflow, Wno-format-truncation): Same.
1286 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
1287 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
1288 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
1289 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
1290 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
1291 (Wno-misspelled-isr): Same.
1293 2018-06-11 Martin Sebor <msebor@redhat.com>
1295 * PR tree-optimization/86083
1296 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
1298 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
1300 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
1302 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
1305 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
1306 on the correct operand.
1307 (*movdi_internal64): Ditto.
1309 2018-06-11 Martin Liska <mliska@suse.cz>
1311 PR tree-optimization/86089
1312 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
1314 2018-06-11 Julia Koval <julia.koval@intel.com>
1316 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
1317 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
1318 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
1320 2018-06-11 Olivier Hainque <hainque@adacore.com>
1322 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
1323 for Ada with strict dwarf2.
1325 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
1328 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
1331 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
1333 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
1335 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
1337 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
1340 2018-06-08 Martin Liska <mliska@suse.cz>
1342 * tree-cfg.h (debug_function): Fix argument type to match
1345 2018-06-08 Martin Liska <mliska@suse.cz>
1347 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
1348 Remove usage of MPX-related (and removed) fields.
1349 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
1351 2018-06-08 David Malcolm <dmalcolm@redhat.com>
1353 * cfg.c (debug): Use TDF_NONE rather than 0.
1354 * cfghooks.c (debug): Likewise.
1355 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
1356 (struct dump_option_value_info): Convert to...
1357 (struct kv_pair): ...this template type.
1358 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
1360 (optinfo_verbosity_options): Likewise.
1361 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
1363 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
1364 than int for "optgroup_flags" param.
1365 (dump_generic_expr_loc): Use dump_flags_t rather than int for
1367 (dump_dec): Likewise.
1368 (dump_finish): Use TDF_NONE rather than 0.
1369 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
1370 rather than int for "optgroup_flags" param. Use TDF_NONE rather
1371 than 0. Update for change to option_ptr.
1372 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
1373 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
1374 0. Update for changes to optinfo_verbosity_options and
1376 (opt_info_switch_p): Convert optgroup_flags from int to
1378 (dump_basic_block): Use dump_flags_t rather than int
1379 for "dump_kind" param.
1380 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
1381 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
1382 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
1383 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
1384 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
1385 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
1386 TDF_NONE): Convert from macros to...
1387 (enum dump_flag): ...this new enum.
1388 (dump_flags_t): Update to use enum.
1389 (operator|, operator&, operator~, operator|=, operator&=):
1390 Implement for dump_flags_t.
1391 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
1392 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
1393 Convert from macros to...
1394 (enum optgroup_flag): ...this new enum.
1395 (optgroup_flags_t): New typedef.
1396 (operator|, operator|=): Implement for optgroup_flags_t.
1397 (struct dump_file_info): Convert field "alt_flags" to
1398 dump_flags_t. Convert field "optgroup_flags" to
1400 (dump_basic_block): Use dump_flags_t rather than int for param.
1401 (dump_generic_expr_loc): Likewise.
1402 (dump_dec): Likewise.
1403 (dump_register): Convert param "optgroup_flags" to
1405 (opt_info_enable_passes): Likewise.
1406 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
1408 * gimple-pretty-print.c (debug): Likewise.
1409 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
1410 (merged_store_group::apply_stores): Likewise.
1411 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
1412 * gimple.c (verify_gimple_pp): Likewise.
1413 * graphite-poly.c (print_pbb_body): Likewise.
1414 * passes.c (pass_manager::register_one_dump_file): Convert
1415 local "optgroup_flags" to optgroup_flags_t.
1416 * print-tree.c (print_node): Use TDF_NONE rather than 0.
1418 (debug_body): Likewise.
1419 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
1420 to optgroup_flags_t.
1421 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
1423 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
1424 (convert_mult_to_fma): Likewise.
1425 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
1426 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
1427 * tree-vect-data-refs.c (dump_lower_bound): Convert param
1428 "dump_kind" to dump_flags_t.
1430 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1432 * config/rs6000/rs6000.c (min, max): Delete.
1434 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1436 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
1437 -mabi=spe and -mabi=no-spe.
1439 2018-06-08 Martin Liska <mliska@suse.cz>
1441 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
1442 where we expect an existing summary.
1444 2018-06-08 Martin Liska <mliska@suse.cz>
1446 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
1447 * ipa-inline.h (estimate_edge_growth): Likewise.
1449 2018-06-08 Martin Liska <mliska@suse.cz>
1451 * cgraph.c (function_version_hasher::hash): Use
1452 cgraph_node::get_uid ().
1453 (function_version_hasher::equal):
1454 * cgraph.h (cgraph_node::get_uid): New method.
1455 * ipa-inline.c (update_caller_keys): Use
1456 cgraph_node::get_uid ().
1457 (update_callee_keys): Likewise.
1458 * ipa-utils.c (searchc): Likewise.
1459 (ipa_reduced_postorder): Likewise.
1460 * lto-cgraph.c (input_node): Likewise.
1461 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
1462 * symbol-summary.h (symtab_insertion): Likewise.
1463 (symtab_removal): Likewise.
1464 (symtab_duplication): Likewise.
1465 * tree-pretty-print.c (dump_function_header): Likewise.
1466 * tree-sra.c (convert_callers_for_node): Likewise.
1468 2018-06-08 Martin Liska <mliska@suse.cz>
1470 * cgraph.c (symbol_table::create_edge): Always assign a new
1472 (symbol_table::free_edge): Do not recycle numbers.
1473 * cgraph.h (cgraph_edge::get): New method.
1474 * symbol-summary.h (symtab_removal): Use it.
1475 (symtab_duplication): Likewise.
1476 (call_summary::hashable_uid): Remove.
1478 2018-06-08 Martin Liska <mliska@suse.cz>
1480 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
1481 (initialize_growth_caches): Remove.
1482 (free_growth_caches): Likewise.
1483 (do_estimate_edge_time): Use edge_growth_cache.
1484 (do_estimate_edge_size): Likewise.
1485 (do_estimate_edge_hints): Likewise.
1486 * ipa-inline.c (reset_edge_caches): Likewise.
1487 (recursive_inlining): Likewise.
1488 (inline_small_functions): Likewise.
1489 * ipa-inline.h (initialize_growth_caches): Remove.
1490 (estimate_edge_size): Likewise.
1491 (estimate_edge_time): Likewise.
1492 (estimate_edge_hints): Likewise.
1493 (reset_edge_growth_cache): Likewise.
1494 * symbol-summary.h (call_summary::remove): New method.
1496 2018-06-08 Martin Liska <mliska@suse.cz>
1498 * ipa-cp.c (class edge_clone_summary): New summary.
1499 (grow_edge_clone_vectors): Remove.
1500 (ipcp_edge_duplication_hook): Remove.
1501 (class edge_clone_summary_t): New call_summary class.
1502 (ipcp_edge_removal_hook): Remove.
1503 (edge_clone_summary_t::duplicate): New function.
1504 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
1505 (create_specialized_node): Likewise.
1506 (ipcp_driver): Initialize edge_clone_summaries and do not
1509 2018-06-08 Martin Liska <mliska@suse.cz>
1511 * symbol-summary.h (get): New function.
1512 (call_summary::m_initialize_when_cloning): New class member.
1514 2018-06-08 Martin Liska <mliska@suse.cz>
1516 * cgraph.c (cgraph_node::remove): Do not recycle uid.
1517 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
1518 (symbol_table::allocate_cgraph_symbol): Do not set uid.
1519 * passes.c (uid_hash_t): Record removed_nodes by their uids.
1520 (remove_cgraph_node_from_order): Use the removed_nodes set.
1521 (do_per_function_toporder): Likwise.
1522 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
1523 instead of summary_uid.
1524 (symtab_removal): Likewise.
1525 (symtab_duplication): Likewise.
1527 2018-06-08 Martin Liska <mliska@suse.cz>
1529 * ipa-cp.c (ipcp_store_bits_results): Use
1530 ipcp_transformation_sum.
1531 (ipcp_store_vr_results): Likewise.
1532 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
1534 (ipcp_transformation_initialize): ... this.
1535 (ipa_set_node_agg_value_chain):
1536 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
1537 (write_ipcp_transformation_info): Likewise.
1538 (read_ipcp_transformation_info): Likewise.
1539 (ipcp_update_bits): Likewise.
1540 (ipcp_update_vr): Likewise.
1541 (ipcp_transform_function): Likewise.
1542 * ipa-prop.h: Rename ipcp_transformation_summary to
1543 ipcp_transformation.
1544 (class ipcp_transformation_t): New function summary.
1545 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
1546 (ipa_get_agg_replacements_for_node): Likewise.
1548 2018-06-08 Martin Liska <mliska@suse.cz>
1550 * ipa-pure-const.c (struct funct_state_d): Do it class instead
1552 (class funct_state_summary_t): New function_summary class.
1553 (has_function_state): Remove.
1554 (get_function_state): Likewise.
1555 (set_function_state): Likewise.
1556 (add_new_function): Likewise.
1557 (funct_state_summary_t::insert): New function.
1558 (duplicate_node_data): Remove.
1559 (remove_node_data): Remove.
1560 (funct_state_summary_t::duplicate): New function.
1561 (register_hooks): Create new funct_state_summaries.
1562 (pure_const_generate_summary): Use it.
1563 (pure_const_write_summary): Likewise.
1564 (pure_const_read_summary): Likewise.
1565 (propagate_pure_const): Likewise.
1566 (propagate_nothrow): Likewise.
1567 (dump_malloc_lattice): Likewise.
1568 (propagate_malloc): Likewise.
1569 (execute): Do not register hooks, just remove summary
1571 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
1574 2018-06-08 Martin Liska <mliska@suse.cz>
1576 * ipa-reference.c (remove_node_data): Remove.
1577 (duplicate_node_data): Likewise.
1578 (class ipa_ref_var_info_summary_t): New class.
1579 (class ipa_ref_opt_summary_t): Likewise.
1580 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
1581 (get_reference_optimization_summary): Use
1582 ipa_ref_opt_sum_summaries.
1583 (set_reference_vars_info): Remove.
1584 (set_reference_optimization_summary): Likewise.
1585 (ipa_init): Create summaries.
1586 (init_function_info): Use function summary.
1587 (ipa_ref_opt_summary_t::duplicate): New function.
1588 (ipa_ref_opt_summary_t::remove): New function.
1589 (get_read_write_all_from_node): Fix GNU coding style.
1590 (propagate): Use function summary.
1591 (write_node_summary_p): Fix GNU coding style.
1592 (stream_out_bitmap): Likewise.
1593 (ipa_reference_read_optimization_summary): Use function summary.
1594 (ipa_reference_c_finalize): Do not release hooks.
1596 2018-06-08 Martin Liska <mliska@suse.cz>
1598 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
1599 (analyze_function_body): Extract multiple calls of get_create.
1600 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1601 * ipa-inline.c (recursive_inlining): Use ::get method.
1602 * ipa-inline.h (estimate_edge_growth): Likewise.
1604 2018-06-08 Martin Liska <mliska@suse.cz>
1606 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
1608 (hsa_function_summary::hsa_function_summary): Use the new enum
1610 (hsa_gpu_implementation_p): Use hsa_summaries::get.
1611 * hsa-gen.c (hsa_get_host_function): Likewise.
1612 (get_brig_function_name): Likewise.
1613 * ipa-hsa.c (process_hsa_functions): Likewise.
1614 (ipa_hsa_write_summary): Likewise.
1615 * symbol-summary.h (symtab_duplication): Use ::get function/
1616 (get): New function.
1618 2018-06-08 Martin Liska <mliska@suse.cz>
1620 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
1622 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
1623 (hsa_register_kernel): Likewise.
1624 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
1625 * hsa-gen.c (hsa_get_host_function): Likewise.
1626 (get_brig_function_name): Likewise.
1627 (generate_hsa): Likewise.
1628 (pass_gen_hsail::execute): Likewise.
1629 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
1630 (devirtualization_time_bonus): Likewise.
1631 (ipcp_propagate_stage): Likewise.
1632 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1633 (edge_set_predicate): Likewise.
1634 (evaluate_conditions_for_known_args): Likewise.
1635 (evaluate_properties_for_edge): Likewise.
1636 (ipa_fn_summary::reset): Likewise.
1637 (ipa_fn_summary_t::duplicate): Likewise.
1638 (dump_ipa_call_summary): Likewise.
1639 (ipa_dump_fn_summary): Likewise.
1640 (analyze_function_body): Likewise.
1641 (compute_fn_summary): Likewise.
1642 (estimate_edge_devirt_benefit): Likewise.
1643 (estimate_edge_size_and_time): Likewise.
1644 (estimate_calls_size_and_time): Likewise.
1645 (estimate_node_size_and_time): Likewise.
1646 (inline_update_callee_summaries): Likewise.
1647 (remap_edge_change_prob): Likewise.
1648 (remap_edge_summaries): Likewise.
1649 (ipa_merge_fn_summary_after_inlining): Likewise.
1650 (ipa_update_overall_fn_summary): Likewise.
1651 (read_ipa_call_summary): Likewise.
1652 (inline_read_section): Likewise.
1653 (write_ipa_call_summary): Likewise.
1654 (ipa_fn_summary_write): Likewise.
1655 (ipa_free_fn_summary): Likewise.
1656 * ipa-hsa.c (process_hsa_functions): Likewise.
1657 (ipa_hsa_write_summary): Likewise.
1658 (ipa_hsa_read_section): Likewise.
1659 * ipa-icf.c (sem_function::merge): Likewise.
1660 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1661 (do_estimate_edge_time): Likewise.
1662 (estimate_size_after_inlining): Likewise.
1663 (estimate_growth): Likewise.
1664 (growth_likely_positive): Likewise.
1665 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1666 (inline_call): Likewise.
1667 * ipa-inline.c (caller_growth_limits): Likewise.
1668 (can_inline_edge_p): Likewise.
1669 (can_inline_edge_by_limits_p): Likewise.
1670 (compute_uninlined_call_time): Likewise.
1671 (compute_inlined_call_time): Likewise.
1672 (want_inline_small_function_p): Likewise.
1673 (edge_badness): Likewise.
1674 (update_caller_keys): Likewise.
1675 (update_callee_keys): Likewise.
1676 (recursive_inlining): Likewise.
1677 (inline_small_functions): Likewise.
1678 (inline_to_all_callers_1): Likewise.
1679 (dump_overall_stats): Likewise.
1680 (early_inline_small_functions): Likewise.
1681 (early_inliner): Likewise.
1682 * ipa-inline.h (estimate_edge_growth): Likewise.
1683 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1684 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1685 * ipa-prop.h (IPA_NODE_REF): Likewise.
1686 (IPA_EDGE_REF): Likewise.
1687 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1688 (propagate_malloc): Likewise.
1689 * ipa-split.c (execute_split_functions): Likewise.
1690 * symbol-summary.h: Rename get to get_create.
1692 (get_create): Likewise.
1693 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1695 2018-06-08 Martin Liska <mliska@suse.cz>
1697 * symbol-summary.h (release): Move definition out of class
1699 (symtab_removal): Likewise.
1700 (symtab_duplication): Likewise.
1702 2018-06-08 Martin Liska <mliska@suse.cz>
1704 * symbol-summary.h (function_summary): Move constructor
1705 implementation out of class declaration.
1706 (release): Likewise.
1707 (symtab_insertion): Likewise.
1708 (symtab_removal): Likewise.
1709 (symtab_duplication): Likewise.
1712 2018-06-08 Martin Liska <mliska@suse.cz>
1714 * Makefile.in: Remove support for MPX (macros, related functions,
1715 fields in cgraph_node, ...).
1716 * builtin-types.def (BT_BND): Likewise.
1717 (BT_FN_BND_CONST_PTR): Likewise.
1718 (BT_FN_CONST_PTR_BND): Likewise.
1719 (BT_FN_VOID_PTR_BND): Likewise.
1720 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
1721 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
1722 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
1723 (expand_builtin_mempcpy_with_bounds): Likewise.
1724 (expand_builtin_memset_with_bounds): Likewise.
1725 (expand_builtin_memset_args): Likewise.
1726 (std_expand_builtin_va_start): Likewise.
1727 (expand_builtin): Likewise.
1728 (expand_builtin_with_bounds): Likewise.
1729 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
1730 (DEF_LIB_BUILTIN_CHKP): Likewise.
1731 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
1732 (DEF_CHKP_BUILTIN): Likewise.
1733 (BUILT_IN_MEMCPY): Likewise.
1734 (BUILT_IN_MEMMOVE): Likewise.
1735 (BUILT_IN_MEMPCPY): Likewise.
1736 (BUILT_IN_MEMSET): Likewise.
1737 (BUILT_IN_STPCPY): Likewise.
1738 (BUILT_IN_STRCAT): Likewise.
1739 (BUILT_IN_STRCHR): Likewise.
1740 (BUILT_IN_STRCPY): Likewise.
1741 (BUILT_IN_STRLEN): Likewise.
1742 (BUILT_IN_MEMCPY_CHK): Likewise.
1743 (BUILT_IN_MEMMOVE_CHK): Likewise.
1744 (BUILT_IN_MEMPCPY_CHK): Likewise.
1745 (BUILT_IN_MEMSET_CHK): Likewise.
1746 (BUILT_IN_STPCPY_CHK): Likewise.
1747 (BUILT_IN_STRCAT_CHK): Likewise.
1748 (BUILT_IN_STRCPY_CHK): Likewise.
1749 * calls.c (store_bounds): Likewise.
1750 (emit_call_1): Likewise.
1751 (special_function_p): Likewise.
1752 (maybe_warn_nonstring_arg): Likewise.
1753 (initialize_argument_information): Likewise.
1754 (finalize_must_preallocate): Likewise.
1755 (compute_argument_addresses): Likewise.
1756 (expand_call): Likewise.
1757 * cfgexpand.c (expand_call_stmt): Likewise.
1758 (expand_return): Likewise.
1759 (expand_gimple_stmt_1): Likewise.
1760 (pass_expand::execute): Likewise.
1761 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
1762 (cgraph_node::remove): Likewise.
1763 (cgraph_node::dump): Likewise.
1764 (cgraph_node::verify_node): Likewise.
1765 * cgraph.h (chkp_function_instrumented_p): Likewise.
1766 (symtab_node::get_alias_target): Likewise.
1767 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
1768 (cgraph_local_p): Likewise.
1769 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
1770 (cgraph_edge::rebuild_references): Likewise.
1771 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
1772 (walk_polymorphic_call_targets): Likewise.
1773 (cgraph_node::expand_thunk): Likewise.
1774 (symbol_table::output_weakrefs): Likewise.
1775 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
1776 (ix86_handle_option): Likewise.
1777 * config/i386/constraints.md: Likewise.
1778 * config/i386/i386-builtin-types.def (BND): Likewise.
1782 * config/i386/i386-builtin.def (BDESC_END): Likewise.
1783 (BDESC_FIRST): Likewise.
1785 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
1786 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
1787 * config/i386/i386.c (enum reg_class): Likewise.
1788 (ix86_target_string): Likewise.
1789 (ix86_option_override_internal): Likewise.
1790 (ix86_conditional_register_usage): Likewise.
1791 (ix86_valid_target_attribute_inner_p): Likewise.
1792 (ix86_set_indirect_branch_type): Likewise.
1793 (ix86_set_current_function): Likewise.
1794 (ix86_function_arg_regno_p): Likewise.
1795 (init_cumulative_args): Likewise.
1796 (ix86_function_arg_advance): Likewise.
1797 (ix86_function_arg): Likewise.
1798 (ix86_pass_by_reference): Likewise.
1799 (ix86_function_value_regno_p): Likewise.
1800 (ix86_function_value_1): Likewise.
1801 (ix86_function_value_bounds): Likewise.
1802 (ix86_return_in_memory): Likewise.
1803 (ix86_setup_incoming_vararg_bounds): Likewise.
1804 (ix86_va_start): Likewise.
1805 (indirect_thunk_need_prefix): Likewise.
1806 (print_reg): Likewise.
1807 (ix86_print_operand): Likewise.
1808 (ix86_expand_call): Likewise.
1809 (ix86_output_function_return): Likewise.
1810 (reg_encoded_number): Likewise.
1811 (BDESC_VERIFYS): Likewise.
1812 (ix86_init_mpx_builtins): Likewise.
1813 (ix86_init_builtins): Likewise.
1814 (ix86_emit_cmove): Likewise.
1815 (ix86_emit_move_max): Likewise.
1816 (ix86_expand_builtin): Likewise.
1817 (ix86_builtin_mpx_function): Likewise.
1818 (ix86_get_arg_address_for_bt): Likewise.
1819 (ix86_load_bounds): Likewise.
1820 (ix86_store_bounds): Likewise.
1821 (ix86_load_returned_bounds): Likewise.
1822 (ix86_store_returned_bounds): Likewise.
1823 (ix86_class_likely_spilled_p): Likewise.
1824 (ix86_hard_regno_mode_ok): Likewise.
1825 (x86_order_regs_for_local_alloc): Likewise.
1826 (ix86_mitigate_rop): Likewise.
1827 (ix86_bnd_prefixed_insn_p): Likewise.
1828 (ix86_mpx_bound_mode): Likewise.
1829 (ix86_make_bounds_constant): Likewise.
1830 (ix86_initialize_bounds): Likewise.
1831 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
1832 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
1833 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
1834 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
1835 (TARGET_CHKP_BOUND_MODE): Likewise.
1836 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
1837 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
1838 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
1839 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
1840 * config/i386/i386.h (TARGET_MPX): Likewise.
1841 (TARGET_MPX_P): Likewise.
1842 (VALID_BND_REG_MODE): Likewise.
1843 (FIRST_BND_REG): Likewise.
1844 (LAST_BND_REG): Likewise.
1845 (enum reg_class): Likewise.
1846 (BND_REG_P): Likewise.
1847 (BND_REGNO_P): Likewise.
1848 (BNDmode): Likewise.
1849 (ADJUST_INSN_LENGTH): Likewise.
1850 * config/i386/i386.md: Likewise.
1851 * config/i386/i386.opt: Likewise.
1852 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
1853 (defined): Likewise.
1854 (LINK_MPX): Likewise.
1855 (MPX_SPEC): Likewise.
1856 (LIBMPX_SPEC): Likewise.
1857 (LIBMPXWRAPPERS_SPEC): Likewise.
1858 (CHKP_SPEC): Likewise.
1859 * config/i386/predicates.md: Likewise.
1860 * dbxout.c (dbxout_type): Likewise.
1861 * doc/extend.texi: Likewise.
1862 * doc/invoke.texi: Likewise.
1863 * doc/md.texi: Likewise.
1864 * doc/tm.texi: Likewise.
1865 * doc/tm.texi.in: Likewise.
1866 * dwarf2out.c (is_base_type): Likewise.
1867 (gen_formal_types_die): Likewise.
1868 (gen_subprogram_die): Likewise.
1869 (gen_type_die_with_usage): Likewise.
1870 (gen_decl_die): Likewise.
1871 (dwarf2out_late_global_decl): Likewise.
1872 * expr.c (expand_assignment): Likewise.
1873 (emit_storent_insn): Likewise.
1874 (store_expr_with_bounds): Likewise.
1875 (store_expr): Likewise.
1876 (expand_expr_real_1): Likewise.
1877 * expr.h (store_expr_with_bounds): Likewise.
1878 * function.c (use_register_for_decl): Likewise.
1879 (struct bounds_parm_data): Likewise.
1880 (assign_parms_augmented_arg_list): Likewise.
1881 (assign_parm_find_entry_rtl): Likewise.
1882 (assign_parm_is_stack_parm): Likewise.
1883 (assign_parm_load_bounds): Likewise.
1884 (assign_bounds): Likewise.
1885 (assign_parms): Likewise.
1886 (expand_function_start): Likewise.
1887 * gcc.c (CHKP_SPEC): Likewise.
1888 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1889 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
1890 (wrestrict_dom_walker::check_call): Likewise.
1891 * gimple.c (gimple_build_call_from_tree): Likewise.
1892 * gimple.h (enum gf_mask): Likewise.
1893 (gimple_call_with_bounds_p): Likewise.
1894 (gimple_call_set_with_bounds): Likewise.
1895 * gimplify.c (gimplify_init_constructor): Likewise.
1896 * ipa-cp.c (initialize_node_lattices): Likewise.
1897 (propagate_constants_across_call): Likewise.
1898 (find_more_scalar_values_for_callers_subset): Likewise.
1899 * ipa-hsa.c (process_hsa_functions): Likewise.
1900 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
1901 * ipa-icf.c (sem_function::merge): Likewise.
1902 * ipa-inline.c (early_inliner): Likewise.
1903 * ipa-pure-const.c (warn_function_noreturn): Likewise.
1904 (warn_function_cold): Likewise.
1905 (propagate_pure_const): Likewise.
1906 * ipa-ref.h (enum GTY): Likewise.
1907 * ipa-split.c (find_retbnd): Likewise.
1908 (consider_split): Likewise.
1909 (split_function): Likewise.
1910 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1911 * ipa.c (walk_polymorphic_call_targets): Likewise.
1912 (symbol_table::remove_unreachable_nodes): Likewise.
1913 (process_references): Likewise.
1914 (cgraph_build_static_cdtor_1): Likewise.
1915 * lto-cgraph.c (lto_output_node): Likewise.
1916 (output_refs): Likewise.
1917 (compute_ltrans_boundary): Likewise.
1918 (input_overwrite_node): Likewise.
1919 (input_node): Likewise.
1920 (input_cgraph_1): Likewise.
1921 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
1922 * passes.c (pass_manager::execute_early_local_passes): Likewise.
1923 (class pass_chkp_instrumentation_passes): Likewise.
1924 (make_pass_chkp_instrumentation_passes): Likewise.
1925 * passes.def: Likewise.
1926 * rtl.h (struct GTY): Likewise.
1927 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
1928 * stor-layout.c (layout_type): Likewise.
1929 * symtab.c: Likewise.
1930 * target.def: Likewise.
1931 * targhooks.c (default_chkp_bound_type): Likewise.
1932 (default_chkp_bound_mode): Likewise.
1933 (default_builtin_chkp_function): Likewise.
1934 (default_chkp_function_value_bounds): Likewise.
1935 (default_chkp_make_bounds_constant): Likewise.
1936 (default_chkp_initialize_bounds): Likewise.
1937 * targhooks.h (default_chkp_bound_type): Likewise.
1938 (default_chkp_bound_mode): Likewise.
1939 (default_builtin_chkp_function): Likewise.
1940 (default_chkp_function_value_bounds): Likewise.
1941 (default_chkp_make_bounds_constant): Likewise.
1942 (default_chkp_initialize_bounds): Likewise.
1943 * toplev.c (compile_file): Likewise.
1944 (process_options): Likewise.
1945 * tree-core.h (DEF_BUILTIN): Likewise.
1946 (DEF_BUILTIN_CHKP): Likewise.
1947 * tree-inline.c (declare_return_variable): Likewise.
1948 (remap_gimple_stmt): Likewise.
1949 (copy_bb): Likewise.
1950 (initialize_inlined_parameters): Likewise.
1951 (expand_call_inline): Likewise.
1952 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
1953 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
1954 (make_pass_ipa_chkp_produce_thunks): Likewise.
1955 (make_pass_chkp): Likewise.
1956 (make_pass_chkp_opt): Likewise.
1957 (make_pass_chkp_instrumentation_passes): Likewise.
1958 * tree-pretty-print.c (dump_generic_node): Likewise.
1959 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
1960 * tree-ssa-dce.c (propagate_necessity): Likewise.
1961 (eliminate_unnecessary_stmts): Likewise.
1962 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
1963 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1964 * tree-ssa-sccvn.h: Likewise.
1965 * tree-ssa-strlen.c (get_string_length): Likewise.
1966 (valid_builtin_call): Likewise.
1967 (adjust_last_stmt): Likewise.
1968 (handle_builtin_strchr): Likewise.
1969 (handle_builtin_strcpy): Likewise.
1970 (handle_builtin_stxncpy): Likewise.
1971 (handle_builtin_memcpy): Likewise.
1972 (handle_builtin_strcat): Likewise.
1973 (strlen_check_and_optimize_stmt): Likewise.
1974 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
1975 * tree-streamer-in.c: Likewise.
1976 * tree-streamer.c (record_common_node): Likewise.
1977 * tree.c (tree_code_size): Likewise.
1978 (wide_int_to_tree_1): Likewise.
1979 (type_contains_placeholder_1): Likewise.
1980 (build_common_tree_nodes): Likewise.
1981 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
1982 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
1983 (POINTER_BOUNDS_P): Likewise.
1984 (BOUNDED_TYPE_P): Likewise.
1985 (BOUNDED_P): Likewise.
1986 (CALL_WITH_BOUNDS_P): Likewise.
1987 (pointer_bounds_type_node): Likewise.
1988 * value-prof.c (gimple_ic): Likewise.
1989 * var-tracking.c (vt_add_function_parameters): Likewise.
1990 * varasm.c (make_decl_rtl): Likewise.
1991 (assemble_start_function): Likewise.
1992 (output_constant): Likewise.
1993 (maybe_assemble_visibility): Likewise.
1994 * varpool.c (ctor_for_folding): Likewise.
1995 * chkp-builtins.def: Remove.
1996 * ipa-chkp.c: Remove.
1997 * ipa-chkp.h: Remove.
1998 * rtl-chkp.c: Remove.
1999 * rtl-chkp.h: Remove.
2000 * tree-chkp-opt.c: Remove.
2001 * tree-chkp.c: Remove.
2002 * tree-chkp.h: Remove.
2004 2018-06-07 Carl Love <cel@us.ibm.com>
2006 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
2007 vextract_fp_from_shortl): Add BE support.
2009 2018-06-07 Paul Koning <ni1d@arrl.net>
2011 * compare-elim.c (try_merge_compare): Don't merge compare if
2012 address contains a side effect.
2013 (try_eliminate_compare): Likewise.
2015 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
2017 * config.gcc: Support "tremont".
2018 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
2019 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2021 * config/i386/i386.c (m_TREMONT): Define.
2022 (processor_target_table): Add "tremont".
2023 (PTA_TREMONT): Define.
2024 (ix86_lea_outperforms): Add TARGET_TREMONT.
2025 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
2026 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
2027 and M_INTEL_GOLDMONT_PLUS.
2028 (fold_builtin_cpu): Add "tremont".
2029 (ix86_add_stmt_cost): Add TARGET_TREMONT.
2030 (ix86_option_override_internal): Add "tremont".
2031 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
2032 (processor_type): Add PROCESSOR_TREMONT.
2033 * config/i386/x86-tune.def: Add m_TREMONT.
2034 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
2036 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2038 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
2039 symbol defined for msp430i* devices to be lower case.
2041 gcc/testsuite/gcc.target/msp430/
2042 * msp430i-device-symbol.c: New test.
2043 * msp430f-device-symbol.c: New test.
2044 * msp430.h: New test header file.
2046 2018-06-07 Richard Biener <rguenther@suse.de>
2048 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
2049 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
2050 Properly wrap signed arithmetic if overflow wraps.
2052 2018-06-07 Jakub Jelinek <jakub@redhat.com>
2054 PR tree-optimization/69615
2055 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
2056 of a cast from a same precision integral SSA_NAME in a bb dominated
2057 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
2058 cast to utype if rhs2 has already a compatible type.
2060 2018-06-07 Richard Biener <rguenther@suse.de>
2062 PR tree-optimization/85935
2063 * graphite-scop-detection.c (find_params_in_bb): Analyze
2064 condition operands with respect to the correct loop. Assert
2065 the analysis doesn't fail.
2067 2018-06-04 Carl Love <cel@us.ibm.com>
2069 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
2070 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
2071 as it is slightly cheaper.
2072 (first_match_or_eos_index_<mode>):
2073 Calculate index using natural element order.
2074 (first_match_index_<mode>):
2075 Calculate index using natural element order.
2076 (first_match_or_eos_index_<mode>):
2077 Calculate index using natural order.
2078 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
2079 for BE and LE modes.
2080 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
2081 P9V_BUILTIN_VCLZLSBB_V16QI.
2082 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
2085 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2087 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
2088 indentation and line wrap for many prototypes. Add missing
2089 @smallexample directives around block of prototypes for vec_xl and
2092 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
2094 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
2095 track if we pass or return IEEE 128-bit floating point.
2096 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
2097 C++ mangling that is compatible with GCC 8.1.
2098 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
2099 (init_cumulative_args): Note if we pass or return IEEE 128-bit
2100 floating point types.
2101 (rs6000_function_arg_advance_1): Likewise.
2102 (rs6000_mangle_type): Optionally generate mangled names that match
2103 what GCC 8.1 generated for IEEE 128-bit floating point types.
2104 (rs6000_globalize_decl_name): If we have an external function that
2105 passes or returns IEEE 128-bit types, generate a weak reference
2106 from the mangled name used in GCC 8.1 to the current mangled
2108 (rs6000_init_builtins): Make __ibm128 use the long double type if
2109 long double is IBM extended double. Make __float128 use the long
2110 double type if long double is IEEE 128-bit.
2113 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
2114 macro for __ibm128 built-in functions.
2115 (PACK_IF): Add __ibm128 pack/unpack functions.
2116 (UNPACK_IF): Likewise.
2117 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
2118 enable long double built-in functions if long double is IEEE
2119 128-bit floating point.
2120 (rs6000_invalid_builtin): Update long double built-in function
2122 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
2123 functions, adjust the built-in function to use the long double
2124 built-in function if __ibm128 and long double are the same type.
2125 * doc/extend.texi (PowerPC builtins): Update documention for
2126 __builtin_{,un}pack_longdouble. Add documentation for
2127 __builtin_{,un}pack_ibm128.
2129 2018-06-06 Jim Wilson <jimw@sifive.com>
2131 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
2132 (struct machine_function): New field interrupt_mode.
2133 (riscv_handle_type_attribute): New function. Add forward declaration.
2134 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
2135 (riscv_expand_epilogue): Check interrupt_mode field.
2136 (riscv_set_current_function): Check interrupt attribute args and
2137 set interrupt_mode field.
2138 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
2139 (riscv_sret, riscv_uret): New.
2140 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
2141 new arguments to interrupt attribute.
2143 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
2146 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
2147 Don't handle -mcpu=power8 if -mpower9-vector is also used.
2149 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2151 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
2152 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
2153 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2154 several redundant entries.
2156 2018-06-06 David Malcolm <dmalcolm@redhat.com>
2158 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
2159 type from "rtx" to "rtx_insn *".
2160 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
2161 for local "call_insn", removing cast.
2162 (ix86_expand_call): Likewise, introducing a "call_insn" local.
2164 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
2166 PR tree-optimization/86066
2167 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
2168 for BIT_INSERT_EXPR stores.
2170 2018-06-06 Richard Biener <rguenther@suse.de>
2172 PR tree-optimization/86062
2173 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2174 component refs ontop
2175 of to be offsetted base.
2177 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2179 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
2180 to be static and remove check on interrupt attribute name.
2182 gcc/testsuite/gcc.target/msp430/
2183 * function-attributes-4.c: New test.
2184 * static-interrupts.c: New test.
2186 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
2188 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2189 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
2191 2018-06-05 Steve Ellcey <sellcey@cavium.com>
2194 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
2196 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
2197 Remove second argument, change how error is called.
2198 (aarch64_layout_arg): Remove second argument from
2199 aarch64_err_no_fpadvsimd call.
2200 (aarch64_init_cumulative_args): Ditto.
2201 (aarch64_gimplify_va_arg_expr): Ditto.
2202 * config/aarch64/aarch64.md (mov<mode>): Ditto.
2204 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
2206 * config/i386/i386.md (simple_return_indirect_internal): New expander.
2207 (*simple_return_indirect_internal<mode>): Rename from
2208 simple_return_indirect_internal. Use W mode iterator.
2209 (rstorssp): New expander.
2210 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
2211 (clrssbsy): New expander.
2212 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
2214 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
2216 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
2218 (cmse_check_pointed_object): Likewise.
2220 2018-06-05 Martin Liska <mliska@suse.cz>
2222 PR gcov-profile/47618
2223 * doc/invoke.texi: Document how -fprofile-dir format
2226 2018-06-05 Richard Biener <rguenther@suse.de>
2228 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
2229 removal pretend DOM info isn't available so we do not update
2230 it and only remove edges, not dominated blocks. Actually free
2231 DOM info in case we removed something. Remove unreachable blocks.
2232 (mfb_keep_latches): Work with either DOM info or marked backedges.
2233 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
2234 first. Mark backedges if DOM info isn't available.
2235 (Re-)compute DOM info after cleanup_control_flow_pre.
2237 2018-06-05 Richard Biener <rguenther@suse.de>
2239 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
2240 (locus_discrim_hasher::hash): Adjust.
2241 (locus_discrim_hasher::equal): Likewise.
2242 (next_discriminator_for_locus): Work on line directly.
2243 (same_line_p): Pass in expanded locus1 as well.
2244 (assign_discriminators): Avoid redundant location expansions.
2246 2018-06-05 Richard Biener <rguenther@suse.de>
2248 PR tree-optimization/86046
2249 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
2250 if required after clearing TREE_ADDRESSABLE.
2252 2018-06-05 Richard Biener <rguenther@suse.de>
2254 PR tree-optimization/86047
2255 * tree-ssa-loop.c (for_each_index): Glob handling of all
2256 decls and constants and really handle all of them.
2258 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2261 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
2262 qualifier_void_pointer and qualifier_const_void_pointer.
2263 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
2264 (arm_init_builtins): Handle the above.
2265 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
2266 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
2267 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
2270 2018-06-05 Martin Liska <mliska@suse.cz>
2272 * auto-profile.c (read_autofdo_file): Do not use
2273 gcov_ctr_summary struct.
2274 (afdo_callsite_hot_enough_for_early_inline): Likewise.
2275 * coverage.c (struct counts_entry): Likewise.
2276 (read_counts_file): Read just single summary entry.
2277 (get_coverage_counts): Use gcov_summary struct.
2278 * coverage.h (get_coverage_counts): Likewise.
2279 * gcov-dump.c (dump_working_sets): Likewise.
2280 (tag_summary): Dump just single summary.
2281 * gcov-io.c (gcov_write_summary): Write just histogram
2283 (gcov_read_summary): Read just single summary.
2284 (compute_working_sets): Use gcov_summary struct.
2285 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
2286 of GCOV_COUNTERS_SUMMABLE.
2287 (GCOV_COUNTERS_SUMMABLE): Remove.
2288 (GCOV_FIRST_VALUE_COUNTER): Replace with
2289 GCOV_COUNTER_V_INTERVAL.
2290 (struct gcov_ctr_summary): Remove.
2291 (struct gcov_summary): Directly use fields of former
2293 (compute_working_sets): Use gcov_summary struct.
2294 * gcov.c (read_count_file): Do not use ctrs fields.
2295 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
2297 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
2299 * profile.c: Likewise.
2300 * profile.h: Likewise.
2302 2018-06-05 Martin Liska <mliska@suse.cz>
2304 PR gcov-profile/84846
2305 * gcov.c (output_lines): Print working directory only
2306 in intermediate format.
2308 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
2310 * config/s390/s390-builtin-types.def: Add void function type.
2311 * config/s390/s390-builtins.def: Use the function type for the
2314 2018-06-04 Jim Wilson <jimw@sifive.com>
2316 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
2318 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
2319 and maybe_eh_return. Change regno to unsigned int. Use new args to
2320 handle EH_RETURN_DATA_REGNO registers properly.
2321 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
2322 (riscv_expand_epilogue): Update comment. Change argument name and
2323 type. Update code to use new name and type. Pass new args to
2324 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
2326 * config/riscv/riscv.md (NORMAL_RETURN): New.
2327 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
2328 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
2329 (eh_return): Call gen_eh_return_internal and emit barrier.
2330 (eh_return_internal): Call riscv_expand_epilogue.
2332 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
2334 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
2335 bit_insertion field and declare can_be_merged_into method.
2336 (merged_store_group::can_be_merged_into): New method.
2337 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
2338 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
2339 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
2341 2018-06-04 Richard Biener <rguenther@suse.de>
2343 PR tree-optimization/85955
2344 * builtins.c (fold_builtin_sincos): Convert pointers to
2345 destination to appropriate type before dereferencing.
2347 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
2349 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
2351 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
2353 * expr.c (expand_expr_real_1): Force the operand into memory if
2354 its TYPE_MODE is BLKmode and if there is no integer mode for
2355 the number of bits being extracted.
2357 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2361 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2362 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
2364 2018-06-04 Richard Biener <rguenther@suse.de>
2366 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
2367 (cleanup_tree_cfg_noloop): ... single caller. Do
2368 start_recording_case_labels later.
2370 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
2372 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
2373 to _IMMINTRIN_H_INCLUDED.
2374 * config/i386/pconfigintrin.h: Ditto.
2375 * config/i386/waitpkgintrin.h: Ditto.
2376 * config/i386/immintrin.h: Add includes for sgxintrin.h,
2377 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
2378 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
2379 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
2380 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
2381 waitpkgintrin.h and cldemoteintrin.h.
2383 2018-06-04 Richard Biener <rguenther@suse.de>
2385 PR tree-optimization/86038
2386 * tracer.c (find_best_successor): Check probability for
2387 being initialized, bail out if not.
2389 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
2392 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
2393 of bits to ignore when comparing architectures.
2395 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2397 PR tree-optimization/69615
2398 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
2399 maximum or minimum of the type, try to merge it also as if
2400 range1 is + [-, x - 1] or + [x + 1, -].
2403 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
2405 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
2407 PR tree-optimization/86034
2408 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
2409 the unsigned bitfield type in a bit insertion sequence if it does not
2410 have a larger precision than the bitfield size.
2411 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
2413 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
2415 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
2417 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2419 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
2420 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
2421 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
2422 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
2424 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2426 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
2427 Disable -fdelete-null-pointer-checks for ELF toolchain.
2429 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2430 Kito Cheng <kito.cheng@gmail.com>
2432 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
2433 (nds32le-*-*, nds32be-*-*): Integrate checking process.
2434 (nds32*-*-*): Add glibc and uclibc conditions.
2435 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
2436 (TARGET_EXCEPT_UNWIND_INFO): Define.
2437 * config/nds32/elf.h: New file.
2438 * config/nds32/linux.h: New file.
2439 * config/nds32/nds32-elf.opt: New file.
2440 * config/nds32/nds32-linux.opt: New file.
2441 * config/nds32/nds32-fp-as-gp.c
2442 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
2443 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
2445 (nds32_asm_file_end): Ditto.
2446 (nds32_print_operand): Ditto.
2447 (nds32_insert_attributes): Ditto.
2448 (nds32_init_libfuncs): New function.
2449 (TARGET_HAVE_TLS): Define.
2450 (TARGET_INIT_LIBFUNCS): Define.
2451 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
2453 (TARGET_ELF): Apply different mcmodel setting.
2454 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
2455 been migrated into elf.h and linux.h files.
2456 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
2457 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
2458 (mcmodel): The content has been migrated into nds32-elf.opt and
2459 nds32-linux.opt files.
2460 * config/nds32/t-elf: New file.
2461 * config/nds32/t-linux: New file.
2463 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2464 Shiva Chen <shiva0217@gmail.com>
2466 * config/nds32/constants.md (unspec_volatile_element): Add
2467 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
2468 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
2470 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
2471 (make_pass_nds32_fp_as_gp): Declare.
2472 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
2474 (nds32_asm_function_end_prologue): Remove unused asm output.
2475 (nds32_asm_function_begin_epilogue): Remove unused asm output.
2476 (nds32_asm_file_start): Output necessary fp_as_gp information.
2477 (nds32_option_override): Adjust register usage.
2478 (nds32_expand_prologue): Consider fp_as_gp situation.
2479 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
2480 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
2483 (simple_return): Ditto.
2484 (omit_fp_begin): Output special directive for fp_as_gp.
2485 (omit_fp_end): Output special directive for fp_as_gp.
2486 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
2487 mforbid-fp-as-gp): New options.
2489 2018-06-01 Mark Wielaard <mark@klomp.org>
2491 * dwarf2out.c (dwarf2out_finish): Remove generation of
2492 DW_AT_loclists_base.
2494 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
2496 * gimple-ssa-store-merging.c: Include gimple-fold.h.
2497 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
2498 (struct merged_store_group): Add bit_insertion field.
2499 (dump_char_array): Use standard hexadecimal format.
2500 (merged_store_group::merged_store_group): Set bit_insertion to false.
2501 (merged_store_group::apply_stores): Use optimal buffer size. Deal
2502 with BIT_INSERT_EXPR stores. Move up code updating the mask and
2503 also print the mask in the dump file.
2504 (pass_store_merging::gate): Minor tweak.
2505 (imm_store_chain_info::coalesce_immediate): Fix wrong association
2506 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
2507 stores with INTEGER_CST stores.
2508 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
2509 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
2510 and use it throughout. Generate bit insertion sequences if need be.
2511 (pass_store_merging::process_store): Remove redundant condition.
2512 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
2514 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
2516 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
2517 the 128-bit floating point types. Fix function comment.
2519 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2521 * config/aarch64/aarch64-simd.md
2522 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
2524 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
2527 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2529 PR tree-optimization/85989
2530 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
2532 (backprop::intersect_uses): Check it when deciding whether this
2533 is a backedge reference.
2534 (backprop::process_block): Add each phi to m_visited_phis
2535 after visiting it, then clear it at the end.
2537 2018-06-01 Richard Biener <rguenther@suse.de>
2539 * tree-vectorizer.h (vect_dr_stmt): New function.
2540 (vect_get_load_cost): Adjust.
2541 (vect_get_store_cost): Likewise.
2542 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
2543 Use vect_dr_stmt instead of DR_SMTT.
2544 (vect_record_base_alignments): Likewise.
2545 (vect_calculate_target_alignment): Likewise.
2546 (vect_compute_data_ref_alignment): Likewise and make static.
2547 (vect_update_misalignment_for_peel): Likewise.
2548 (vect_verify_datarefs_alignment): Likewise.
2549 (vector_alignment_reachable_p): Likewise.
2550 (vect_get_data_access_cost): Likewise. Pass down
2551 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
2552 (vect_get_peeling_costs_all_drs): Likewise.
2553 (vect_peeling_hash_get_lowest_cost): Likewise.
2554 (vect_enhance_data_refs_alignment): Likewise.
2555 (vect_find_same_alignment_drs): Likewise.
2556 (vect_analyze_data_refs_alignment): Likewise.
2557 (vect_analyze_group_access_1): Likewise.
2558 (vect_analyze_group_access): Likewise.
2559 (vect_analyze_data_ref_access): Likewise.
2560 (vect_analyze_data_ref_accesses): Likewise.
2561 (vect_vfa_segment_size): Likewise.
2562 (vect_small_gap_p): Likewise.
2563 (vectorizable_with_step_bound_p): Likewise.
2564 (vect_prune_runtime_alias_test_list): Likewise.
2565 (vect_analyze_data_refs): Likewise.
2566 (vect_supportable_dr_alignment): Likewise.
2567 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2568 (vect_gen_prolog_loop_niters): Likewise.
2569 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2570 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
2572 (vect_recog_mask_conversion_pattern): Likewise.
2573 (vect_try_gather_scatter_pattern): Likewise.
2574 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
2575 to vect_get_store_cost.
2576 (vect_get_store_cost): Get stmt_info instead of DR.
2577 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
2578 (vect_get_load_cost): Get stmt_info instead of DR.
2580 2018-06-01 Richard Biener <rguenther@suse.de>
2583 * gimple-fold.c (var_decl_component_p): Also allow offsetted
2584 vars wrapped in MEM_REFs.
2586 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2588 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2589 Fix subreg tests so that we only return a choice between
2590 GENERAL_REGS and FP_REGS if the original classes included both.
2592 2018-06-01 Richard Biener <rguenther@suse.de>
2595 * tree-ssa-structalias.c (get_function_part_constraint):
2596 Handle NULL fi->decl.
2597 (find_func_aliases_for_call): Properly handle indirect
2598 fi from direct call.
2599 (find_func_clobbers): Likewise.
2600 (ipa_pta_execute): Likewise.
2601 (create_variable_info_for): For functions that are ifunc_resolver
2602 resolve to a varinfo that contains the result of the resolver call.
2603 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
2606 2018-05-31 Michael Collison <michael.collison@arm.com>
2608 * config/aarch64/aarch64.md:
2609 (*fix_to_zero_extenddfdi2): New pattern.
2610 * gcc.target/aarch64/fix_extend1.c: New testcase.
2612 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
2616 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
2617 and BUILT_IN_STRNCMP_EQ.
2618 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
2619 BUILT_IN_STRNCMP_EQ.
2620 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
2621 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2622 (gimple_fold_builtin): Likewise.
2623 * tree-ssa-strlen.c (compute_string_length): New function.
2624 (determine_min_obsize): New function.
2625 (handle_builtin_string_cmp): New function to handle calls to
2626 string compare functions.
2627 (strlen_optimize_stmt): Add handling to builtin string compare
2629 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2630 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2631 * tree.c (build_common_builtin_nodes): Add new defines of
2632 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
2634 2018-05-31 Jakub Jelinek <jakub@redhat.com>
2637 * bb-reorder.c (pass_partition_blocks::gate): Return false for
2638 functions with naked attribute.
2640 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
2642 * config/i386/sse.md (avx_vec_concat<mode>):
2643 Substitute concat_tg_mode mode attribute with xtg_mode.
2644 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
2645 (concat_tg_mode): Remove mode attribute.
2647 2018-05-31 Martin Sebor <msebor@redhat.com>
2650 * calls.c (alloc_max_size): Correct a logic error/typo.
2651 Treat excessive arguments as infinite. Warn for invalid arguments.
2652 * doc/invoke.texi (-Walloc-size-larger-than): Update.
2654 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
2657 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
2658 and movx for Haswell.
2660 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
2661 Cesar Philippidis <cesar@codesourcery.com>
2664 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
2665 when emitting error on private/firstprivate reductions.
2666 * omp-low.c (lower_omp_target): Avoid reference-type processing
2667 on pointers for firstprivate clause.
2669 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2671 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
2674 * config/aarch64/aarch64-simd.md
2675 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
2676 (aarch64_ld1_x3_<mode>): Likewise
2677 (aarch64_st1x2<VALLDIF:mode>): Likewise
2678 (aarch64_st1_x2_<mode>): Likewise
2679 (aarch64_st1x3<VALLDIF:mode>): Likewise
2680 (aarch64_st1_x3_<mode>): Likewise
2681 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
2682 (vld1_s8_x3): Likewise.
2683 (vld1_u16_x3): Likewise.
2684 (vld1_s16_x3): Likewise.
2685 (vld1_u32_x3): Likewise.
2686 (vld1_s32_x3): Likewise.
2687 (vld1_u64_x3): Likewise.
2688 (vld1_s64_x3): Likewise.
2689 (vld1_f16_x3): Likewise.
2690 (vld1_f32_x3): Likewise.
2691 (vld1_f64_x3): Likewise.
2692 (vld1_p8_x3): Likewise.
2693 (vld1_p16_x3): Likewise.
2694 (vld1_p64_x3): Likewise.
2695 (vld1q_u8_x3): Likewise.
2696 (vld1q_s8_x3): Likewise.
2697 (vld1q_u16_x3): Likewise.
2698 (vld1q_s16_x3): Likewise.
2699 (vld1q_u32_x3): Likewise.
2700 (vld1q_s32_x3): Likewise.
2701 (vld1q_u64_x3): Likewise.
2702 (vld1q_s64_x3): Likewise.
2703 (vld1q_f16_x3): Likewise.
2704 (vld1q_f32_x3): Likewise.
2705 (vld1q_f64_x3): Likewise.
2706 (vld1q_p8_x3): Likewise.
2707 (vld1q_p16_x3): Likewise.
2708 (vld1q_p64_x3): Likewise.
2709 (vst1_s64_x2): Likewise.
2710 (vst1_u64_x2): Likewise.
2711 (vst1_f64_x2): Likewise.
2712 (vst1_s8_x2): Likewise.
2713 (vst1_p8_x2): Likewise.
2714 (vst1_s16_x2): Likewise.
2715 (vst1_p16_x2): Likewise.
2716 (vst1_s32_x2): Likewise.
2717 (vst1_u8_x2): Likewise.
2718 (vst1_u16_x2): Likewise.
2719 (vst1_u32_x2): Likewise.
2720 (vst1_f16_x2): Likewise.
2721 (vst1_f32_x2): Likewise.
2722 (vst1_p64_x2): Likewise.
2723 (vst1q_s8_x2): Likewise.
2724 (vst1q_p8_x2): Likewise.
2725 (vst1q_s16_x2): Likewise.
2726 (vst1q_p16_x2): Likewise.
2727 (vst1q_s32_x2): Likewise.
2728 (vst1q_s64_x2): Likewise.
2729 (vst1q_u8_x2): Likewise.
2730 (vst1q_u16_x2): Likewise.
2731 (vst1q_u32_x2): Likewise.
2732 (vst1q_u64_x2): Likewise.
2733 (vst1q_f16_x2): Likewise.
2734 (vst1q_f32_x2): Likewise.
2735 (vst1q_f64_x2): Likewise.
2736 (vst1q_p64_x2): Likewise.
2737 (vst1_s64_x3): Likewise.
2738 (vst1_u64_x3): Likewise.
2739 (vst1_f64_x3): Likewise.
2740 (vst1_s8_x3): Likewise.
2741 (vst1_p8_x3): Likewise.
2742 (vst1_s16_x3): Likewise.
2743 (vst1_p16_x3): Likewise.
2744 (vst1_s32_x3): Likewise.
2745 (vst1_u8_x3): Likewise.
2746 (vst1_u16_x3): Likewise.
2747 (vst1_u32_x3): Likewise.
2748 (vst1_f16_x3): Likewise.
2749 (vst1_f32_x3): Likewise.
2750 (vst1_p64_x3): Likewise.
2751 (vst1q_s8_x3): Likewise.
2752 (vst1q_p8_x3): Likewise.
2753 (vst1q_s16_x3): Likewise.
2754 (vst1q_p16_x3): Likewise.
2755 (vst1q_s32_x3): Likewise.
2756 (vst1q_s64_x3): Likewise.
2757 (vst1q_u8_x3): Likewise.
2758 (vst1q_u16_x3): Likewise.
2759 (vst1q_u32_x3): Likewise.
2760 (vst1q_u64_x3): Likewise.
2761 (vst1q_f16_x3): Likewise.
2762 (vst1q_f32_x3): Likewise.
2763 (vst1q_f64_x3): Likewise.
2764 (vst1q_p64_x3): Likewise.
2766 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2768 * config/msp430/msp430.c (msp430_output_labelref): Prepend
2769 user_label_prefix to name.
2771 * tree-core.h: Update comment about the format of NAME string
2772 passed to handler in attribute_spec.
2774 * config/msp430/msp430.md: Remove erroneous subreg expression from
2775 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
2776 zero_extend{q,h}isi2.
2778 2018-05-30 Borislav Petkov <bp@suse.de>
2780 * doc/extend.texi: Document some architecture specific
2781 constraints and sort entries.
2783 2018-05-30 Martin Sebor <msebor@redhat.com>
2786 * builtins.c (expand_builtin_stpcpy_1): New function.
2787 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
2788 only if the former succeeds.
2790 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2792 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
2795 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2797 * doc/invoke.texi (-flinker-output): Document
2799 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2801 * passes.c (ipa_write_summaries): Only modify statements if body
2803 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
2804 incrementally linking.
2805 (ipa_passes): Likewise.
2806 * lto-cgraph.c (lto_output_node): When incrementally linking do not
2807 pass down resolution info.
2808 * common.opt (flag_incremental_link): Update info.
2809 * gcc.c (plugin specs): Turn flinker-output=* to
2810 -plugin-opt=-linker-output-known
2811 * toplev.c (compile_file): Also cut compilation when doing incremental
2813 * flag-types. (enum lto_partition_model): Add
2814 LTO_LINKER_OUTPUT_NOLTOREL.
2815 (invoke.texi): Add -flinker-output docs.
2816 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
2817 link same way as WPA; do not stream in dead initializers.
2819 * dwarf2out.c (dwarf2out_die_ref_for_decl,
2820 darf2out_register_external_decl): Support incremental link.
2822 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2824 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
2826 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2828 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
2829 it down to simple_object_copy_lto_debug_sections.
2830 (run_gcc): Determine incremental LTO link time and configure
2831 lto1 into non-wpa mode, disable renaming of debug sections.
2833 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
2835 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2836 descriptions of various incorrectly documented functions.
2838 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
2841 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2842 address check not strict.
2844 2018-05-30 Richard Biener <rguenther@suse.de>
2846 PR tree-optimization/85964
2847 * tracer.c (better_p): Drop initialized count check, we only
2848 call the function with initialized counts now.
2849 (find_best_successor): Do find a best edge if one
2850 has uninitialized count.
2851 (find_best_predecessor): Likewise. Do BB frequency check only
2852 if count is initialized.
2854 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
2856 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
2857 (aarch64_ldrstr_offset_compare): New.
2858 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
2859 load/store orderings.
2860 (aarch64_gen_adjusted_ldpstp): Likewise.
2862 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
2864 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2865 Check for subset of GENERAL_REGS and FP_REGS.
2866 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
2869 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
2871 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
2872 and wi::to_poly_offset. Add the current offset and then check
2873 whether the sum fits, rather than using an unchecked addition of
2874 a checked term. Check for a shwi rather than a uhwi.
2875 * expr.c (get_bit_range): Use tree_to_poly_uint64.
2876 (store_constructor): Use poly_int_tree_p.
2877 (expand_expr_real_1): Likewise.
2878 * function.c (assign_temp): Likewise.
2879 * fold-const.c (const_binop): Use poly_int_tree_p and
2881 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
2883 * ipa-icf-gimple.c (func_checker::compare_operand): Use
2884 to_poly_offset for MEM offsets.
2885 * ipa-icf.c (sem_variable::equals): Likewise.
2886 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
2887 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
2888 wi::to_poly_offset for BIT_FIELD_REF offsets.
2889 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
2891 * var-tracking.c (emit_note_insn_var_location): Use
2892 tree_to_poly_uint64.
2894 2018-05-29 Jim Wilson <jimw@sifive.com>
2896 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
2898 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
2901 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
2902 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
2904 (sse4_1_round<mode>2): Use nonimmediate_operand
2905 for operand 1 predicate.
2907 2018-05-29 Martin Sebor <msebor@redhat.com>
2908 Richard Biener <rguenther@suse.de>
2911 * calls.c (get_size_range): Call determine_value_range instead
2912 of get_value_range..
2913 * tree-vrp.h (determine_value_range): Declared new function.
2914 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
2916 2018-05-29 Richard Biener <rguenther@suse.de>
2918 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
2919 sure to use non-pattern stmts for get_earlier_stmt arguments.
2920 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
2921 called on pattern stmts.
2922 (get_later_stmt): Likewise.
2924 2018-05-29 Martin Liska <mliska@suse.cz>
2926 PR gcov-profile/85759
2927 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
2930 2018-05-29 Jakub Jelinek <jakub@redhat.com>
2932 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
2934 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
2938 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2939 VEC_PACK_FLOAT_EXPR): New tree codes.
2940 * tree-pretty-print.c (op_code_prio): Handle
2941 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2942 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2943 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2944 * tree-inline.c (estimate_operator_cost): Likewise.
2945 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
2946 * fold-const.c (const_binop): Likewise.
2947 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
2948 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2949 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2950 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
2951 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2952 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2953 * expr.c (expand_expr_real_2): Likewise.
2954 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
2955 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
2956 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
2958 * optabs.c (expand_widen_pattern_expr): For
2959 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
2960 sign from result type rather than operand's type.
2961 (expand_binop_directly): For vec_packu_float_optab and
2962 vec_packs_float_optab allow result type to be different from operand's
2964 * optabs-tree.c (optab_for_tree_code): Handle
2965 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2966 VEC_PACK_FLOAT_EXPR. Formatting fixes.
2967 * tree-vect-generic.c (expand_vector_operations_1): Handle
2968 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2969 VEC_PACK_FLOAT_EXPR.
2970 * tree-vect-stmts.c (supportable_widening_operation): Handle
2972 (supportable_narrowing_operation): Handle FLOAT_EXPR.
2973 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
2974 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
2975 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
2976 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
2978 (vec_pack<floatprefix>_float_<mode>): New expander.
2979 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
2981 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
2982 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
2983 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
2984 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
2985 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
2987 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
2988 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
2989 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2990 VEC_PACK_FLOAT_EXPR): Document.
2992 2018-05-29 Richard Biener <rguenther@suse.de>
2994 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
2996 (stmt_vec_info_vec): Make pointer.
2997 (init_stmt_vec_info_vec): Remove.
2998 (free_stmt_vec_info_vec): Likewise.
2999 (set_stmt_vec_info_vec): New function.
3000 (free_stmt_vec_infos): Likewise.
3001 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
3002 (set_vinfo_for_stmt): Likewise.
3003 (get_earlier_stmt): Likewise.
3004 (get_later_stmt): Likewise.
3005 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
3006 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
3007 (vec_info::~vec_info): Free stmt_vec_infos.
3008 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
3009 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
3010 (pass_slp_vectorize::execute): Likewise.
3011 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
3012 (free_stmt_vec_info_vec): Likewise.
3013 (set_stmt_vec_info_vec): New function.
3014 (free_stmt_vec_infos): Likewise.
3015 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
3016 the global stmt_vec_info_vec.
3017 * tree-parloops.c (gather_scalar_reductions): Use
3018 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
3021 2018-05-29 Richard Biener <rguenther@suse.de>
3023 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
3025 2018-05-29 Martin Liska <mliska@suse.cz>
3026 David Malcolm <dmalcolm@redhat.com>
3028 * vec.c (test_reverse): New.
3029 (vec_c_tests): Add new test.
3030 * vec.h (vl_ptr>::reverse): New function.
3032 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
3034 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
3036 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
3039 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
3041 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
3043 2018-05-28 Richard Biener <rguenther@suse.de>
3045 PR tree-optimization/85933
3046 * tree-vect-data-refs.c (vect_record_base_alignments): Only
3047 look at stmts marked as vectorizable.
3049 2018-05-28 Richard Biener <rguenther@suse.de>
3051 PR tree-optimization/85934
3052 * tree-vect-generic.c (expand_vector_operations_1): Hoist
3053 vector boolean check before scalar optimization.
3055 2018-05-28 Jakub Jelinek <jakub@redhat.com>
3057 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
3060 2018-05-28 Mark Wielaard <mark@klomp.org>
3062 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
3063 if it is an expression containing a minus sign.
3065 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
3067 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
3069 2018-05-27 Paul Koning <ni1d@arrl.net>
3071 * config/pdp11/pdp11.md (truncsihi2): Remove.
3073 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
3074 Chung-Ju Wu <jasonwucj@gmail.com>
3076 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
3078 (unaligned_store_dw): Ditto.
3079 * config/nds32/nds32-memory-manipulation.c
3080 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
3081 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
3082 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
3083 (emit_setmem_word_loop): Rename to ...
3084 (emit_setmem_doubleword_loop): ... this.
3085 (nds32_gen_dup_4_byte_to_word_value): New function.
3086 (nds32_gen_dup_8_byte_to_double_word_value): New function.
3087 (nds32_expand_setmem_loop): Refine implementation.
3088 (nds32_expand_setmem_loop_v3m): Ditto.
3089 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
3092 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3094 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
3096 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3098 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
3099 (nds32_init_machine_status): Initialize machine->attr_naked_p and
3100 machine->attr_no_prologue_p.
3101 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
3102 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
3103 (nds32_expand_epilogue): Consider attr_naked_p.
3104 (nds32_expand_epilogue_v3pop): Likewise.
3105 (nds32_can_use_return_insn): Likewise.
3106 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
3107 attr_no_prologue_p fields.
3108 * config/nds32/nds32.opt (mret-in-naked-func): New option.
3110 2018-05-27 Jakub Jelinek <jakub@redhat.com>
3113 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
3115 * config/i386/sse.md
3116 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
3118 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
3120 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
3122 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
3124 (*<floatsuffix>floatv2div2sf2): Rename to ...
3125 (*float<floatunssuffix>v2div2sf2): ... this.
3126 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
3127 (float<floatunssuffix>v2div2sf2_mask): ... this.
3128 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
3129 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
3130 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
3132 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
3134 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3136 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3138 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3140 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3142 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
3143 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
3144 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
3145 gen_ufix_truncv8dfv8si2.
3146 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
3147 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
3148 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
3149 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
3150 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
3151 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
3152 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
3153 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
3155 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3159 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
3161 2018-05-25 Jim Wilson <jimw@sifive.com>
3163 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
3164 * config/riscv/riscv.c (struct machine_function): Add
3165 interrupt_handler_p and attribute_checked_p fields.
3166 (riscv_attribute_table): Add interrupt.
3167 (riscv_interrupt_type_p): New.
3168 (riscv_save_reg_p): Save extra regs for interrupt handler.
3169 (riscv_use_save_libcall): Return false for interrupt handler.
3170 (riscv_first_stack_step): Add forward declaration.
3171 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
3172 for interrupt handler with large frame. Use it for saved reg list.
3173 (riscv_expand_prologue): Move flag_stack_usage_info support to
3174 eliminate duplication.
3175 (riscv_expand_epilogue): Generate mret for interrupt handler.
3176 (riscv_epilogue_uses): New.
3177 (riscv_can_use_return_insn): Return false for interrupt handler.
3178 (riscv_function_ok_for_sibcall): Likewise.
3179 (riscv_set_current_function): Add interrupt handler support.
3180 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
3181 * config/riscv/riscv.md (UNSPECV_MRET): New.
3183 (riscv_frflags, riscv_fsflags): Use tab after opcode.
3185 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
3187 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
3189 PR tree-optimization/85712
3190 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
3191 this candidate has already been replaced in-situ by a copy.
3193 2018-05-25 Jason Merrill <jason@redhat.com>
3195 PR c++/80485 - inline function non-zero address.
3196 * symtab.c (nonzero_address): Check DECL_COMDAT.
3198 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
3201 * config/alpha/alpha.md (ashlsi3): New insn pattern.
3202 (*ashlsi_se): Rename from *ashldi_se. Define as sign
3203 extension of SImode operation. Use const123_operand predicate.
3204 (*saddsi_1): Remove.
3205 (*saddl_se_1): Ditto.
3207 (*ssubl_se_1): Ditto.
3208 * config/alpha/predicates.md (const123_operand): New predicate.
3209 * config/alpha/constraints.md (P): Use IN_RANGE.
3211 2018-05-25 Richard Biener <rguenther@suse.de>
3213 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
3215 (ref_maybe_used_by_stmt_p): Likewise.
3216 (stmt_may_clobber_ref_p): Likewise.
3217 (stmt_may_clobber_ref_p_1): Likewise.
3218 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
3220 (ref_maybe_used_by_stmt_p): Likewise.
3221 (stmt_may_clobber_ref_p): Likewise.
3222 (stmt_may_clobber_ref_p_1): Likewise.
3223 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
3224 the alias oracle to disambiguate DRs with stmts DR analysis
3226 (vect_analyze_data_refs): Do not give up on not analyzable
3227 DRs for BB vectorization. Remove code truncating the dataref
3230 2018-05-25 Jakub Jelinek <jakub@redhat.com>
3233 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3234 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
3235 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
3237 2018-05-25 Richard Biener <rguenther@suse.de>
3239 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
3240 function, combining stmt data ref gathering and fatal analysis
3242 (vect_analyze_data_refs): Remove now redudnant code and simplify.
3243 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
3244 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
3245 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
3246 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
3248 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3250 PR tree-optimization/85720
3251 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
3252 SCC if all partitions are builtins.
3253 (version_loop_by_alias_check): New parameter. Generate cancelable
3254 runtime alias check if all partitions are builtins.
3255 (distribute_loop): Update call to above function.
3257 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3259 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
3260 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
3261 (parm_default_def_partition_arg): Ditto.
3262 (set_parm_default_def_partition): Ditto.
3263 (get_parm_default_def_partitions): Ditto and make it static.
3264 (get_undefined_value_partitions): Ditto and make it static.
3265 (remove_ssa_form): Refactor call to init_var_map here.
3266 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
3267 computation for loop region.
3268 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
3269 (register_default_def): Delete.
3270 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
3271 (parm_default_def_partition_arg): Ditto.
3272 (set_parm_default_def_partition): Ditto.
3273 (get_parm_default_def_partitions): Ditto and make it static.
3274 (get_undefined_value_partitions): Ditto and make it static.
3275 (coalesce_with_default, coalesce_with_default): Update comment.
3276 (create_coalesce_list_for_region): New func factored out from
3277 create_outofssa_var_map.
3278 (populate_coalesce_list_for_outofssa): New func factored out from
3279 create_outofssa_var_map and coalesce_ssa_name.
3280 (create_outofssa_var_map): Delete.
3281 (coalesce_ssa_name): Refactor to support live range computation.
3282 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
3283 (get_parm_default_def_partitions): Delete.
3284 (get_undefined_value_partitions): Ditto.
3285 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
3286 computation for loop region.
3287 (new_tree_live_info, loe_visit_block): Ditto.
3288 (live_worklist, set_var_live_on_entry): Ditto.
3289 (calculate_live_on_exit, verify_live_on_entry): Ditto.
3290 * tree-ssa-live.h (struct _var_map): New fields.
3291 (init_var_map): Change decl.
3292 (region_contains_p): New.
3294 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3296 * tree-ssa-live.h (live_merge_and_clear): Delete.
3298 2018-05-25 Richard Biener <rguenther@suse.de>
3301 * tree-dump.c (dequeue_and_dump): Remove access to removed
3302 operand 2 of a SWITCH_EXPR.
3304 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3306 * doc/sourcebuild.texi (vect_double_cond_arith): Include
3307 multiplication and division.
3308 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
3309 (cond_udiv@var{m}, cond_umod@var{m}): Document.
3310 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
3311 (cond_udiv_optab, cond_umod_optab): New optabs.
3312 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
3313 (IFN_COND_RDIV): New internal functions.
3314 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
3315 TRUNC_MOD_EXPR and RDIV_EXPR.
3316 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
3317 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
3319 (SVE_INT_BINARY): Include mult.
3320 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
3321 (optab, sve_int_op): Handle mult.
3322 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
3324 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
3325 for SVE_INT_BINARY_SD.
3327 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3329 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
3330 (optab, sve_int_op): Handle div and udiv.
3331 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
3332 for SVE_INT_BINARY_SD.
3333 (*<optab><mode>3): New insn for the same.
3335 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3337 * tree-vect-patterns.c: Include predict.h.
3338 (vect_recog_divmod_pattern): Restrict check for division support
3339 to when optimizing for size.
3341 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3343 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
3344 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
3345 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
3346 (gimple_match_op::set_op): Likewise.
3347 (gimple_resimplify4): Declare.
3348 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
3349 (expr::gen_transform): Likewise.
3350 (decision_tree::gen): Generate a simplification routine for 4 operands.
3351 * gimple-match-head.c (gimple_simplify): Add an overload for
3352 4 operands. In the top-level function, handle up to 4 call
3353 arguments and call gimple_resimplify4.
3354 (gimple_resimplify4): New function.
3355 (build_call_internal): Pass a fourth operand.
3356 (maybe_push_to_seq): Likewise.
3357 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
3358 Fold VEC_COND_EXPRs of an operation and a default value into
3359 an IFN_COND_* function if possible.
3360 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
3362 (SVE_COND_FP_BINARY): Include them.
3363 (optab, sve_fp_op): Handle them.
3364 (SVE_INT_BINARY_REV): New code iterator.
3365 (SVE_COND_FP_BINARY_REV): New int iterator.
3366 (commutative): New int attribute.
3367 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
3369 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
3371 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
3372 (*cond_<optab><mode>): New patterns for reversed operands.
3374 2018-05-25 Richard Biener <rguenther@suse.de>
3376 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
3377 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
3378 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
3379 (STMT_VINFO_GROUPED_ACCESS): Adjust.
3380 * tree-vect-data-refs.c (everywhere): Adjust users.
3381 * tree-vect-loop.c (everywhere): Likewise.
3382 * tree-vect-slp.c (everywhere): Likewise.
3383 * tree-vect-stmts.c (everywhere): Likewise.
3384 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
3386 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3388 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
3390 (gcc_cv_as_section_exclude): ... this.
3391 Try Solaris as #exclude syntax.
3392 * configure: Regenerate.
3393 * config.in: Regenerate.
3394 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
3396 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
3397 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
3399 * varasm.c (default_elf_asm_named_section): Don't check if
3400 HAVE_GAS_SECTION_EXCLUDE is defined.
3402 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3404 * doc/md.texi: Update the documentation of the cond_* optabs
3405 to mention the new final operand. Fix GET_MODE_NUNITS call.
3406 Describe the scalar case too.
3407 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
3408 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
3410 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
3411 (get_conditional_internal_fn): Update comment.
3412 * tree-vect-loop.c (vectorizable_reduction): Pass the original
3413 accumulator value as a final argument to conditional functions.
3414 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
3415 a define_expand and add an "else" operand. Assert for now that
3416 the else operand is equal to operand 2. Use SVE_INT_BINARY and
3417 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
3418 (*cond_<optab><mode>): New patterns.
3419 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
3420 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
3421 (UNSPEC_COND_EOR): Delete.
3422 (optab): Remove associated mappings.
3423 (SVE_INT_BINARY): New code iterator.
3424 (sve_int_op): Remove int attribute and add "minus" to the code
3426 (SVE_COND_INT_OP): Delete.
3427 (SVE_COND_FP_OP): Rename to...
3428 (SVE_COND_FP_BINARY): ...this.
3430 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3432 * optabs.c (can_reuse_operands_p): New function.
3433 (maybe_legitimize_operands): Try to reuse the results for
3436 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3438 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
3439 Add {q} suffix to insn mnemonic.
3441 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3443 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
3444 (msp430_warn_func_return): New.
3446 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
3448 * fold-const.c (tree_nonzero_bits): New function.
3449 * fold-const.h (tree_nonzero_bits): Likewise.
3450 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
3451 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
3453 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3457 * varasm.c (assemble_alias): Check ifunc_resolver only on
3460 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3463 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
3464 when memory input operand is handled.
3466 2018-05-24 Luis Machado <luis.machado@linaro.org>
3468 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
3470 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
3472 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3474 * match.pd: Delay FMA folds until after vectorization.
3476 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
3479 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3480 address check not strict.
3482 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3484 * gimple-match.h (gimple_match_op): New class.
3485 (mprts_hook): Replace parameters with a gimple_match_op *.
3486 (maybe_build_generic_op): Likewise.
3487 (gimple_simplified_result_is_gimple_val): Replace parameters with
3488 a const gimple_match_op *.
3489 (gimple_simplify): Replace code_helper * and tree * parameters with
3490 a gimple_match_op * parameter.
3491 (gimple_resimplify1): Replace code_helper *, tree and tree *
3492 parameters with a gimple_match_op * parameter.
3493 (gimple_resimplify2): Likewise.
3494 (gimple_resimplify3): Likewise.
3495 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
3496 parameters with a gimple_match_op * parameter.
3497 * gimple-match-head.c (gimple_simplify): Change prototypes of
3498 auto-generated functions to take a gimple_match_op * instead of
3499 separate code_helper * and tree * parameters. Make the same
3500 change in the top-level overload and update calls to the
3501 gimple_resimplify routines. Update calls to the auto-generated
3502 functions and to maybe_push_res_to_seq in the publicly-facing
3503 operation-specific gimple_simplify overloads.
3504 (gimple_match_op::MAX_NUM_OPS): Define.
3505 (gimple_resimplify1): Replace rcode and ops with a single res_op
3506 parameter. Update call to gimple_simplify.
3507 (gimple_resimplify2): Likewise.
3508 (gimple_resimplify3): Likewise.
3509 (mprts_hook): Replace parameters with a gimple_match_op *.
3510 (maybe_build_generic_op): Likewise.
3511 (build_call_internal): Replace type, nargs and ops with
3512 a gimple_match_op *.
3513 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
3514 with a single gimple_match_op *. Update calls to mprts_hook,
3515 build_call_internal and gimple_simplified_result_is_gimple_val.
3516 Factor out code that is common to the tree_code and combined_fn cases.
3517 * genmatch.c (expr::gen_transform): Replace tem_code and
3518 tem_ops with a gimple_match_op called tem_op. Update calls
3519 to the gimple_resimplify functions and maybe_push_res_to_seq.
3520 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
3521 res_ops. Update call to the gimple_resimplify functions.
3522 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
3523 (decision_tree::gen): Make the functions take a gimple_match_op *
3524 called res_op instead of separate res_code and res_ops parameters.
3525 Update call accordingly.
3526 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
3527 and ops with a single res_op parameter. Update calls to
3528 maybe_build_generic_op and maybe_push_res_to_seq.
3529 (fold_stmt_1): Update calls to gimple_simplify and
3530 replace_stmt_with_simplification.
3531 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
3532 and gimple_simplified_result_is_gimple_val.
3533 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
3535 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
3536 with a gimple_match_op *.
3537 (vn_nary_build_or_lookup): Likewise. Update call to
3538 vn_nary_build_or_lookup_1.
3539 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
3540 gimple_match_op *. Update calls to the gimple_resimplify routines
3541 and to gimple_simplified_result_is_gimple_val.
3542 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
3543 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
3544 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
3545 (visit_nary_op): Likewise.
3546 (visit_reference_op_load): Likewise.
3548 2018-05-23 Luis Machado <luis.machado@linaro.org>
3550 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
3551 modifier for printing the step amount.
3553 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
3556 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
3559 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
3561 * doc/sourcebuild.texi (Endianness): New subsubsection.
3563 2018-05-23 Luis Machado <luis.machado@linaro.org>
3565 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3566 <prefetch_dynamic_strides>: New const bool field.
3567 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3568 prefetch_dynamic_strides.
3569 (exynosm1_prefetch_tune): Likewise.
3570 (thunderxt88_prefetch_tune): Likewise.
3571 (thunderx_prefetch_tune): Likewise.
3572 (thunderx2t99_prefetch_tune): Likewise.
3573 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
3575 (aarch64_override_options_internal): Update to set
3576 PARAM_PREFETCH_DYNAMIC_STRIDES.
3577 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3578 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3579 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3580 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3581 prefetch-dynamic-strides setting.
3583 2018-05-23 Luis Machado <luis.machado@linaro.org>
3585 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3586 <minimum_stride>: New const int field.
3587 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3588 minimum_stride field defaulting to -1.
3589 (exynosm1_prefetch_tune): Likewise.
3590 (thunderxt88_prefetch_tune): Likewise.
3591 (thunderx_prefetch_tune): Likewise.
3592 (thunderx2t99_prefetch_tune): Likewise.
3593 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
3594 <default_opt_level>: Set to 3.
3595 (aarch64_override_options_internal): Update to set
3596 PARAM_PREFETCH_MINIMUM_STRIDE.
3597 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3598 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3599 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3600 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3601 stride is constant and is below the minimum stride threshold.
3603 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3605 * config/arm/arm-cpus.in (mode26): Delete.
3606 (armv4): Delete mode26 reference.
3607 * config/arm/arm.c (arm_configure_build_target): Delete use of
3610 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3612 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
3614 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
3615 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
3617 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
3618 (floatunsdidf2): Ditto.
3620 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3622 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
3623 (fixuns_trunc<mode>si2_avx512f): Ditto.
3624 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
3625 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
3626 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
3628 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3630 PR rtl-optimization/79985
3631 * df-scan.c (df_insn_refs_collect): Remove special case for
3632 global registers and asm statements.
3634 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3636 * extend.texi (Global Register Variables): Rewrite the bullet list.
3637 Note that the register is available for allocation. Note that access
3638 via inline asm must use constraints. Add note about async-signal
3639 handlers. Remove paragraph about automagic register selection.
3641 2018-05-23 Richard Biener <rguenther@suse.de>
3643 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
3644 of fixed offset from memset VN.
3646 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
3648 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
3650 (alloc_cand_and_find_basis): Initialize first_interp field.
3651 (slsr_process_mul): Modify first_interp field.
3652 (slsr_process_add): Likewise.
3653 (slsr_process_cast): Modify first_interp field for each new
3655 (slsr_process_copy): Likewise.
3656 (dump_candidate): Dump first_interp field.
3657 (replace_mult_candidate): Process all interpretations, not just
3659 (replace_rhs_if_not_dup): Likewise.
3660 (replace_one_candidate): Likewise.
3662 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
3664 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
3666 (aarch64_needs_frame_chain): New function.
3667 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
3669 2018-05-23 Sudakshina Das <sudi.das@arm.com>
3672 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
3673 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
3674 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
3675 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
3676 as true for strict-align.
3677 (aarch64_can_inline_p): Perform checks even when callee has no
3678 attributes to check for strict alignment.
3679 * doc/extend.texi (AArch64 Function Attributes): Document
3681 * doc/invoke.texi: (AArch64 Options): Likewise.
3683 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
3685 PR tree-optimization/85853
3686 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
3687 the handling of the root of the node to...
3688 (vect_slp_analyze_node_operations_1): ...this new function,
3689 and run the whole thing with the child nodes' def types
3690 set according to their SLP node's def type.
3692 2018-05-23 Richard Biener <rguenther@suse.de>
3695 * tree-data-ref.c (create_runtime_alias_checks): Defer
3696 and ignore overflow warnings.
3698 2018-05-23 Yury Gribov <tetra2005@gmail.com>
3700 PR tree-optimization/85822
3701 * tree-vrp.c (is_masked_range_test): Fix handling of negative
3704 2018-05-23 Richard Biener <rguenther@suse.de>
3706 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3707 memset constants via native_interpret_expr.
3709 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
3712 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
3714 (cgraph_node::create_alias): Likewise.
3715 (cgraph_node::get_availability): Check ifunc_resolver instead
3716 of looking up ifunc attribute.
3717 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3718 * varasm.c (do_assemble_alias): Likewise.
3719 (assemble_alias): Likewise.
3720 (default_binds_local_p_3): Likewise.
3721 * cgraph.h (cgraph_node): Add ifunc_resolver.
3722 (cgraph_node::only_called_directly_or_aliased_p): Return false
3724 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
3726 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
3727 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
3728 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
3729 instead of looking up ifunc attribute.
3731 2018-05-22 Luis Machado <luis.machado@linaro.org>
3733 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
3735 2018-05-22 Martin Sebor <msebor@redhat.com>
3738 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
3739 only when expasion succeeds.
3740 (expand_builtin_strcmp): Same.
3741 (expand_builtin_strncmp): Same.
3743 2018-05-22 Martin Sebor <msebor@redhat.com>
3745 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
3747 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3748 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3750 * config/aarch64/aarch64-ldpstp.md: Replace uses of
3751 aarch64_mem_pair_operand with memory_operand and delete operand swapping
3753 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3754 Add check for legitimate_address.
3755 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
3756 (aarch64_swap_ldrstr_operands): New.
3757 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
3760 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3761 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3763 * config/aarch64/aarch64.md: New patterns to generate stp
3765 (store_pair_sw, store_pair_dw): New patterns to generate stp for
3766 single words and double words.
3767 (load_pair_sw, load_pair_dw): Likewise.
3768 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
3770 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
3772 * config/aarch64/aarch64-ldpstp.md: Modify peephole
3773 for different mode ldpstp and add peephole for merged zero stores.
3775 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3777 (aarch64_gen_store_pair): Rename calls to match new patterns.
3778 (aarch64_gen_load_pair): Rename calls to match new patterns.
3779 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
3780 (load_pair<DREG:mode><DREG2:mode>): ... This.
3781 (store_pair<mode>): Rename to...
3782 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
3783 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
3785 (V_INT_EQUIV): Handle SImode.
3786 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
3789 2018-05-22 Martin Sebor <msebor@redhat.com>
3792 * calls.c (maybe_warn_nonstring_arg): Use string length to set
3793 or ajust the presumed bound on an operation to avoid unnecessary
3796 2018-05-22 Martin Sebor <msebor@redhat.com>
3798 PR tree-optimization/85826
3799 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
3800 assuming that a DECL necesarily has a constant size.
3802 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
3805 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
3807 2018-05-22 Richard Biener <rguenther@suse.de>
3809 PR tree-optimization/85834
3810 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
3811 non-constant and non-zero memset arguments.
3813 2018-05-22 Martin Liska <mliska@suse.cz>
3816 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
3818 2018-05-22 Richard Biener <rguenther@suse.de>
3820 PR tree-optimization/85863
3821 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
3822 comparisons when vectype is specified.
3823 (vectorizable_condition): Do not specify vectype for
3824 vect_is_simple_cond when SLP vectorizing.
3826 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
3829 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
3830 define __ibm128 as long double.
3831 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
3832 as a distinct type when IEEE 128-bit support is enabled.
3833 (init_float128_ieee): Fix up conversions between IFmode and IEEE
3834 128-bit types to use the correct functions.
3835 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
3836 convert between 128-bit floating point types that have different
3837 modes but the same representation, instead of using gen_lowpart to
3839 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
3841 (IFKF_reg): New attributes to give the register constraints for
3843 (extend<mode>tf2_internal): New insns to mark an explicit
3844 conversion between 128-bit floating point types that have a
3845 different mode but share the same representation.
3847 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
3849 PR tree-optimization/85814
3850 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
3851 a null return from get_strinfo when unsharing the next
3852 strinfo in the chain.
3854 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3857 * varasm.c (weak_finish): Clean up weak_decls.
3859 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3861 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
3862 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
3863 UNSPEC_UADALP values.
3864 * config/aarch64/iterators.md (ABAL): New int iterator.
3867 (sur): Add mappings for the above.
3868 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
3870 (aarch64_<sur>abal<mode>_4): Likewise.
3871 (aarch64_<sur>adalp<mode>_3): Likewise.
3872 (<sur>sadv16qi): New define_expand.
3874 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
3876 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
3877 (*movdf_internal): Ditto.
3878 (*rcpsf2_sse): Ditto.
3879 (*rsqrtsf2_sse): Ditto.
3880 (*sqrt<mode>2_sse): Ditto.
3882 2018-05-21 Tamar Christina <tamar.christina@arm.com>
3884 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
3886 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
3887 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
3888 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
3889 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
3891 * config/aarch64/arm_neon.h: Likewise.
3892 * config/aarch64/iterators.md (VQ_I): New.
3894 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
3896 * config.gcc: Add arc/t-multilib-linux to tmake_file for
3898 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
3901 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3903 * config/nds32/constraints.md (S): New constraint.
3904 * config/nds32/nds32.md (call_internal): Use constraint S.
3905 (call_value_internal): Likewise.
3906 (sibcall_internal): Likewise.
3907 (sibcall_value_internal): Likewise.
3909 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3910 Chung-Ju Wu <jasonwucj@gmail.com>
3912 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
3915 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3916 Chung-Ju Wu <jasonwucj@gmail.com>
3918 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
3919 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
3920 (nds32_rtx_costs_impl): Simplify.
3921 (nds32_address_cost_impl): Simplify.
3922 (nds32_init_rtx_costs): New function.
3923 (nds32_rtx_costs_speed_prefer): Likewise.
3924 (nds32_rtx_costs_size_prefer): Likewise.
3925 (nds32_address_cost_speed_prefer): Likewise.
3926 (nds32_address_cost_speed_fwprop): Likewise.
3927 (nds32_address_cost_size_prefer): Likewise.
3928 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
3929 * config/nds32/nds32.c (nds32_option_override): Use
3930 nds32_init_rtx_costs function.
3932 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3934 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
3935 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
3936 (TARGET_PIPELINE_N8): Likewise.
3937 (TARGET_PIPELINE_N10): Likewise.
3938 (TARGET_PIPELINE_N13): Likewise.
3939 (TARGET_PIPELINE_GRAYWOLF): Likewise.
3941 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
3943 * config/nds32/nds32-fpu.md: Update copyright year.
3945 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3947 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
3949 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3951 * config/nds32/nds32.c
3952 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
3953 * config/nds32/nds32.opt (minline-asm-r15): New option.
3955 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3957 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
3959 * config/nds32/nds32.md (abssi2): New pattern.
3961 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
3963 * config/i386/i386.md (rex64namesuffix): New mode attribute.
3964 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
3965 Merge insn pattern from sse_cvtsi2ss<round_name> and
3966 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
3967 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
3968 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
3969 using SWI48 mode iterator.
3970 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
3971 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
3972 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
3973 pattern from sse_cvttss2si<round_saeonly_name>
3974 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
3975 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
3976 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
3977 using SWI48 mode iterator.
3978 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
3979 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
3980 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
3981 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
3982 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
3983 using SWI48 mode iterator.
3984 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
3985 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
3986 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
3987 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
3988 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
3989 SWI48 mode iterator.
3990 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
3991 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
3992 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
3993 pattern from sse_cvttsd2si<round_saeonly_name>
3994 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
3996 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3998 * config/nds32/nds32-md-auxiliary.c
3999 (nds32_valid_smw_lwm_base_p): Refine.
4000 (nds32_output_smw_single_word): Refine.
4001 (nds32_output_smw_double_word): New.
4002 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
4004 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4006 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
4007 (nds32_output_stack_pop): Refine.
4008 (nds32_expand_unaligned_load): Refine.
4009 (nds32_expand_unaligned_store): Refine.
4011 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4012 Chung-Ju Wu <jasonwucj@gmail.com>
4014 * config/nds32/constants.md: Add TP_REGNUM constant.
4015 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
4016 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
4018 * config/nds32/nds32-doubleword.md: Consider flag_pic.
4019 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
4020 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
4021 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
4022 and PIC code generation.
4023 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
4025 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
4027 * config/nds32/nds32.md: Support TLS and PIC.
4028 * config/nds32/nds32.c: Support TLS and PIC.
4029 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
4030 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
4033 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4035 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
4036 mode with E_ prefix.
4038 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4039 Chung-Ju Wu <jasonwucj@gmail.com>
4041 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
4042 * config/nds32/nds32-md-auxiliary.c
4043 (symbolic_reference_mentioned_p): New.
4044 (nds32_legitimize_ict_address): New.
4045 (nds32_expand_ict_move): New.
4046 (nds32_indirect_call_referenced_p): New.
4047 (nds32_symbol_binds_local_p): Delete.
4048 (nds32_long_call_p): Modify.
4049 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
4050 * config/nds32/nds32-protos.h
4051 (symbolic_reference_mentioned_p): Declare.
4052 (nds32_legitimize_ict_address): Declare.
4053 (nds32_expand_ict_move): Declare.
4054 (nds32_indirect_call_referenced_p): Declare.
4055 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
4056 (nds32_relax_group): Use nds32_ict_const_p as condition.
4057 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
4058 (nds32_asm_file_start): Output ict_model directive in asm code.
4059 (nds32_legitimate_address_p): Consider indirect call.
4060 (nds32_print_operand): Consider indirect call.
4061 (nds32_print_operand_address): Consider indirect call.
4062 (nds32_insert_attributes): Handle "indirect_call" attribute.
4063 (TARGET_LEGITIMATE_ADDRESS_P): Define.
4064 (TARGET_LEGITIMATE_CONSTANT_P): Define.
4065 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
4066 (TARGET_DELEGITIMIZE_ADDRESS): Define.
4067 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
4068 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
4069 (TARGET_ICT_MODEL_SMALL): Define.
4070 (TARGET_ICT_MODEL_LARGE): Define.
4071 * config/nds32/nds32.md (movsi): Consider ict model.
4072 (call, call_value): Consider ict model.
4073 (sibcall, sibcall_value): Consider ict model.
4074 * config/nds32/nds32.opt (mict-model): New option.
4075 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
4078 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
4079 Monk Chiang <sh.chiang04@gmail.com>
4080 Jim Wilson <jimw@sifive.com>
4082 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
4083 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
4084 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
4085 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
4086 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
4087 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
4088 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
4089 compute save_libcall_adjustment properly.
4090 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
4091 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
4092 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
4093 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
4094 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
4095 (ABI_SPEC): Handle mabi=ilp32e.
4096 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
4097 (RVE): Add RVE mask.
4098 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
4099 <-march>: Add rv32e as an example.
4101 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
4104 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
4105 (intra_create_variable_infos): Handle C++ constructors.
4107 2018-05-18 Martin Liska <mliska@suse.cz>
4109 * passes.def: Remove a redundant pass.
4111 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
4114 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
4116 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4118 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
4120 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
4122 (armv2, armv2a, armv3, armv3m): Delete architectures.
4123 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
4124 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
4125 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
4127 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
4128 (*mulsidi3adddi): Likewise.
4129 (mulsidi3): Likewise.
4130 (*mulsidi3_nov6): Likewise.
4131 (umulsidi3): Likewise.
4132 (umulsidi3_nov6): Likewise.
4133 (umaddsidi4): Likewise.
4134 (*umulsidi3adddi): Likewise.
4135 (smulsi3_highpart): Likewise.
4136 (*smulsi3_highpart_nov6): Likewise.
4137 (umulsi3_highpart): Likewise.
4138 (*umulsi3_highpart_nov6): Likewise.
4139 * config/arm/arm.h (arm_arch3m): Delete.
4140 * config/arm/arm.c (arm_arch3m): Delete.
4141 (arm_option_override_internal): Update armv3-related comment.
4142 (arm_configure_build_target): Delete use of isa_bit_mode32.
4143 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
4144 (arm_rtx_costs_internal): Delete check of arm_arch3m.
4145 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
4147 (mulusa3): Likewise.
4148 * config/arm/arm-protos.h (arm_arch3m): Delete.
4149 * config/arm/arm-tables.opt: Regenerate.
4150 * config/arm/arm-tune.md: Likewise.
4151 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
4152 deleted architectures.
4154 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4156 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
4157 (armv5t, armv5te): New features.
4158 (ARMv5, ARMv5e): Delete fgroups.
4159 (ARMv5t, ARMv5te): Adjust for above changes.
4161 (armv5, armv5e): Delete arches.
4162 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
4164 (*call_reg_arm): Likewise.
4165 (*call_value_reg_armv5): Likewise.
4166 (*call_value_reg_arm): Likewise.
4167 (*call_symbol): Likewise.
4168 (*call_value_symbol): Likewise.
4169 (*sibcall_insn): Likewise.
4170 (*sibcall_value_insn): Likewise.
4172 (prefetch): Likewise.
4173 (define_split and define_peephole2 dependent on arm_arch5):
4175 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
4177 (TARGET_ARM_QBIT): Likewise.
4178 (TARGET_DSP_MULTIPLY): Likewise.
4179 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
4180 (arm_arch5, arm_arch5e): Delete.
4181 (arm_arch5t, arm_arch5te): Declare.
4182 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
4183 (arm_arch5t): Declare.
4184 (arm_option_reconfigure_globals): Update for the above.
4185 (arm_options_perform_arch_sanity_checks): Update comment, replace
4186 use of arm_arch5 with arm_arch5t.
4187 (use_return_insn): Likewise.
4188 (arm_emit_call_insn): Likewise.
4189 (output_return_instruction): Likewise.
4190 (arm_final_prescan_insn): Likewise.
4191 (arm_coproc_builtin_available): Likewise.
4192 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
4193 arm_arch5e with arm_arch5t and arm_arch5te.
4194 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
4195 (arm_arch5t, arm_arch5te): Declare.
4196 * config/arm/arm-tables.opt: Regenerate.
4197 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
4198 * config/arm/t-multilib: Likewise.
4199 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
4200 instead of arm_arch5.
4201 (*call_reg_thumb1): Likewise.
4202 (*call_value_reg_thumb1_v5): Likewise.
4203 (*call_value_reg_thumb1): Likewise.
4204 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
4206 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
4208 2018-05-18 Martin Liska <mliska@suse.cz>
4210 PR gcov-profile/84846
4211 * doc/gcov.texi: Document -t option of gcov tool.
4213 2018-05-18 Martin Liska <mliska@suse.cz>
4215 PR gcov-profile/84846
4216 * gcov.c (print_usage): Add new -t option.
4217 (process_args): Handle the option.
4218 (generate_results): Use stdout as output when requested by
4221 2018-05-18 Martin Liska <mliska@suse.cz>
4223 PR gcov-profile/84846
4224 * coverage.c (coverage_init): Write PWD to .gcno file.
4225 * doc/gcov.texi: Document how working directory is printed.
4226 * gcov-dump.c (dump_gcov_file): Print PWD.
4227 * gcov.c (output_intermediate_file): Likewise.
4228 (read_graph_file): Read PWD string.
4229 (output_lines): Print PWD.
4231 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4234 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
4235 for retval and return false if all args to phi are zero.
4237 2018-05-18 Richard Biener <rguenther@suse.de>
4239 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
4241 (evrp_dom_walker::before_dom_children): Call it.
4243 2018-05-18 Richard Biener <rguenther@suse.de>
4245 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
4246 results when processing array refs with variable index.
4248 2018-05-18 Toon Moene <toon@moene.org>
4250 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
4251 directly after that of -floop-interchange. Indicate that both
4252 options are enabled by default when specifying -O3.
4254 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4256 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
4257 iterator. Delete separate integer-mode vec_set<mode> expander.
4258 (aarch64_simd_vec_setv2di): Delete.
4259 (vec_setv2di): Delete.
4260 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
4261 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
4262 the "w, r" alternative.
4264 2018-05-18 Martin Liska <mliska@suse.cz>
4266 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
4267 * tree-pass.h (make_pass_lower_switch_O0): New function.
4268 * tree-switch-conversion.c (node_has_low_bound): Remove.
4269 (node_has_high_bound): Likewise.
4270 (node_is_bounded): Likewise.
4271 (class pass_lower_switch): Make it a template type and create
4273 (pass_lower_switch::execute): Add template argument.
4274 (make_pass_lower_switch): New function.
4275 (make_pass_lower_switch_O0): New function.
4276 (do_jump_if_equal): Remove.
4277 (emit_case_nodes): Simplify to just handle all 3 cases and leave
4278 all the hard work to tree optimization passes.
4280 2018-05-18 Martin Liska <mliska@suse.cz>
4282 * dbgcnt.c (limit_low): Renamed from limit.
4283 (limit_high): New variable.
4284 (dbg_cnt_is_enabled): Check for upper limit.
4285 (dbg_cnt): Adjust dumping.
4286 (dbg_cnt_set_limit_by_index): Add new argument for high
4288 (dbg_cnt_set_limit_by_name): Likewise.
4289 (dbg_cnt_process_single_pair): Parse new format.
4290 (dbg_cnt_process_opt): Use strtok.
4291 (dbg_cnt_list_all_counters): Remove 'value' and add
4293 * doc/invoke.texi: Document changes.
4295 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
4297 * doc/sourcebuild.texi (scalar_all_fma): Document.
4298 * tree.def (FMA_EXPR): Delete.
4299 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
4300 * internal-fn.c (ternary_direct): New macro.
4301 (expand_ternary_optab_fn): Likewise.
4302 (direct_ternary_optab_supported_p): Likewise.
4303 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
4304 * builtins.c (fold_builtin_fma): Delete.
4305 (fold_builtin_3): Don't call it.
4306 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
4307 * expr.c (expand_expr_real_2): Likewise.
4308 * fold-const.c (operand_equal_p): Likewise.
4309 (fold_ternary_loc): Likewise.
4310 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
4311 * gimple.c (DEFTREECODE): Likewise.
4312 * gimplify.c (gimplify_expr): Likewise.
4313 * optabs-tree.c (optab_for_tree_code): Likewise.
4314 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
4315 * tree-eh.c (operation_could_trap_p): Likewise.
4316 (stmt_could_throw_1_p): Likewise.
4317 * tree-inline.c (estimate_operator_cost): Likewise.
4318 * tree-pretty-print.c (dump_generic_node): Likewise.
4319 (op_code_prio): Likewise.
4320 * tree-ssa-loop-im.c (stmt_cost): Likewise.
4321 * tree-ssa-operands.c (get_expr_operands): Likewise.
4322 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
4323 * fold-const-call.h (fold_fma): Delete.
4324 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
4325 CFN_FNMA and CFN_FNMS.
4327 * genmatch.c (combined_fn): New enum.
4328 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
4329 (commutative_op): New function.
4330 (commutate): Use it. Handle more than 2 operands.
4331 (dt_operand::gen_gimple_expr): Use commutative_op.
4332 (parser::parse_expr): Allow :c to be used with non-binary
4333 operators if the commutative operand is known.
4334 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
4335 CFN_FMS, CFN_FNMA and CFN_FNMS.
4336 (backprop::process_assign_use): Remove FMA_EXPR handling.
4337 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
4338 (gen_hsa_fma): New function.
4339 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
4340 IFN_FNMA and IFN_FNMS.
4341 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
4342 * gimple-fold.h (follow_all_ssa_edges): Declare.
4343 * gimple-fold.c (follow_all_ssa_edges): New function.
4344 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
4345 gimple_build interface and use follow_all_ssa_edges to fold the result.
4346 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
4347 instead of checking for optabs directly.
4348 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
4349 rather than FMA_EXPRs.
4350 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
4351 call to IFN_FMA instead of an FMA_EXPR.
4353 2018-05-17 Jim Wilson <jimw@sifive.com>
4355 * expr.c (do_tablejump): When converting index to Pmode, if we have a
4356 sign extended promoted subreg, and the range does not have the sign bit
4357 set, then do a sign extend.
4359 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
4360 test, check for sign extended subreg and/or constant operands, and
4361 do a sign extend in that case.
4363 2018-05-17 Steve Ellcey <sellcey@cavium.com>
4365 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
4366 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
4368 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
4369 Change logics_shift_reg to logics_shift_imm.
4370 (thunderx2t99_fp_loadpair_basic): Delete.
4371 (thunderx2t99_fp_storepair_basic): Delete.
4372 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
4373 (thunderx2t99_asimd_polynomial): Delete.
4374 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
4375 and neon_fp_mul_d_scalar_q.
4376 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
4377 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
4378 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
4379 (thunderx2t99_asimd_lut): Add missing tbl types.
4380 (thunderx2t99_asimd_ext): Delete.
4381 (thunderx2t99_asimd_load1_1_mult): Delete.
4382 (thunderx2t99_asimd_load1_2_mult): Delete.
4383 (thunderx2t99_asimd_load1_ldp): New.
4384 (thunderx2t99_asimd_load1): New.
4385 (thunderx2t99_asimd_load2): Add missing *load2* types.
4386 (thunderx2t99_asimd_load3): New.
4387 (thunderx2t99_asimd_load4): New.
4388 (thunderx2t99_asimd_store1_1_mult): Delete.
4389 (thunderx2t99_asimd_store1_2_mult): Delete.
4390 (thunderx2t99_asimd_store2_mult): Delete.
4391 (thunderx2t99_asimd_store2_onelane): Delete.
4392 (thunderx2t99_asimd_store_stp): New.
4393 (thunderx2t99_asimd_store1): New.
4394 (thunderx2t99_asimd_store2): New.
4395 (thunderx2t99_asimd_store3): New.
4396 (thunderx2t99_asimd_store4): New.
4398 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
4400 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
4401 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
4403 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
4404 Segher Boessenkool <segher@kernel.crashing.org>
4407 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
4410 2018-05-17 Richard Biener <rguenther@suse.de>
4412 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
4413 for pruning loop and prune defs feeding only already visited PHIs.
4415 2018-05-17 Richard Biener <rguenther@suse.de>
4417 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
4419 2018-05-17 Bin Cheng <bin.cheng@arm.com>
4420 Richard Biener <rguenther@suse.de>
4422 PR tree-optimization/85793
4423 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
4424 for VMAT_ELEMENTWISE.
4426 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4428 * internal-fn.h (lookup_internal_fn): Declare
4429 * internal-fn.c (lookup_internal_fn): New function.
4430 * gimple.c (gimple_build_call_from_tree): Handle calls to
4432 * gimple-pretty-print.c (dump_gimple_call): Print "." before
4433 internal function names.
4434 * tree-pretty-print.c (dump_generic_node): Likewise.
4435 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
4437 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4439 * gimple-fold.h (gimple_build): Make the function forms take
4440 combined_fn rather than built_in_function.
4441 (gimple_simplify): Likewise.
4442 * gimple-match-head.c (gimple_simplify): Likewise.
4443 * gimple-fold.c (gimple_build): Likewise.
4444 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
4445 rather than gimple_build_call_internal.
4446 (get_initial_defs_for_reduction): Likewise.
4447 (vect_create_epilog_for_reduction): Likewise.
4448 (vectorizable_live_operation): Likewise.
4450 2018-05-17 Martin Liska <mliska@suse.cz>
4452 * gimple-ssa-sprintf.c (format_directive): Do not use
4453 space in between 'G_' and '('.
4455 2018-05-17 Jakub Jelinek <jakub@redhat.com>
4458 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
4459 even if the mask is not all ones.
4462 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
4464 (ix86_gimple_fold_builtin): Likewise.
4467 * config/i386/i386.c: Include tree-vector-builder.h.
4468 (ix86_vector_shift_count): New function.
4469 (ix86_fold_builtin): Fold shift builtins by scalar count.
4470 (ix86_gimple_fold_builtin): Likewise.
4472 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
4473 _mm512_setzero): New intrinsics.
4475 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
4476 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4478 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
4479 code generation for cases where splatting a value is not useful.
4480 * simplify-rtx.c (simplify_ternary_operation): Simplify
4481 vec_merge across a vec_duplicate and a paradoxical subreg forming
4482 a vector mode to a vec_concat.
4484 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
4486 * config.gcc: Support "goldmont-plus".
4487 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4489 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4490 PROCESSOR_GOLDMONT_PLUS.
4491 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
4492 (processor_target_table): Add "goldmont-plus".
4493 (PTA_GOLDMONT_PLUS): Define.
4494 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
4495 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
4496 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
4497 (fold_builtin_cpu): Add "goldmont-plus".
4498 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
4499 (ix86_option_override_internal): Add "goldmont-plus".
4500 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
4501 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
4502 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
4503 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
4505 2018-05-17 Richard Biener <rguenther@suse.de>
4507 PR tree-optimization/85757
4508 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
4509 remove defs that only feed that PHI from further processing.
4511 2018-05-16 Jim Wilson <jimw@sifive.com>
4513 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
4515 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
4517 2018-05-16 Mark Wielaard <mark@klomp.org>
4519 * dwarf2out.c (count_index_strings): New function.
4520 (output_indirect_strings): Call count_index_strings and generate
4521 header for dwarf_version >= 5.
4523 2018-05-16 Mark Wielaard <mark@klomp.org>
4525 * dwarf2out.c (dwarf_FORM): New function.
4526 (set_indirect_string): Use dwarf_FORM.
4527 (reset_indirect_string): Likewise.
4528 (size_of_die): Likewise.
4529 (value_format): Likewise.
4530 (output_die): Likewise.
4531 (add_skeleton_AT_string): Likewise.
4532 (output_macinfo_op): Likewise.
4533 (index_string): Likewise.
4534 (output_index_string_offset): Likewise.
4535 (output_index_string): Likewise.
4536 (count_index_strings): Likewise.
4538 2018-05-16 Carl Love <cel@us.ibm.com>
4540 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4541 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
4543 2018-05-16 Martin Jambor <mjambor@suse.cz>
4545 * ipa-prop.c (ipa_free_all_edge_args): Remove.
4546 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
4548 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
4550 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
4551 (fnma<mode>4): Likewise.
4552 (fms<mode>4): Likewise.
4553 (fnms<mode>4): Likewise.
4554 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
4555 (aarch64_fnma<mode>4): Likewise.
4556 (aarch64_fms<mode>4): Likewise.
4557 (aarch64_fnms<mode>4): Likewise.
4558 (aarch64_fnmadd<mode>4): Likewise.
4560 2018-05-16 Jason Merrill <jason@redhat.com>
4562 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
4564 2018-05-16 Richard Biener <rguenther@suse.de>
4566 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
4567 (dump_stmt_cost): Declare.
4568 (add_stmt_cost): Dump cost we add.
4569 (add_stmt_costs): New function.
4570 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
4572 (vect_analyze_stmt): Adjust prototype.
4573 (vectorizable_condition): Likewise.
4574 (vectorizable_live_operation): Likewise.
4575 (vectorizable_reduction): Likewise.
4576 (vectorizable_induction): Likewise.
4577 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
4578 cost vector to pass to vectorizable_ and record afterwards.
4579 (vect_model_reduction_cost): Take cost vector argument and adjust.
4580 (vect_model_induction_cost): Likewise.
4581 (vectorizable_reduction): Likewise.
4582 (vectorizable_induction): Likewise.
4583 (vectorizable_live_operation): Likewise.
4584 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
4585 SLP_TREE_NUMBER_OF_VEC_STMTS.
4586 (vect_analyze_slp_cost_1): Remove.
4587 (vect_analyze_slp_cost): Likewise.
4588 (vect_slp_analyze_node_operations): Take visited args and
4589 a target cost vector. Avoid processing already visited stmt sets.
4590 (vect_slp_analyze_operations): Use a local cost vector to gather
4591 costs and register those of non-discarded instances.
4592 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
4593 (vect_schedule_slp_instance): Remove copying of
4594 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
4596 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
4597 adding cost. Record cost entry location.
4598 (vect_prologue_cost_for_slp_op): Function to compute cost of
4599 a constant or invariant generated for SLP vect in the prologue,
4600 split out from vect_analyze_slp_cost_1.
4601 (vect_model_simple_cost): Make static. Adjust for SLP costing.
4602 (vect_model_promotion_demotion_cost): Likewise.
4603 (vect_model_store_cost): Likewise, make static.
4604 (vect_model_load_cost): Likewise.
4605 (vectorizable_bswap): Add cost vector arg and adjust.
4606 (vectorizable_call): Likewise.
4607 (vectorizable_simd_clone_call): Likewise.
4608 (vectorizable_conversion): Likewise.
4609 (vectorizable_assignment): Likewise.
4610 (vectorizable_shift): Likewise.
4611 (vectorizable_operation): Likewise.
4612 (vectorizable_store): Likewise.
4613 (vectorizable_load): Likewise.
4614 (vectorizable_condition): Likewise.
4615 (vectorizable_comparison): Likewise.
4616 (can_vectorize_live_stmts): Likewise.
4617 (vect_analyze_stmt): Likewise.
4618 (vect_transform_stmt): Adjust calls to vectorizable_*.
4619 * tree-vectorizer.c: Include gimple-pretty-print.h.
4620 (dump_stmt_cost): New function.
4622 2018-05-16 Richard Biener <rguenther@suse.de>
4624 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
4625 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
4626 * tree-ssa-dse.c: Include tree-ssa-loop.h.
4627 (check_name): New callback.
4628 (dse_classify_store): Track cycles via a visited bitmap of PHI
4629 defs and simplify handling of in-loop and across loop dead stores
4630 and properly fail for loop-variant refs. Handle byte-tracking with
4631 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
4634 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
4636 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
4637 (vect_get_mask_type_for_stmt): Likewise.
4638 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
4640 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
4641 to determine the statement's vector type and the vector type that
4642 should be used for calculating nunits. Deal with cases in which
4643 the type has to be deferred.
4644 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
4645 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
4646 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
4647 (vect_determine_vf_for_stmt): New functions, split out from...
4648 (vect_determine_vectorization_factor): ...here.
4649 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
4650 (vect_get_mask_type_for_stmt): New functions, split out from
4651 vect_determine_vectorization_factor.
4653 2018-05-16 Richard Biener <rguenther@suse.de>
4655 * tree-cfg.c (verify_gimple_assign_ternary): Properly
4656 verify the [VEC_]COND_EXPR embedded comparison.
4658 2018-05-15 Martin Sebor <msebor@redhat.com>
4660 PR tree-optimization/85753
4661 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
4662 RECORD_TYPE in addition to ARRAY_TYPE.
4664 2018-05-15 Martin Sebor <msebor@redhat.com>
4667 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
4669 2018-05-15 Richard Biener <rguenther@suse.de>
4671 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
4672 add by_clobber_p one. Change algorithm to collect all defs
4673 representing uses we need to walk and try reducing them to
4674 a single one before failing.
4675 (dse_dom_walker::dse_optimize_stmt): Adjust.
4677 2018-05-13 Mark Wielaard <mark@klomp.org>
4679 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
4680 (size_of_loc_descr): Likewise.
4681 (output_loc_operands): Likewise.
4682 (output_loc_operands_raw): Likewise.
4683 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
4684 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
4685 (hash_loc_operands): Likewise.
4686 (compare_loc_operands): Likewise.
4688 2018-05-14 Mark Wielaard <mark@klomp.org>
4690 * dwarf2out.c (count_index_addrs): New function.
4691 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
4693 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4695 PR tree-optimization/83648
4696 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
4697 return value as malloc candidate.
4699 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4702 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
4703 param as true in call to suggest_attribute.
4705 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
4707 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
4708 -mreadonly-in-sdata.
4710 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4712 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
4714 (aarch64_crypto_aesd_fused): Likewise.
4716 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
4718 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
4719 (movsi_aarch64): Likewise.
4720 (load_pairsi): Likewise.
4721 (load_pairdi): Likewise.
4722 (store_pairsi): Likewise.
4723 (store_pairdi): Likewise.
4724 (load_pairsf): Likewise.
4725 (load_pairdf): Likewise.
4726 (store_pairsf): Likewise.
4727 (store_pairdf): Likewise.
4728 (zero_extend): Likewise.
4729 (trunc): Swap alternatives.
4730 (fcvt_target): Add '?' to prefer w over r.
4732 2018-05-14 Jakub Jelinek <jakub@redhat.com>
4735 * config/i386/i386.md: Disallow non-commutative arithmetics in
4736 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
4737 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
4738 in the peephole2 before it.
4740 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
4742 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
4743 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
4744 (ix86_handle_option): Handle -mcldemote.
4745 * config.gcc: New header.
4746 * config/i386/cldemoteintrin.h: New file.
4747 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
4748 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4750 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4751 OPTION_MASK_ISA_CLDEMOTE.
4752 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
4753 (ix86_valid_target_attribute_inner_p): Ditto.
4754 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
4755 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
4756 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
4757 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
4758 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
4760 * config/i386/i386.opt: Add -mcldemote.
4761 * config/i386/x86intrin.h: New header.
4762 * doc/invoke.texi: Add -mcldemote.
4764 2018-05-14 Richard Biener <rguenther@suse.de>
4766 * doc/match-and-simplify.texi: Adjust :s documentation.
4768 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
4770 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
4771 intended memcpy size.
4772 (REORDER_45): Likewise.
4774 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
4776 * sort.cc: New file.
4777 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
4778 * vec.c (qsort_chk): Use gcc_qsort.
4779 * Makefile.in (OBJS-libcommon): Add sort.o.
4780 (build/sort.o): New target. Use it...
4781 (BUILD_RTL): ... here, and...
4782 (build/gencfn-macros): ... here, and...
4783 (build/genmatch): ... here.
4785 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4786 Chung-Ju Wu <jasonwucj@gmail.com>
4788 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
4789 * config/nds32/nds32-graywolf.md: New file.
4790 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
4791 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
4793 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
4794 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
4795 * config/nds32/nds32.md (pipeline_model): Add graywolf.
4796 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
4797 * config/nds32/pipelines.md: Include n15 settings.
4799 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4800 Chung-Ju Wu <jasonwucj@gmail.com>
4802 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
4803 * config/nds32/nds32-n13.md: New file.
4804 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
4805 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
4807 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
4808 * config/nds32/nds32.md (pipeline_model): Add n13.
4809 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
4810 * config/nds32/pipelines.md: Include n13 settings.
4812 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4813 Chung-Ju Wu <jasonwucj@gmail.com>
4815 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
4816 * config/nds32/nds32-n10.md: New file.
4817 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
4818 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
4820 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
4821 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
4822 * config/nds32/nds32.md (pipeline_model): Add n10.
4823 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
4824 * config/nds32/pipelines.md: Include n10 settings.
4826 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
4827 Kito Cheng <kito.cheng@gmail.com>
4828 Chung-Ju Wu <jasonwucj@gmail.com>
4830 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
4831 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4832 Add enum values for DSP extension instructions.
4833 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
4835 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
4836 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
4838 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
4839 * config/nds32/nds32-dspext.md: New file for DSP implementation.
4840 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
4841 * config/nds32/nds32-intrinsic.md: Likewise.
4842 * config/nds32/nds32_intrinsic.h: Likewise.
4843 * config/nds32/nds32-md-auxiliary.c: Likewise.
4844 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
4845 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
4846 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
4847 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
4848 * config/nds32/nds32-protos.h: New declarations for DSP extension.
4849 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
4850 TYPE_DMAC in switch statement.
4851 * config/nds32/nds32.c: New checking and implementation for DSP
4852 extension instructions.
4853 * config/nds32/nds32.h: Likewise.
4854 * config/nds32/nds32.md: Likewise.
4855 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
4856 * config/nds32/predicates.md: Implement new predicates for DSP
4859 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4861 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
4862 Reformat alternatives and attributes so it is easier to identify
4863 which constraints/attributes go with which instruction.
4864 (mov<mode>_hardfloat32, FMOVE64): Likewise.
4865 (mov<mode>_softfloat32, FMOVE64): Likewise.
4866 (mov<mode>_hardfloat64, FMOVE64): Likewise.
4867 (mov<mode>_softfloat64, FMOVE64): Likewise.
4869 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
4871 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4873 (Basic PowerPC Built-in Functions): The new name of the
4874 subsection previously known as "PowerPC Built-in Functions".
4875 (Basic PowerPC Built-in Functions Available on all Configurations):
4877 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
4878 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
4879 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
4880 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
4882 2018-05-11 Martin Jambor <mjambor@suse.cz>
4885 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
4888 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4891 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
4893 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
4895 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
4896 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
4897 (ix86_handle_option): Handle -mwaitpkg.
4898 * config.gcc: New header.
4899 * config/i386/cpuid.h (bit_WAITPKG): New bit.
4900 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
4901 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
4903 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4904 OPTION_MASK_ISA_WAITPKG.
4905 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
4906 (ix86_option_override_internal): Add PTA_WAITPKG.
4907 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
4908 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
4909 IX86_BUILTIN_TPAUSE.
4910 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
4911 __builtin_ia32_umwait and __builtin_ia32_tpause.
4912 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
4913 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
4914 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
4915 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
4916 UNSPECV_TPAUSE): New.
4917 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
4918 * config/i386/i386.opt: Add -mwaitpkg.
4919 * config/i386/waitpkgintrin.h: New file.
4920 * config/i386/x86intrin.h: New header.
4921 * doc/invoke.texi: Add -mwaitpkg.
4923 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4926 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
4928 (cortex-m0): Use armv6s-m isa.
4929 (cortex-m0plus): Likewise.
4930 (cortex-m1): Likewise.
4931 (cortex-m0.small-multiply): Likewise.
4932 (cortex-m0plus.small-multiply): Likewise.
4933 (cortex-m1.small-multiply): Likewise.
4935 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
4936 Jakub Jelinek <jakub@redhat.com>
4938 PR tree-optimization/85692
4939 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
4940 source permute as well.
4942 2018-05-11 Martin Liska <mliska@suse.cz>
4945 * doc/extend.texi: Document LLVM style format for no_sanitize
4948 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
4950 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
4951 mode_supports_vsx_dform_quad to mode_supports_dq_form.
4952 (mode_supports_vsx_dform_quad): Likewise.
4953 (mode_supports_vmx_dform): Move these functions to be next to the
4954 other mode_supports functions.
4955 (mode_supports_dq_form): Likewise.
4956 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
4957 mode_supports_dq_form.
4958 (reg_offset_addressing_ok_p): Likewise.
4959 (offsettable_ok_by_alignment): Likewise.
4960 (rs6000_legitimate_offset_address_p): Likewise.
4961 (legitimate_lo_sum_address_p): Likewise.
4962 (rs6000_legitimize_address): Likewise.
4963 (rs6000_legitimize_reload_address): Likewise.
4964 (rs6000_secondary_reload_inner): Likewise.
4965 (rs6000_preferred_reload_class): Likewise.
4966 (rs6000_output_move_128bit): Likewise.
4968 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4970 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
4971 Generate SImode target register for null target.
4972 <case IX86_BUILTIN_XGETBV>: Ditto.
4973 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
4974 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
4976 2018-05-10 Carl Love <cel@us.ibm.com>
4978 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4979 dcbtt and dcbtstt if operands[2] is 0.
4981 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4984 * config/i386/sse.md (usadv64qi): New expander.
4986 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
4988 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
4989 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
4990 -maltivec=be support.
4991 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
4992 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
4993 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
4994 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
4995 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
4996 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
4997 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
4998 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
4999 altivec_vsumsws): Adjust.
5000 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
5001 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
5002 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
5004 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
5005 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
5006 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
5007 (altivec_lve<VI_char>x): Delete expand.
5008 (*altivec_lve<VI_char>x_internal): Rename to...
5009 (altivec_lve<VI_char>x): ... this.
5010 (altivec_lvxl_<mode>): Delete expand.
5011 (*altivec_lvxl_<mode>_internal): Rename to ...
5012 (altivec_lvxl_<mode>): ... this.
5013 (altivec_stvxl_<mode>): Delete expand.
5014 (*altivec_stvxl_<mode>_internal): Rename to ...
5015 (altivec_stvxl_<mode>): ... this.
5016 (altivec_stve<VI_char>x): Delete expand.
5017 (*altivec_stve<VI_char>x_internal): Rename to ...
5018 (altivec_stve<VI_char>x): ... this.
5019 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
5020 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
5021 reduc_plus_scal_<mode>): Adjust.
5022 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
5024 (rs6000_cpu_cpp_builtins): Adjust.
5025 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
5026 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
5027 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
5028 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
5029 -maltivec=be support.
5030 (rs6000_split_vec_extract_var): Adjust.
5031 (rs6000_split_v4si_init): Adjust.
5032 (swap_selector_for_mode): Delete.
5033 (altivec_expand_lvx_be, altivec_expand_stvx_be,
5034 altivec_expand_stvex_be): Delete.
5035 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
5036 -maltivec=be support.
5037 (rs6000_gimple_fold_builtin): Ditto.
5038 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
5040 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
5041 (TARGET_DIRECT_MOVE_64BIT): Adjust.
5042 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
5043 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
5044 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
5045 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
5046 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
5047 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
5048 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
5049 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
5050 anonymous split): Adjust.
5051 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
5052 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
5054 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
5056 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
5057 when --with-gxx-include-dir is also specified.
5058 * configure: Regenerate.
5060 2018-05-09 Jim Wilson <jimw@sifive.com>
5063 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
5064 * config/riscv/t-withmultilib: New.
5065 * config/riscv/withmultilib.h: New.
5066 * doc/install.texi: Document RISC-V --with-multilib-list support.
5068 2018-05-09 Richard Biener <rguenther@suse.de>
5070 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
5072 (vect_bb_vectorization_profitable_p): Adjust. Compute
5073 actual scalar cost using the cost vector and the add_stmt_cost
5076 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5078 PR rtl-optimization/85645
5079 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
5080 in the REG_CFA_REGISTER note for LR, don't leave it empty.
5082 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5084 PR rtl-optimization/85645
5085 * shrink-wrap.c (spread_components): Return a boolean saying if
5086 anything was changed.
5087 (try_shrink_wrapping_separate): Iterate spread_components until
5088 nothing changes anymore.
5090 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5092 PR rtl-optimization/85645
5093 * regrename.c (build_def_use): Also kill the chains that include the
5094 destination of a REG_CFA_REGISTER note.
5096 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5098 PR rtl-optimization/85645
5099 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
5100 insn that has a REG_CFA_REGISTER note.
5102 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
5104 * cfgexpand.c (expand_clobber): New function.
5105 (expand_gimple_stmt_1): Use it.
5106 * tree-vect-stmts.c (vect_clobber_variable): New function,
5108 (vectorizable_simd_clone_call): ...here.
5109 (vectorizable_store): Emit a clobber either side of an
5110 IFN_STORE_LANES sequence.
5111 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
5113 2018-05-09 Tom de Vries <tom@codesourcery.com>
5116 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
5117 (define_insn "trap_if_false"): Add exit after trap.
5119 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
5121 PR rtl-optimization/85638
5122 * bb-reorder.c: Include common/common-target.h.
5123 (create_forwarder_block): New function extracted from...
5124 (fix_up_crossing_landing_pad): ...here. Rename into...
5125 (dw2_fix_up_crossing_landing_pad): ...this.
5126 (sjlj_fix_up_crossing_landing_pad): New function.
5127 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
5128 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
5129 from both partitions and exit the loop after one iteration.
5131 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5134 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5136 (Basic PowerPC Built-in Functions): The new name of the
5137 subsection previously known as "PowerPC Built-in Functions".
5138 (Basic PowerPC Built-in Functions Available on all Configurations):
5140 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5142 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5144 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5146 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5149 2018-05-08 Jim Wilson <jimw@sifive.com>
5151 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
5152 (LD_EMUL_SUFFIX): New.
5153 (LINK_SPEC): Use it.
5155 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5157 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5159 (Basic PowerPC Built-in Functions): The new name of the
5160 subsection previously known as "PowerPC Built-in Functions".
5161 (Basic PowerPC Built-in Functions Available on all Configurations):
5163 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5165 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5167 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5169 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5172 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5175 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
5176 after cmpelim optimization.
5178 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
5180 * config.gcc: Support "goldmont".
5181 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
5182 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5184 * config/i386/i386.c (m_GOLDMONT): Define.
5185 (processor_target_table): Add "goldmont".
5186 (PTA_GOLDMONT): Define.
5187 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
5188 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
5189 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
5190 (fold_builtin_cpu): Add "goldmont".
5191 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
5192 (ix86_option_override_internal): Add "goldmont".
5193 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
5194 (processor_type): Add PROCESSOR_GOLDMONT.
5195 * config/i386/i386.md: Add CPU "glm".
5196 * config/i386/glm.md: New file.
5197 * config/i386/x86-tune.def: Add m_GOLDMONT.
5198 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
5200 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5203 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
5205 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
5206 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
5207 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
5210 * config/i386/i386.c (ix86_fold_builtin): Handle
5211 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
5214 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
5215 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
5217 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
5220 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
5221 (check_arch): Likewise.
5222 (check_fpu): Return the result rather than printing it.
5223 (end arch): Fix operator precedence.
5224 (end cpu): Likewise.
5225 (END): Print the result from check_fpu.
5227 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5228 Alan Hayward <alan.hayward@arm.com>
5229 David Sherwood <david.sherwood@arm.com>
5231 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
5232 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
5233 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
5234 (*fcmuo<mode>_and): New patterns.
5236 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5238 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
5239 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
5240 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
5241 (cmp_op, sve_imm_con): New code attributes.
5242 (SVE_COND_INT_CMP, imm_con): Delete.
5243 (cmp_op): Remove above unspecs from int attribute.
5244 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
5246 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
5247 comparison-specific unspecs.
5248 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
5249 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
5250 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
5251 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
5252 (*vec_fcm<cmp_op><mode>): Rename to...
5253 (*fcm<cmp_op><mode>): ...this and adjust likewise.
5254 (*vec_fcmuo<mode>): Rename to...
5255 (*fcmuo<mode>): ...this and adjust likewise.
5256 (*pred_fcm<cmp_op><mode>): New pattern.
5257 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
5258 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
5260 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
5262 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
5263 (aarch64_emit_sve_predicated_cond): New function.
5264 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
5265 (aarch64_emit_unspec_cond_or): Replace with...
5266 (aarch64_emit_sve_or_conds): ...this new function. Use
5267 aarch64_emit_sve_ptrue_op for the individual comparisons and
5268 aarch64_emit_binop to OR them together.
5269 (aarch64_emit_inverted_unspec_cond): Replace with...
5270 (aarch64_emit_sve_inverted_cond): ...this new function. Use
5271 aarch64_emit_sve_ptrue_op for the comparison and
5272 aarch64_emit_unop to invert the result.
5273 (aarch64_expand_sve_vec_cmp_float): Update after the above
5274 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
5276 2018-05-07 Nathan Sidwell <nathan@acm.org>
5278 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
5279 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
5280 (Backwards Compatibility): Likewise.
5282 2018-05-07 Luis Machado <luis.machado@linaro.org>
5286 2018-05-07 Luis Machado <luis.machado@linaro.org>
5288 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5289 <prefetch_dynamic_strides>: New const bool field.
5290 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5291 prefetch_dynamic_strides.
5292 (exynosm1_prefetch_tune): Likewise.
5293 (thunderxt88_prefetch_tune): Likewise.
5294 (thunderx_prefetch_tune): Likewise.
5295 (thunderx2t99_prefetch_tune): Likewise.
5296 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5298 (aarch64_override_options_internal): Update to set
5299 PARAM_PREFETCH_DYNAMIC_STRIDES.
5300 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5301 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5302 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5303 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5304 prefetch-dynamic-strides setting.
5306 2018-05-07 Luis Machado <luis.machado@linaro.org>
5308 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5309 <minimum_stride>: New const int field.
5310 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5311 minimum_stride field.
5312 (exynosm1_prefetch_tune): Likewise.
5313 (thunderxt88_prefetch_tune): Likewise.
5314 (thunderx_prefetch_tune): Likewise.
5315 (thunderx2t99_prefetch_tune): Likewise.
5316 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5317 (aarch64_override_options_internal): Update to set
5318 PARAM_PREFETCH_MINIMUM_STRIDE.
5319 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5320 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5321 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5322 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5323 stride is constant and is below the minimum stride threshold.
5325 2018-05-07 Luis Machado <luis.machado@linaro.org>
5327 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
5330 2018-05-07 Luis Machado <luis.machado@linaro.org>
5332 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5333 <prefetch_dynamic_strides>: New const bool field.
5334 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5335 prefetch_dynamic_strides.
5336 (exynosm1_prefetch_tune): Likewise.
5337 (thunderxt88_prefetch_tune): Likewise.
5338 (thunderx_prefetch_tune): Likewise.
5339 (thunderx2t99_prefetch_tune): Likewise.
5340 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5342 (aarch64_override_options_internal): Update to set
5343 PARAM_PREFETCH_DYNAMIC_STRIDES.
5344 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5345 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5346 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5347 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5348 prefetch-dynamic-strides setting.
5350 2018-05-07 Luis Machado <luis.machado@linaro.org>
5352 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5353 <minimum_stride>: New const int field.
5354 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5355 minimum_stride field.
5356 (exynosm1_prefetch_tune): Likewise.
5357 (thunderxt88_prefetch_tune): Likewise.
5358 (thunderx_prefetch_tune): Likewise.
5359 (thunderx2t99_prefetch_tune): Likewise.
5360 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5361 (aarch64_override_options_internal): Update to set
5362 PARAM_PREFETCH_MINIMUM_STRIDE.
5363 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5364 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5365 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5366 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5367 stride is constant and is below the minimum stride threshold.
5369 2018-05-06 Jakub Jelinek <jakub@redhat.com>
5372 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
5373 the type is addressable. Don't force op into register if it has
5376 2018-05-05 Roland McGrath <mcgrathr@google.com>
5379 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
5380 any section for which we don't know a specific type it should have,
5381 regardless of name. Previously this was done only for the exact
5382 names ".init_array", ".fini_array", and ".preinit_array".
5383 (default_elf_asm_named_section): Add comment about
5384 relationship with default_section_type_flags and SECTION_NOTYPE.
5385 (get_section): Don't consider it a type conflict if one side has
5386 SECTION_NOTYPE and the other doesn't, as long as neither has the
5387 SECTION_BSS et al used in the default_section_type_flags logic.
5389 2018-05-05 Tom de Vries <tom@codesourcery.com>
5392 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
5393 (workaround_barsyncs): New function.
5394 (nvptx_reorg): Use workaround_barsyncs.
5395 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5396 (define_expand "nvptx_membar_cta"): New define_expand.
5397 (define_insn "*nvptx_membar_cta"): New insn.
5399 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
5401 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
5402 To improve optimization opportunities.
5403 * builtin-types.def: The new needed builtin types for the above.
5405 2018-05-04 Richard Biener <rguenther@suse.de>
5407 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
5408 * gimple-ssa-store-merging.c
5409 (imm_store_chain_info::output_merged_store): Remove redundant create,
5410 release split_store vector contents on failure.
5411 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
5412 scalar stmt vector on cache hit.
5414 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
5416 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
5418 * config.gcc (powerpc-xilinx-eabi*): Remove.
5419 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
5421 (fusion_addis_mem_combo_load): Ditto.
5422 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
5424 (rs6000_cpu_cpp_builtins): Ditto.
5425 * config/rs6000/rs6000-linux.c
5426 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
5427 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
5428 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
5430 (rs6000_setup_reg_addr_masks): Ditto.
5431 (rs6000_init_hard_regno_mode_ok): Ditto.
5432 (rs6000_option_override_internal): Ditto.
5433 (legitimate_lo_sum_address_p): Ditto.
5434 (rs6000_legitimize_address): Ditto.
5435 (rs6000_legitimize_reload_address): Ditto.
5436 (rs6000_legitimate_address_p): Ditto.
5437 (abi_v4_pass_in_fpr): Ditto.
5438 (setup_incoming_varargs): Ditto.
5439 (rs6000_gimplify_va_arg): Ditto.
5440 (rs6000_split_multireg_move): Ditto.
5441 (rs6000_savres_strategy): Ditto.
5442 (rs6000_emit_prologue_components): Ditto.
5443 (rs6000_emit_epilogue_components): Ditto.
5444 (rs6000_emit_prologue): Ditto.
5445 (rs6000_emit_epilogue): Ditto.
5446 (rs6000_elf_file_end): Ditto.
5447 (rs6000_function_value): Ditto.
5448 (rs6000_libcall_value): Ditto.
5449 * config/rs6000/rs6000.h: Ditto.
5450 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
5451 (TARGET_MINMAX): ... this. New.
5452 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
5453 * config/rs6000/rs6000.md: Remove Xilinx FP support.
5454 (*movsi_internal1_single): Delete.
5455 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
5456 mfpu=, mxilinx-fpu): Delete.
5457 * config/rs6000/singlefp.h: Delete.
5458 * config/rs6000/sysv4.h: Remove Xilinx FP support.
5459 * config/rs6000/t-rs6000: Ditto.
5460 * config/rs6000/t-xilinx: Delete.
5461 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
5462 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
5463 (VStype_simple): Delete.
5464 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
5465 * config/rs6000/xfpu.h: Delete.
5466 * config/rs6000/xfpu.md: Delete.
5467 * config/rs6000/xilinx.h: Delete.
5468 * config/rs6000/xilinx.opt: Delete.
5469 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
5470 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
5472 2018-05-04 Tom de Vries <tom@codesourcery.com>
5475 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
5478 2018-05-04 Richard Biener <rguenther@suse.de>
5481 * tree-complex.c (update_complex_assignment): We are always in SSA form.
5482 (expand_complex_div_wide): Likewise.
5483 (expand_complex_operations_1): Likewise.
5484 (expand_complex_libcall): Preserve EH info of the original stmt.
5485 (tree_lower_complex): Handle removed blocks.
5486 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
5487 on complex multiplication and division libcall builtins.
5489 2018-05-04 Richard Biener <rguenther@suse.de>
5492 * fold-const.c (negate_expr_p): Restrict negation of operand
5493 zero of a division to when we know that can happen without
5495 (fold_negate_expr_1): Likewise.
5497 2018-05-04 Jakub Jelinek <jakub@redhat.com>
5500 * real.h (real_nextafter): Declare.
5501 * real.c (real_nextafter): New function.
5502 * fold-const-call.c (fold_const_nextafter): New function.
5503 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
5504 CASE_CFN_NEXTTOWARD.
5505 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
5506 even when arg1_mode is different from arg0_mode.
5508 2018-05-03 Nathan Sidwell <nathan@acm.org>
5510 * doc/extend.texi (Deprecated Features): Remove
5512 (Backwards Compatibility): Likewise.
5513 * doc/invoke.texi (C++ Language Options): Likewise.
5514 (C++ Dialect Options): Likewise.
5516 2018-05-03 Jakub Jelinek <jakub@redhat.com>
5519 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
5520 _mm512_mask_mullox_epi64): New intrinsics.
5522 2018-05-03 Tom de Vries <tom@codesourcery.com>
5525 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5526 dump files): Add offload-tree.
5528 2018-05-03 Richard Biener <rguenther@suse.de>
5530 PR tree-optimization/85615
5531 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
5532 to loops not nested in BBs loop father to avoid creating multi-entry
5535 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5537 PR tree-optimization/70291
5538 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
5539 arguments. Change return type to tree. Emit libcall as a new
5540 statement rather than replacing existing one when inplace_p is true.
5541 (expand_complex_multiplication_components): New function.
5542 (expand_complex_multiplication): Expand floating-point complex
5543 multiplication using the above.
5544 (expand_complex_division): Rename inner_type parameter to type.
5545 Update expand_complex_libcall call-site.
5546 (expand_complex_operations_1): Update expand_complex_multiplication
5547 and expand_complex_division call-sites.
5549 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5552 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5553 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5554 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
5555 the highest significant bit of the shift count mask is clear. In
5556 check whether and[sq]i3 is needed verify that all significant bits
5557 of the shift count other than the highest are set.
5559 2018-05-02 Tom de Vries <tom@codesourcery.com>
5562 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
5563 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
5564 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
5565 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
5566 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
5567 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
5568 __builtin_goacc_parlevel_size.
5570 2018-05-02 Richard Biener <rguenther@suse.de>
5572 PR tree-optimization/85597
5573 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
5574 do not use split vect_get_vec_defs call but call vect_get_slp_defs
5577 2018-05-02 Tom de Vries <tom@codesourcery.com>
5580 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5581 dump files): Add ltrans-tree.
5583 2018-05-02 Tom de Vries <tom@codesourcery.com>
5586 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5587 dump files): Add wpa-ipa.
5589 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
5591 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
5592 powerpc*-*-linux*paired* target.
5593 * config/rs6000/750cl.h: Delete.
5594 * config/rs6000/paired.h: Delete.
5595 * config/rs6000/paired.md: Delete.
5596 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
5598 * config/rs6000/rs6000-builtin.def: Remove paired float support.
5599 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
5600 comment. Remove paired float support.
5601 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
5602 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
5604 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
5605 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
5607 * config/rs6000/rs6000.c: Remove paired float support.
5608 (paired_expand_vector_init, paired_expand_vector_move,
5609 paired_emit_vector_compare, paired_emit_vector_cond_expr,
5610 (paired_expand_lv_builtin, paired_expand_stv_builtin,
5611 paired_expand_builtin, paired_expand_predicate_builtin,
5612 paired_init_builtins): Delete.
5613 * config/rs6000/rs6000.h: Remove paired float support.
5614 * config/rs6000/rs6000.md: Remove paired float support.
5615 (move_from_CR_ov_bit): Delete.
5616 * config/rs6000/rs6000.opt (mpaired): Delete.
5617 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
5618 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
5620 2018-05-02 Richard Biener <rguenther@suse.de>
5623 * gimplify.c (gimplify_save_expr): When in SSA form allow
5624 SAVE_EXPRs to compute to SSA vars.
5626 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5629 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5630 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5631 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
5632 clobber operands[2], instead use a new pseudo. Formatting fixes.
5634 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
5636 PR tree-optimization/85586
5637 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
5638 exit early for statements in the same group if the accesses are
5641 2018-05-02 Tom de Vries <tom@codesourcery.com>
5644 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
5647 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
5649 PR tree-optimization/85143
5650 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
5652 2018-05-01 Tom de Vries <tom@codesourcery.com>
5655 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
5656 not found" error message.
5658 2018-05-01 Tom de Vries <tom@codesourcery.com>
5661 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
5662 * vec.c (test_ordered_remove_if): New function.
5663 (vec_c_tests): Call test_ordered_remove_if.
5664 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
5665 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
5666 * tree-vect-patterns.c (vect_pattern_recog_1): Use
5667 VEC_ORDERED_REMOVE_IF.
5669 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5671 PR tree-optimization/82665
5672 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
5673 pointer subtraction where arguments come from a memchr call.
5675 2018-05-01 Jakub Jelinek <jakub@redhat.com>
5677 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
5678 --push-state --as-needed and --pop-state instead of --as-needed and
5679 --no-as-needed if ld supports it.
5680 * configure: Regenerated.
5683 * doc/install.texi2html: Replace _002d with - and _002a with * in
5684 generated html files using sed.
5686 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5689 * gcc-rich-location.c (blank_line_before_p): New function.
5690 (use_new_line): New function.
5691 (gcc_rich_location::add_fixit_insert_formatted): New function.
5692 * gcc-rich-location.h
5693 (gcc_rich_location::add_fixit_insert_formatted): New function.
5695 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5697 * selftest.c (assert_streq): Rename "expected" and "actual" to
5698 "val1" and "val2". Extend NULL-handling to cover both inputs
5699 symmetrically, while still requiring both to be non-NULL for a pass.
5700 * selftest.h (assert_streq): Rename "expected" and "actual" to
5702 (ASSERT_EQ): Likewise.
5703 (ASSERT_EQ_AT): Likewise.
5704 (ASSERT_KNOWN_EQ): Likewise.
5705 (ASSERT_KNOWN_EQ_AT): Likewise.
5706 (ASSERT_NE): Likewise.
5707 (ASSERT_MAYBE_NE): Likewise.
5708 (ASSERT_MAYBE_NE_AT): Likewise.
5709 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
5710 the assertion to pass.
5711 (ASSERT_STREQ_AT): Likewise.
5713 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5715 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
5716 interaction with -pie.
5718 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5720 * selftest.h: Fix alphabetization of per-source-file selftest
5723 2018-04-30 Jason Merrill <jason@redhat.com>
5725 PR c++/61982 - dead stores to destroyed objects.
5726 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
5729 2018-04-30 Jason Merrill <jason@redhat.com>
5731 * tree.c (build_clobber): New.
5732 * tree.h: Declare it.
5733 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
5735 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5737 * diagnostic-show-locus.c (layout::layout): Update for
5738 location_get_source_line returning a char_span.
5739 (struct char_span): Move to input.h.
5740 (struct correction): Update for fields in char_span becoming
5742 (struct source_line): Update for location_get_source_line
5743 returning a char_span.
5744 (layout::print_line): Likewise.
5745 * edit-context.c (edited_file::print_content): Likewise.
5746 (edited_file::print_diff_hunk): Likewise.
5747 (edited_file::print_run_of_changed_lines): Likewise.
5748 (edited_file::get_num_lines): Likewise.
5749 (edited_line::edited_line): Likewise.
5750 * final.c (asm_show_source): Likewise.
5751 * input.c (location_get_source_line): Convert return type
5752 from const char * to char_span, losing the final "line_len"
5754 (dump_location_info): Update for the above.
5755 (get_substring_ranges_for_loc): Likewise. Use a char_span
5756 when handling the literal within the line.
5757 (test_reading_source_line): Update for location_get_source_line
5758 returning a char_span.
5759 * input.h (class char_span): Move here from
5760 diagnostic-show-locus.c, converting from a struct to a class.
5761 Make data members private.
5762 (char_span::operator bool): New.
5763 (char_span::length): New.
5764 (char_span::get_buffer): New.
5765 (char_span::operator[]): New.
5766 (char_span::subspan): Make const.
5767 (char_span::xstrdup): New.
5768 (location_get_source_line): Convert return type from const char *
5769 to char_span, losing the final "line_size" param.
5771 2018-04-30 Jan Hubicka <jh@suse.cz>
5773 * lto-wrapper.c (ltrans_priorities): New static var.
5774 (cmp_priority): New.
5775 (run_gcc): Read priorities and if doing parallel build order
5776 the Makefile by them.
5778 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5780 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
5782 2018-04-30 Richard Biener <rguenther@suse.de>
5784 * tree-cfg.c (verify_address): Remove base argument, add
5785 flag whether to check TREE_ADDRESSABLE and do that.
5786 (verify_expr): Remove.
5787 (verify_types_in_gimple_reference): Add pieces from verify_expr.
5788 (verify_gimple_assign_single): Likewise.
5789 (verify_gimple_switch): Likewise.
5790 (verify_expr_location_1): Dereference tp once. Add (disabled)
5791 piece from verify_expr.
5792 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
5794 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5796 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
5798 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5800 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
5801 (small_data_pattern): Likewise.
5802 (arc_rewrite_small_data): Likewise.
5803 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
5804 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
5805 (get_symbol_alignment): New function.
5806 (legitimate_small_data_address_p): Likewise.
5807 (legitimate_scaled_address): Update, call
5808 legitimate_small_data_address_p.
5809 (output_sdata): New static variable.
5810 (arc_print_operand): Update how we handle small data operands.
5811 (arc_print_operand_address): Likewise.
5812 (arc_legitimate_address_p): Update, use
5813 legitimate_small_data_address_p.
5814 (arc_rewrite_small_data_p): Remove.
5815 (arc_rewrite_small_data_1): Likewise.
5816 (arc_rewrite_small_data): Likewise.
5817 (small_data_pattern): Likewise.
5818 (compact_sda_memory_operand): Update to use
5819 legitimate_small_data_address_p and get_symbol_alignment.
5820 (prepare_move_operands): Don't rewite sdata pattern.
5821 (prepare_extend_operands): Remove.
5822 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
5824 (zero_extendqisi2): Likewise.
5825 (zero_extendhisi2): Likewise.
5826 (extendqihi2): Likewise.
5827 (extendqisi2): Likewise.
5828 (extendhisi2): Likewise.
5832 * config/arc/constraints.md (Usd): Change it to memory constraint.
5834 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5836 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
5837 as source of std instructions.
5838 * config/arc/arc.md (movsi_insn): Update pattern predicate to
5839 allow 6-bit constants as source for store instructions.
5840 (movdi_insn): Update instruction pattern to allow 6-bit constants
5841 as source for store instructions.
5843 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5845 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
5847 2018-04-30 Nathan Sidwell <nathan@acm.org>
5848 Sandra Loosemore <sandra@codesourcery.com>
5850 * dumpfile.c (dump_open): Allow '-' for stdout.
5851 * doc/invoke.texi (Developer Options): Document dump filename
5852 determination early. Document stdin/stdout selection.
5854 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
5856 Microblaze Target: PIC data text relative
5858 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
5859 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
5861 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
5862 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
5863 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
5864 New addressing mode for data-text relative position indepenedent code.
5865 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
5866 'ADDRESS_SYMBOLIC_TXT_REL'.
5867 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
5868 (microblaze_legitimate_pic_operand): Exclude function calls from
5869 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
5870 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
5872 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5873 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5874 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
5875 for 'address + offset'.
5876 (microblaze_expand_prologue): Add new function prologue call for
5878 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
5879 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
5880 table in case of TARGET_PIC_DATA_TEXT_REL.
5881 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
5882 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
5883 Add new macros 'UNSPEC_TEXT',
5884 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
5885 + exclude function calls from 'UNSPEC_PLT' in case of data text
5887 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5888 new target hook for generating address diff vector tables in case of
5890 * doc/tm.texi : Regenerate.
5891 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
5892 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
5893 of addr diff vector generation.
5894 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5895 target hook definition.
5896 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
5897 Add default function for generate_pic_addr_diff_vec -> flag_pic.
5898 * doc/invoke.texi (Add new pic option): Add new microblaze pic
5899 option for data text relative.
5901 2018-04-30 Richard Biener <rguenther@suse.de>
5903 * tree-chrec.h (evolution_function_is_constant_p): Remove
5905 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
5907 2018-04-30 Richard Biener <rguenther@suse.de>
5910 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
5912 2018-04-30 Richard Biener <rguenther@suse.de>
5914 PR tree-optimization/28364
5915 PR tree-optimization/85275
5916 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
5917 copying first exit test.
5919 2018-04-28 Mark Wielaard <mark@klomp.org>
5921 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
5923 (dwarf_AT): Handle DW_AT_addr_base.
5924 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
5926 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
5929 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
5930 (*ashl<dwi>3_doubleword_mask_1): Ditto.
5931 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
5932 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
5934 2018-04-28 Richard Biener <rguenther@suse.de>
5936 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
5937 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
5938 to reflect use. Only add interesting stmts.
5940 2018-04-27 Martin Jambor <mjambor@suse.cz>
5943 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
5944 the jump function allows for passing through aggregate values.
5946 2018-04-27 David Malcolm <dmalcolm@redhat.com>
5948 * input.h (in_system_header_at): Convert from macro to inline
5950 (from_macro_expansion_at): Likewise.
5951 (from_macro_definition_at): Likewise.
5953 2018-04-27 Jeff Law <law@redhat.com>
5955 * config.gcc: Mark tile* targets as deprecated/obsolete.
5957 2018-04-27 Richard Biener <rguenther@suse.de>
5959 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
5962 2018-04-27 Richard Biener <rguenther@suse.de>
5964 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
5966 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
5968 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
5969 with Yd constraint. Set "preferred_for_speed" attribute from
5970 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
5972 (*movdi_internal): Ditto.
5973 (movti_interunit splitters): Remove
5974 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
5975 (movdi_interunit splitters): Ditto.
5976 * config/i386/constraints.md (Ye): Remove.
5977 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
5979 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5982 * config/aarch64/constraints.md (Usg): Limit to 31.
5985 2018-04-27 Jakub Jelinek <jakub@redhat.com>
5987 PR tree-optimization/85529
5988 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
5989 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
5990 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
5991 zero extension or masking of the MSB bit.
5992 (optimize_range_tests): Add FIRST_BB argument, pass it through
5993 to optimize_range_tests_var_bound.
5994 (maybe_optimize_range_tests, reassociate_bb): Adjust
5995 optimize_range_tests callers.
5997 2018-04-26 Richard Biener <rguenther@suse.de>
5998 Jakub Jelinek <jakub@redhat.com>
6000 * cgraph.h (symbol_table): Just declare debug method here.
6001 * symtab.c (symbol_table::debug): Define.
6003 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
6005 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
6007 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
6009 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
6010 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
6011 (*movdi_internal): Substitute Yi and Yj constraint with x
6012 and Ym and Yn constraint with y constraint. Update "isa"
6013 attribute and set "preferred_for_speed" attribute from
6014 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
6015 (*movsi_internal): Ditto.
6016 (*movdf_internal): Ditto.
6017 (*movsf_internal): Ditto.
6018 (*zero_extendsidi2): Ditto.
6019 * config/i386/sse.md (vec_set<mode>_0): Ditto.
6020 (sse2_loadld): Ditto.
6021 (*vec_extract<ssevecmodelower>_0): Ditto.
6022 (*vec_extractv4si_0_zext_sse4): Ditto.
6023 (vec_concatv2di): Ditto.
6024 (*vec_dup<mode>): Ditto.
6025 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
6026 * config/i386/constraints.md (Yi): Remove.
6031 2018-04-26 Nathan Sidwell <nathan@acm.org>
6033 * dumpfile.c (dump_open): New.
6034 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
6035 (dump_finish): Detect stdio/stderr by value not name.
6037 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
6039 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
6041 2018-04-26 Tom de Vries <tom@codesourcery.com>
6044 * config/nvptx/nvptx.c (verify_neutering_jumps)
6045 (verify_neutering_labels): New function
6046 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
6048 2018-04-26 Tom de Vries <tom@codesourcery.com>
6051 * config/nvptx/nvptx.c (needs_neutering_p): New function.
6052 (nvptx_single): Use needs_neutering_p to skip over insns that do not
6055 2018-04-26 Richard Biener <rguenther@suse.de>
6056 Tom de Vries <tom@codesourcery.com>
6059 * lto-streamer-out.c (output_function): Fixup loops if required to match
6060 discovery done in the reader.
6062 2018-04-26 Richard Biener <rguenther@suse.de>
6064 PR tree-optimization/85116
6065 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
6066 have a loop exit from the single latch predecessor. Remove
6067 case of header with just condition.
6068 (ch_base::copy_headers): Exclude infinite loops from any
6070 (pass_ch::execute): Record exits.
6072 2018-04-26 Richard Biener <rguenther@suse.de>
6074 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
6075 prologue cost vector and pass it to vect_get_load_cost.
6076 (vect_get_peeling_costs_all_drs): Likewise.
6077 (vect_peeling_hash_get_lowest_cost): Likewise.
6078 (vect_enhance_data_refs_alignment): Likewise.
6080 2018-04-26 Richard Biener <rguenther@suse.de>
6083 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
6084 checking of integer<->pointer conversions.
6085 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
6086 sign-/zero-extending pointer types.
6087 (expand_omp_for_static_chunk): Likewise.
6089 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
6090 Jean Lee <xiaoyur347@gmail.com>
6092 * config/mips/mips.c (mips_asan_shadow_offset): New function.
6093 (TARGET_ASAN_SHADOW_OFFSET): Define.
6094 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
6095 true for -fsanitize=address.
6097 2018-04-25 Mark Wielaard <mark@klomp.org>
6099 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
6102 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6104 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
6105 than "alu", remove explicit "memory" and "imm_disp" attributes.
6106 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
6109 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
6110 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
6111 gen_lowpart_no_emit.
6113 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
6116 * config/i386/i386.c (ix86_expand_builtin): Change memory
6117 operand to XI, extend p0 to Pmode.
6118 * config/i386/i386.md: Change unspec volatile and operand
6119 1 mode to XI, change operand 0 mode to P.
6121 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6123 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
6124 GET_MODE_MASK before any checking.
6125 (nds32_can_use_bset_p): Likewise.
6126 (nds32_can_use_btgl_p): Likewise.
6128 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6130 * config/nds32/nds32-doubleword.md: New define_split pattern for
6131 illegal register number.
6133 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6135 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
6137 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6139 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
6141 2018-04-25 Richard Biener <rguenther@suse.de>
6143 * lto-streamer.h (LTO_major_version): Bump to 8.
6145 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6147 * BASE-VER: Set to 9.0.0.
6149 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
6151 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
6152 in __abskf2 and __powikf2.
6154 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6157 * config/aarch64/constraints.md (Usg, Usj): New constraints.
6158 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
6159 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
6160 Use the above on operand 2. Reindent.
6161 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
6163 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6166 * common/config/i386/i386-common.c (ix86_handle_option): Don't
6168 * config/i386/i386.opt (mcet): Removed.
6169 * doc/install.texi: Remove -mcet documentation.
6170 * doc/invoke.texi: Likewise.
6172 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6175 * doc/install.texi: Remove -mcet from bootstrap-cet.
6177 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6180 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
6181 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
6185 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
6186 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
6187 containing a CONST_VECTOR.
6189 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
6191 * doc/install.texi: Update newlib dependency for nvptx.
6193 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6196 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
6197 instead of INTVAL when shifting x left.
6199 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
6201 PR tree-optimization/85478
6202 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
6203 vect_grouped_store_supported for single element vectors.
6205 2018-04-24 Richard Biener <rguenther@suse.de>
6208 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
6209 load cost increase to the case of non-constant step.
6211 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6214 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
6215 destination if any_malformed_asm.
6217 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
6220 * expr.c (store_field): In the bitfield case, if the value comes from
6221 a function call and is returned in registers by means of a PARALLEL,
6222 do not change the mode of the temporary unless BLKmode and VOIDmode.
6224 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
6226 PR rtl-optimization/85423
6227 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
6228 dependencies to debug insns when the previous insn is non-debug.
6230 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6232 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
6233 enums into a single definition.
6234 (fls): Fix predicates and printing.
6237 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6239 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
6240 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
6241 and short u6 immediate.
6242 (check_if_valid_sleep_operand): Remove.
6243 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
6245 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6247 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
6248 flag_always_save_lp condition.
6249 * config/nds32/nds32.opt (malways-save-lp): New option.
6251 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6253 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
6254 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
6255 * config/nds32/nds32.h
6256 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
6257 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
6259 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6261 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
6262 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
6264 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6265 Chung-Ju Wu <jasonwucj@gmail.com>
6267 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
6269 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
6270 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
6272 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6274 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
6276 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6278 * config/nds32/nds32-protos.h (nds32_data_alignment,
6279 nds32_local_alignment): Declare.
6280 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
6281 nds32_local_alignment): New functions.
6282 (TARGET_CONSTANT_ALIGNMENT): Define.
6283 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
6285 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6287 * config/nds32/nds32.c
6288 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
6289 (TARGET_MODES_TIEABLE_P): Likewise.
6291 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6293 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
6296 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
6297 Chung-Ju Wu <jasonwucj@gmail.com>
6299 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6300 for unaligned access.
6301 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6303 * config/nds32/nds32-intrinsic.md: Likewise.
6304 * config/nds32/nds32_intrinsic.h: Likewise.
6305 * config/nds32/nds32.h (nds32_builtins): Likewise.
6306 * config/nds32/nds32.opt (munaligned-access): New option.
6307 * config/nds32/nds32.c (nds32_asm_file_start): Display
6308 flag_unaligned_access status.
6310 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6312 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
6313 -mno-relax is present.
6314 * config/riscv/linux.h (LINK_SPEC): Ditto.
6316 2018-04-20 Martin Sebor <msebor@redhat.com>
6319 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
6321 (gimple_fold_builtin_stxcpy_chk): Same.
6322 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
6324 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
6327 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
6328 __powikf2 when long double is IEEE 128-bit.
6330 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6332 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
6333 step to make sure stack always aligned.
6335 2018-04-20 Carl Love <cel@us.ibm.com>
6338 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
6339 size check for arg0.
6341 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
6342 Tom de Vries <tom@codesourcery.com>
6345 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
6346 Emit insns for calls too.
6347 (nvptx_find_par): Always look for worker-level predecessor insn.
6348 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
6350 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
6351 (nvptx_process_pars): Propagate frames for calls.
6353 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
6356 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
6358 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
6359 (ix86_handle_option): Don't handle OPT_mibt.
6360 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
6362 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
6364 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
6365 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
6366 (ix86_target_macros): Define __CET__ with flag_cf_protection
6367 for -fcf-protection.
6368 * config/i386/i386.c (isa2_opts): Remove -mibt.
6369 * config/i386/i386.h (TARGET_IBT): Removed.
6370 (TARGET_IBT_P): Likewise.
6371 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
6372 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
6373 * config/i386/i386.opt (mcet): Update help message.
6376 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
6377 -mcet as an alias for -mshstk.
6379 2018-04-20 Richard Biener <rguenther@suse.de>
6382 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
6383 complexity by forcing a single use of the multiply operand.
6385 2018-04-20 Martin Jambor <mjambor@suse.cz>
6388 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
6389 recursion dependency to only apply to non-clones.
6391 2018-04-20 Martin Jambor <mjambor@suse.cz>
6394 * ipa-cp.c (create_specialized_node): Check that clones of
6395 self-recursive edges exist during IPA-CP.
6397 2018-04-19 Toon Moene <toon@moene.org>
6399 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
6402 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6404 PR tree-optimization/85467
6405 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
6406 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
6407 VECTOR_CST element to type.
6409 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6412 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
6413 * config/i386/i386.md (builtin_setjmp_setup): Removed.
6414 (builtin_longjmp): Likewise.
6415 (save_stack_nonlocal): New pattern.
6416 (restore_stack_nonlocal): Likewise.
6418 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6421 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6422 Replace ASM_OUTPUT_LABEL with fprintf.
6424 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6427 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6428 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
6429 * config/i386/i386-c.c (ix86_target_macros_internal): Also
6430 define __IBT__ and __SHSTK__ for -fcf-protection.
6431 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
6433 (ix86_trampoline_init): Likewise.
6434 (x86_output_mi_thunk): Likewise.
6435 (ix86_notrack_prefixed_insn_p): Likewise.
6436 (ix86_option_override_internal): Don't disallow -fcf-protection.
6437 * config/i386/i386.md (rdssp<mode>): Also enable for
6439 (incssp<mode>): Likewise.
6440 (nop_endbr): Likewise.
6441 * config/i386/i386.opt (mcet): Change help message to built-in
6445 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
6446 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
6447 enable CET built-in functions.
6449 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
6451 * common/config/i386/i386-common.c
6452 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
6453 OPTION_MASK_ISA_MOVDIRI_UNSET,
6454 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
6455 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
6456 * config.gcc (movdirintrin.h): New header.
6457 * config/i386/cpuid.h (bit_MOVDIRI,
6458 bit_MOVDIR64B): New bits.
6459 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
6461 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
6462 (VOID, PVOID, PCVOID)): New function types.
6463 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
6464 __builtin_ia32_directstoreu_u64,
6465 __builtin_ia32_movdir64b): New builtins.
6466 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
6467 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
6469 (ix86_valid_target_attribute_inner_p): Ditto.
6470 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
6471 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
6472 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
6473 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
6474 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
6475 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
6476 (movdiri<mode>, movdir64b_<mode>): New.
6477 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
6478 * config/i386/immintrin.h: Include movdirintrin.h.
6479 * config/i386/movdirintrin.h: New file.
6480 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
6482 2018-04-19 Richard Biener <rguenther@suse.de>
6485 * cfg.c (clear_bb_flags): When loop state says we have
6486 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
6488 2018-04-19 Richard Biener <rguenther@suse.de>
6490 PR tree-optimization/84737
6491 * tree-vect-data-refs.c (vect_copy_ref_info): New function
6492 copying restrict info.
6493 (vect_setup_realignment): Use it.
6494 * tree-vectorizer.h (vect_copy_ref_info): Declare.
6495 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
6496 the first DR to all generated stores.
6497 (vectorizable_load): Likewise for loads.
6499 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6501 PR tree-optimization/85446
6502 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
6503 the integral and pointer types to have the same precision.
6505 * doc/install.texi: Document --disable-cet being the default and
6508 2018-04-18 Martin Liska <mliska@suse.cz>
6510 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
6513 2018-04-18 Martin Liska <mliska@suse.cz>
6516 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6519 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6520 arguments if they are comparable.
6522 2018-04-18 Martin Liska <mliska@suse.cz>
6525 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6528 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6531 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
6534 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
6535 ENDBR after calling __morestack.
6537 2018-04-18 David Malcolm <dmalcolm@redhat.com>
6540 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
6541 by using gcc_base_ver to generate a gcc_driver_version, and use
6542 it when generating GCC_DRIVER_NAME.
6543 * configure: Regenerate.
6545 2018-04-18 Jakub Jelinek <jakub@redhat.com>
6548 * config.gcc: Obsolete powerpc*-*-*spe*.
6550 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6553 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
6554 (stabstr_D): Change type of unum from unsigned int to
6555 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
6558 2018-04-17 Jim Wilson <jimw@sifive.com>
6561 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
6562 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
6563 Set arg_pointer_offset after using pretend_args_size.
6565 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6567 PR rtl-optimization/85431
6568 * dse.c (record_store): Ignore zero width stores.
6571 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
6572 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
6573 __builtin_stack_restore rather than after it.
6574 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
6575 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
6576 argument instead of virtual_dynamic_stack_rtx.
6578 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
6580 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
6582 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
6583 Add note to error message to explain internal mapping of overloaded
6584 built-in function name to non-overloaded built-in function name.
6585 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
6588 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6591 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
6592 where the inputs overlap with the output.
6594 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6596 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
6597 (=v, v) alternative and explicit "memory" attribute.
6598 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
6599 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6601 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
6602 "sselog1" type instead of "sselog".
6603 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
6604 "sselog". Remove explicit "memory" attribute.
6605 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
6606 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6608 (vec_extract_hi_v32hi): Merge all alternatives into one, use
6609 "sselog1" type instead of "sselog". Remove explicit "memory"
6611 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
6612 use "sselog1" type instead of "sselog". Remove explicit "memory"
6614 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
6615 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6617 (vec_extract_hi_v64qi): Merge all alternatives into one, use
6618 "sselog1" type instead of "sselog". Remove explicit "memory"
6620 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
6621 use "sselog1" type instead of "sselog". Remove explicit "memory"
6625 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
6628 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
6631 2018-04-17 Martin Jambor <mjambor@suse.cz>
6634 * ipa-cp.c (create_specialized_node): Call
6635 expand_all_artificial_thunks if necessary.
6637 2018-04-17 Martin Liska <mliska@suse.cz>
6640 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
6641 in message, remote space in between '_G' and '('.
6643 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6646 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
6647 avx512f_vmcmp<mode>3<round_saeonly_name>,
6648 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
6649 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
6650 avx512f_rndscale<mode><round_saeonly_name>,
6651 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
6652 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
6653 Use %<iptr>2 instead of %2 for -masm=intel.
6654 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
6655 avx512f_vcvttss2usi<round_saeonly_name>,
6656 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
6658 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
6659 avx512f_vcvttsd2usi<round_saeonly_name>,
6660 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
6661 Use %q1 instead of %1 for -masm=intel.
6662 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
6663 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
6664 of %3 for -masm=intel.
6665 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
6667 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
6669 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
6671 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
6672 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
6674 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
6676 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
6677 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
6678 %g1 and one with %0 and %1.
6679 (avx512er_vmrcp28<mode><round_saeonly_name>,
6680 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
6682 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
6683 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
6684 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
6685 of %0 and %{%4%} for -masm=intel.
6686 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
6687 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
6688 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
6689 order of %0 and %{%5%}%{z%} for -masm=intel.
6691 2018-04-17 Jan Hubicka <jh@suse.cz>
6694 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
6696 2018-04-17 Martin Liska <mliska@suse.cz>
6699 * multiple_target.c (create_dispatcher_calls): Set apostrophes
6700 for target_clone error message. Make default implementation
6701 clone to be a local declaration.
6702 (separate_attrs): Add new argument and check for an empty
6704 (expand_target_clones): Handle it.
6705 (ipa_target_clone): Make redirection just for target_clones
6708 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
6709 Tom de Vries <tom@codesourcery.com>
6712 * omp-expand.c (expand_oacc_for): Add dummy false branch for
6713 tiled basic blocks without omp continue statements.
6715 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
6718 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
6719 vec_extract expression as having side effects to make sure it gets
6722 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
6725 * config/i386/i386.c (get_builtin_code_for_version): Check
6728 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
6731 * gcc/config.gcc: Support "skylake".
6732 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
6734 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
6735 (processor_target_table): Add "skylake".
6736 (ix86_option_override_internal): Add "skylake".
6737 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
6738 PROCESSOR_CANNONLAKE.
6739 (get_builtin_code_for_version): Fix priority for
6740 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
6741 PROCESSOR_SKYLAKE-AVX512.
6742 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
6743 (processor_type): Add PROCESSOR_SKYLAKE.
6745 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
6746 Jason Merrill <jason@redhat.com>
6749 * convert.c (convert_to_integer_1): Use direct recursion for
6750 enumeral types and types with a precision less than the number
6751 of bits in their mode.
6753 2018-04-16 Julia Koval <julia.koval@intel.com>
6756 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
6757 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
6759 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
6762 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
6763 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
6764 and -mno-direct-move.
6766 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
6769 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
6770 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
6771 Ensure negative shifts result in {0}.
6773 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
6775 PR rtl-optimization/79916
6776 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
6777 regs (if any) to define how to gnerate SD moves when LRA is in
6780 2018-04-13 Jakub Jelinek <jakub@redhat.com>
6782 PR rtl-optimization/85393
6783 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
6784 * except.c (expand_dw2_landing_pad_for_region): Make static.
6785 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
6786 a label and unconditional jump to old_bb, rather than
6787 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
6790 PR rtl-optimization/85376
6791 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
6792 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
6793 instead of a specific value.
6795 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6796 Bin Cheng <bin.cheng@arm.com>
6798 PR tree-optimization/82965
6799 PR tree-optimization/83991
6800 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
6801 by_profile_only parameter.
6802 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
6803 information if the loop was predicted to iterate too many times.
6804 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
6806 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6809 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
6812 2018-04-13 Martin Liska <mliska@suse.cz>
6813 Jakub Jelinek <jakub@redhat.com>
6816 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
6817 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
6818 * builtins.c (expand_builtin_memory_copy_args): Use
6819 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
6820 handle dest_addr == pc_rtx.
6822 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
6825 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
6826 asked to not generate direct moves.
6827 (fix_trunc<mode>si2_stfiwx): Similar.
6828 (fix_trunc<mode>si2_internal): Similar.
6830 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6833 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
6834 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
6835 lookup if dest in some wider mode is known to be const0_rtx and
6836 if so, record permanent equivalence for it to be ZERO_EXTEND of
6837 the narrower mode destination.
6839 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6841 * lto-streamer-out.c (output_function): Revert 259346.
6842 * omp-expand.c (expand_oacc_for): Likewise.
6844 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
6846 PR rtl-optimization/85354
6847 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
6848 * sel-sched.c (sel_global_init): ... here.
6850 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
6853 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
6854 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6855 mode for PE-COFF targets.
6856 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
6857 (i386_pe_asm_lto_end): Likewise.
6858 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
6859 (TARGET_ASM_LTO_END): Likewise.
6860 * config/i386/winnt.c (saved_debug_info_level): New static variable.
6861 (i386_pe_asm_lto_start): New function.
6862 (i386_pe_asm_lto_end): Likewise.
6864 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6865 Richard Biener <rguenther@suse.de>
6868 * lto-streamer-out.c (output_function): Fix CFG loop state before
6870 * omp-expand.c (expand_oacc_for): Handle calls to internal
6871 functions like regular functions.
6873 2018-04-12 Richard Biener <rguenther@suse.de>
6876 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
6877 for the early LTO debug to properly generate references to it
6878 during DIE emission. Do not re-use that for the skeleton for
6880 (dwarf2out_early_finish): Likewise.
6882 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6885 * config/i386/sse.md
6886 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
6887 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
6888 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
6889 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
6890 and output is a reg, avoid creating invalid lowpart subreg, but
6891 instead split into a 512-bit move. Don't split if not AVX512VL,
6892 input is xmm16+ reg and output is a mem.
6893 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
6894 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
6895 xmm16+ reg and output is a mem.
6897 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6899 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
6900 also for flag_dwarf2_cfi_asm.
6902 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6904 PR rtl-optimization/85342
6905 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
6906 a bool scalar var inside of the loop instead. Don't try to update
6907 recog_data.operand after failed apply_change_group.
6909 2018-04-12 Tom de Vries <tom@codesourcery.com>
6912 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
6913 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
6914 array with flexible array member as array without given dimension.
6915 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
6916 argument for undefined param to true.
6918 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
6921 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6922 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
6923 from PowerPC section.
6924 * config/rs6000/sysv4.opt (mcall-): Improve help text.
6925 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
6926 help text that is too long.
6927 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
6928 help text that is too long.
6929 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
6930 help text that is too long.
6932 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
6934 * config/alpha/alpha.md (stack_probe_internal): Rename
6935 from "probe_stack". Update all callers.
6937 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6939 PR rtl-optimization/84566
6940 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
6941 sched_macro_fuse_insns.
6943 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6946 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
6947 (compute_block_dependences): ... from here.
6949 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6951 PR tree-optimization/85331
6952 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
6953 from int to HOST_WIDE_INT.
6955 2018-04-11 Martin Jambor <mjambor@suse.cz>
6958 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
6959 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
6960 not the same as the source val.
6961 (cgraph_edge_brings_value_p): New parameter.
6962 (gather_edges_for_value): Pass destination value to
6963 cgraph_edge_brings_value_p.
6964 (perhaps_add_new_callers): Likewise.
6965 (get_info_about_necessary_edges): Likewise and exclude values brought
6966 only by self-recursive edges.
6967 (create_specialized_node): Redirect only clones of self-calling edges.
6968 (+self_recursive_pass_through_p): New function.
6969 (find_more_scalar_values_for_callers_subset): Use it.
6970 (find_aggregate_values_for_callers_subset): Likewise.
6971 (known_aggs_to_agg_replacement_list): Removed.
6972 (decide_whether_version_node): Re-calculate known constants for all
6973 remaining context clones.
6975 2018-04-11 Richard Biener <rguenther@suse.de>
6978 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
6979 from early DWARF output.
6980 (dwarf2out_early_finish): Output line info unconditionally into
6981 early DWARF and add reference to it.
6983 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6986 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
6987 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
6988 other than V2DFmode using iptr mode attribute.
6989 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
6991 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6993 PR rtl-optimization/84659
6994 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
6996 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6999 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
7001 (output_loc_list): Pass address of a dummy size variable even in the
7002 locview handling loop.
7003 (index_location_lists): Add comment on why skip_loc_list_entry can't
7006 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
7009 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
7012 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
7015 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7016 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
7017 and -mstring-compare-inline-limit.
7019 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7022 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
7023 for stack clash protection in a register whenever we need it to be in
7026 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7028 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
7029 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
7031 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7034 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
7036 (mlong-double-): Ditto.
7037 * config/rs6000/sysv4.opt (msdata=): Ditto.
7038 (mtls-size=): Ditto.
7040 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7042 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
7043 erroneous entries for
7044 "vector int vec_ldl (int, long int *)", and
7045 "vector unsigned int vec_ldl (int, unsigned long int *)".
7046 Add comments and entries for
7047 "vector bool char vec_ldl (int, bool char *)",
7048 "vector bool short vec_ldl (int, bool short *)",
7049 "vector bool int vec_ldl (int, bool int *)",
7050 "vector bool long long vec_ldl (int, bool long long *)",
7051 "vector pixel vec_ldl (int, pixel *)",
7052 "vector long long vec_ldl (int, long long *)",
7053 "vector unsigned long long vec_ldl (int, unsigned long long *)".
7054 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
7055 type tree bool_long_long_type_node and correct definition of
7056 bool_V2DI_type_node to make reference to this new type tree.
7057 (rs6000_mangle_type): Replace erroneous reference to
7058 bool_long_type_node with bool_long_long_type_node.
7059 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
7060 comments to emphasize sign distinctions for char and int types and
7061 replace RS6000_BTI_bool_long constant with
7062 RS6000_BTI_bool_long_long constant. Also add comment to restrict
7063 use of RS6000_BTI_pixel.
7064 (bool_long_type_node): Remove this macro definition.
7065 (bool_long_long_type_node): New macro definition
7067 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7069 PR rtl-optimization/85300
7070 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
7071 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
7072 simplify_unary_operation fails.
7074 2018-04-10 Martin Liska <mliska@suse.cz>
7076 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
7077 cgraph_edge and ipa_ref.
7079 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7083 * config/i386/sse.md
7084 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
7085 computation of the VEC_MERGE selector from mask.
7086 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
7087 Fix decoding of the VEC_MERGE selector into mask.
7089 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7091 PR tree-optimization/85286
7092 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
7094 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7096 * final.c (final_1): Set insn_last_address as well as
7097 insn_current_address.
7099 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7102 * explow.c (emit_stack_probe): Call validize_mem on memory location
7103 before passing it to gen_probe_stack. Create address operand and
7104 legitimize it for the probe_stack_address case.
7106 2018-04-09 Jan Hubicka <jh@suse.cz>
7109 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
7110 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
7111 * tree.c (free_lang_data_in_type): Fix handling of binfos;
7113 (free_lang_data): Rebuild type inheritance graph.
7115 2018-04-09 Martin Sebor <msebor@redhat.com>
7117 * invoke.texi (-finline-small-functions): Mention other optimization
7119 (-findirect-inlining, -fpartial-inlining): Same.
7120 (-finline-functions-called-once): Same.
7121 (-freorder-blocks-and-partition): Same.
7123 2018-04-09 Jan Hubicka <jh@suse.cz>
7126 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
7127 jumps; choose last target that matches the criteria (i.e.
7128 no partition changes for non-crossing jumps).
7129 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
7130 support for redirecting crossing jumps to non-crossing.
7132 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
7134 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
7135 also for naked functions.
7137 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
7139 * config/arc/arc.md (add_shift): New pattern.
7140 (add_shift2): Likewise.
7141 (sub_shift): Likewise.
7142 (sub_shift_cmp0_noout): Likewise.
7143 (compare_si_ashiftsi): Likewise.
7144 (xbfu_cmp0_noout): New combine pattern.
7145 (xbfu_cmp0"): Likewise.
7146 (movsi_set_cc_insn): Place the predicable variant first.
7147 (commutative_binary_cmp0_noout): Remove clobber.
7148 (commutative_binary_cmp0): New pattern.
7149 (noncommutative_binary_cmp0): Likewise.
7150 (noncommutative_binary_cmp0_noout): Likewise.
7151 (noncommutative_binary_comparison_result_used): Removed.
7152 (rsub_cmp0): New pattern.
7153 (rsub_cmp0_noout): Likewise.
7154 (extzvsi): Changed, keep only meaningful variants.
7155 (SQH, SEZ): New iterators.
7156 (SQH_postfix): New mode attribute.
7157 (SEZ_prefix): New code attribute.
7158 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
7159 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
7160 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
7162 (noncommutative_operator): Check the availability of barrel
7165 2018-04-09 Richard Biener <rguenther@suse.de>
7167 PR tree-optimization/85284
7168 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
7169 Only use the niter constraining form of simple_iv when the exit
7172 2018-04-09 Tom de Vries <tom@codesourcery.com>
7175 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7176 (define_expand "*memory_barrier"): New define_expand.
7177 (define_insn "memory_barrier"): New insn.
7179 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7181 PR rtl-optimization/80463
7182 PR rtl-optimization/83972
7183 PR rtl-optimization/83480
7185 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
7186 correct producer for the insn.
7187 (tidy_control_flow): Fixup seqnos in case of debug insns.
7189 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7191 PR rtl-optimization/83913
7193 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
7194 different sched-times when merging exprs.
7196 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7198 PR rtl-optimization/83962
7200 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
7201 tidy_fallthru_edge and tidy_control_flow.
7203 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7205 PR rtl-optimization/83530
7207 * sel-sched.c (force_next_insn): New global variable.
7208 (remove_insn_for_debug): When force_next_insn is true, also leave only
7209 next insn in the ready list.
7210 (sel_sched_region): When the region wasn't scheduled, make another pass
7211 over it with force_next_insn set to 1.
7213 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7215 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
7217 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7218 for interrupt control.
7219 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7220 functions for interrupt control.
7221 * config/nds32/nds32-intrinsic.md: Likewise.
7222 * config/nds32/nds32_intrinsic.h: Likewise.
7223 * config/nds32/nds32.h (nds32_builtins): Likewise.
7225 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7227 * config/nds32/nds32.c (nds32_init_machine_status,
7228 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
7229 strict_aligned_p field.
7230 (nds32_expand_to_rtl_hook): New function.
7231 (TARGET_EXPAND_TO_RTL_HOOK): Define.
7232 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
7234 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7235 Chung-Ju Wu <jasonwucj@gmail.com>
7237 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
7238 * config/nds32/nds32-n7.md: New file.
7239 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
7240 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
7242 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
7243 * config/nds32/nds32.md (pipeline_model): Add n7.
7244 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
7245 * config/nds32/pipelines.md: Include n7 settings.
7247 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7248 Chung-Ju Wu <jasonwucj@gmail.com>
7250 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
7251 * config/nds32/nds32-e8.md: New file.
7252 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
7253 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
7255 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
7256 * config/nds32/nds32.md (pipeline_model): Add e8.
7257 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
7258 * config/nds32/pipelines.md: Include e8 settings.
7260 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7261 Chung-Ju Wu <jasonwucj@gmail.com>
7263 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
7264 * config/nds32/nds32-n8.md: New file.
7265 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
7266 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
7268 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
7269 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
7270 * config/nds32/nds32.md (pipeline_model): Add n8.
7271 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
7272 * config/nds32/pipelines.md: Include n8 settings.
7274 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7275 Chung-Ju Wu <jasonwucj@gmail.com>
7277 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
7278 * config/nds32/nds32-n9-2r1w.md: New file.
7279 * config/nds32/nds32-n9-3r2w.md: New file.
7280 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
7281 nds32_register_ports): New or modify for cpu n9.
7282 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
7284 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
7285 * config/nds32/nds32-utils.c: New file.
7286 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
7287 TARGET_MUL_SLOW): Define.
7288 * config/nds32/nds32.md (pipeline_model): New attribute.
7289 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
7290 New options that support cpu n9.
7291 * config/nds32/pipelines.md: Include n9 settings.
7292 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
7294 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7296 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
7297 information if necessary.
7298 (output_cond_branch_compare_zero): Likewise.
7299 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
7300 (nds32_target_alignment): Refine for alignment.
7301 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
7302 (FUNCTION_BOUNDARY): Modify.
7303 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
7305 * config/nds32/nds32.opt (malways-align, malign-functions): New.
7307 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7309 * config/nds32/constants.md (unspec_volatile_element): Add values for
7310 TLB operation and data prefetch.
7311 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7312 functions for TLB operation and data prefetch.
7313 * config/nds32/nds32-intrinsic.md: Likewise.
7314 * config/nds32/nds32_intrinsic.h: Likewise.
7315 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
7316 (nds32_print_operand): Likewise.
7317 * config/nds32/nds32.h (nds32_builtins): Likewise.
7319 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
7320 Andrew Pinski <pinsika@gcc.gnu.org>
7323 * match.pd: Use constant_boolean_node of correct type instead of
7324 boolean_true_node or boolean_false_node for simplifying
7325 pointer comparisons to zero.
7327 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7329 PR tree-optimization/80021
7330 * tree.c (verify_type_variant): Make error call in verify_variant_match
7331 translatable and remove final full stop.
7333 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7335 * config/nds32/constants.md (unspec_volatile_element): Add
7336 UNSPEC_VOLATILE_EH_RETURN.
7337 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
7338 nds32_output_stack_pop): Support dwarf exception handling process.
7339 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
7340 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
7341 exception handling process.
7342 (nds32_compute_stack_frame): Likewise.
7343 (nds32_return_addr_rtx): Likewise.
7344 (nds32_initial_elimination_offset): Likewise.
7345 (nds32_expand_prologue): Likewise.
7346 (nds32_expand_epilogue): Likewise.
7347 (nds32_dynamic_chain_address): New function.
7348 * config/nds32/nds32.h (machine_function): Add fields for dwarf
7350 (DYNAMIC_CHAIN_ADDRESS): Define.
7351 (EH_RETURN_DATA_REGNO): Define.
7352 (EH_RETURN_STACKADJ_RTX): Define.
7353 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
7354 patterns for dwarf exception handling.
7356 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7358 * config/nds32/nds32.h: Clean up obsolete macros.
7360 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7362 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7363 Add enum values for particular instructions.
7364 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7365 particular intrinsic functions.
7366 * config/nds32/nds32-intrinsic.md: Likewise.
7367 * config/nds32/nds32_intrinsic.h: Likewise.
7368 * config/nds32/nds32.h (nds32_builtins): Likewise.
7369 * config/nds32/nds32.md (type): Add pbsad and pbsada.
7370 (btst, ave): New patterns for particular instructions.
7372 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7374 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7375 Add enum values for atomic load/store and memory sync.
7376 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
7378 * config/nds32/nds32-intrinsic.md: Likewise.
7379 * config/nds32/nds32_intrinsic.h: Likewise.
7380 * config/nds32/nds32.h (nds32_builtins): Likewise.
7382 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7384 PR tree-optimization/85257
7385 * fold-const.c (native_encode_vector): If not all elts could fit
7386 and off is -1, return 0 rather than offset.
7387 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
7388 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
7389 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
7390 adjust buffer in native_interpret_expr call.
7392 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7394 * config/nds32/constants.md (unspec_volatile_element): Add cache
7395 control enum values.
7396 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
7397 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
7398 * config/nds32/nds32.c (nds32_cctl_names): New.
7399 (nds32_print_operand): Handle cache control register names.
7400 * config/nds32/nds32.h (nds32_builtins): New enum values.
7401 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
7403 * config/nds32/nds32.md (type): Add mmu.
7404 * config/nds32/pipelines.md (simple_insn): Add mmu.
7406 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7408 * config/nds32/nds32.md (type): Remove call.
7409 * config/nds32/pipelines.md (simple_insn): Likewise.
7411 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7413 * config/nds32/constants.md (unspec_volatile_element): Add
7414 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
7415 UNSPEC_VOLATILE_FMFCFG.
7416 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
7417 description for fmfcfg and fmfcsr.
7418 (bdesc_1arg): Add fmtcsr.
7419 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
7420 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
7421 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
7422 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
7423 unspec_fmfcfg): New patterns.
7424 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
7425 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
7426 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
7427 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
7428 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
7429 __nds32__fmfcfg): Define.
7431 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7433 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
7434 intrinsic register names.
7435 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
7436 intrinsic register enum values and macros.
7438 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7440 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
7441 for load/store addressing form.
7442 (nds32_print_operand_address): Likewise.
7444 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
7447 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
7448 based on LABEL_REF. Remove useless assertion.
7449 (pic_address_needs_scratch): Fix formatting.
7450 (sparc_legitimize_pic_address): Minor tweaks.
7451 (sparc_delegitimize_address): Adjust assertion accordingly.
7452 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
7453 into symbolic_operand.
7454 (movsi_high_pic_label_ref): Likewise.
7455 (movsi_lo_sum_pic_label_ref): Likewise.
7456 (movdi_pic_label_ref): Likewise.
7457 (movdi_high_pic_label_ref): Likewise.
7458 (movdi_lo_sum_pic_label_ref): Likewise.
7460 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
7462 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
7463 custom LIB_SPEC setup.
7465 2018-04-06 Ruslan Bukin <br@bsdpad.com>
7466 Kito Cheng <kito.cheng@gmail.com>
7468 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
7469 * config/riscv/freebsd.h: New.
7471 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7473 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
7474 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
7477 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7478 Kito Cheng <kito.cheng@gmail.com>
7480 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
7481 nds32_output_call, nds32_symbol_binds_local_p): New functions.
7482 * config/nds32/nds32-protos.h (nds32_output_call,
7483 nds32_output_return): Declare.
7484 * config/nds32/nds32.md: Refine all the call and return patterns.
7486 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7489 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
7490 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
7492 PR rtl-optimization/84872
7493 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
7494 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
7497 2018-04-06 Tamar Christina <tamar.christina@arm.com>
7499 * expr.c (copy_blkmode_to_reg): Revert 254862.
7500 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
7502 2018-04-06 Richard Biener <rguenther@suse.de>
7505 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
7506 after seeing a component reference with an adjacent field. Treat
7507 refs to arrays at struct end of external decls similar to
7508 refs to unconstrained commons.
7510 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7513 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
7514 look through SAVE_EXPRs with non-side-effects argument. Adjust
7516 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
7519 2018-04-06 Richard Biener <rguenther@suse.de>
7522 * alias.c (find_base_term): New wrapper around find_base_term
7523 unwinding CSELIB_VAL_PTR changes.
7524 (find_base_term): Do not restore CSELIB_VAL_PTR during the
7527 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7529 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
7531 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
7532 constant definitions.
7533 ("nop"): lr 0,0 -> nopr r0
7534 ("nop_lr0", "nop_lr1"): New insn definitions.
7536 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7538 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
7539 NDS32_V3PUSH_AVAILABLE_P macro.
7541 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
7542 Chung-Ju Wu <jasonwucj@gmail.com>
7544 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
7545 (nds32*-*-*): Add float and fpu_config into supported_defaults.
7546 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
7547 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
7548 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
7549 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
7550 * config/nds32/constraints.md: New constraints and checking for hard
7551 float configuration.
7552 * config/nds32/iterators.md: New mode iterator and attribute for hard
7553 float configuration.
7554 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
7556 * config/nds32/nds32-fpu.md: New file.
7557 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
7558 deal with hard float code generation.
7559 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
7561 (abi_type, float_reg_number): New enum type.
7562 * config/nds32/nds32-predicates.c: New predicates for hard float.
7563 * config/nds32/nds32-protos.h: Declare functions for hard float.
7564 * config/nds32/nds32.c: Implementation for hard float configuration.
7565 * config/nds32/nds32.h: Definitions for hard float configuration.
7566 * config/nds32/nds32.md: Include hard float machine description and
7567 modify patterns for hard float configuration.
7568 * config/nds32/nds32.opt: New options for hard float configuration.
7569 * config/nds32/predicates.md: New predicates for hard float
7572 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7574 * common/config/nds32/nds32-common.c
7575 (nds32_option_optimization_table): Enable -mreleax-hint by default.
7577 2018-04-05 Jakub Jelinek <jakub@redhat.com>
7580 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
7581 CONSTRUCTOR_ELT (ctor, ...)->value.
7583 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
7586 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
7588 2018-04-05 Tom de Vries <tom@codesourcery.com>
7591 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
7594 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7595 Kito Cheng <kito.cheng@gmail.com>
7597 * config/nds32/constraints.md (U33): Fine-tune checking condition.
7598 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
7599 * config/nds32/nds32.h (nds32_16bit_address_type): Add
7600 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
7602 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7603 Kito Cheng <kito.cheng@gmail.com>
7605 * config/nds32/constraints.md (Ufe): New memory constraint.
7606 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
7607 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
7608 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
7610 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
7611 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
7613 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7615 * config/nds32/nds32.md: Use optimize_size in the condition for
7616 alu-shift instructions.
7618 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7620 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
7622 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7624 * config/nds32/nds32.md (negsi2): Refine pattern.
7626 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7627 Chung-Ju Wu <jasonwucj@gmail.com>
7629 * config/nds32/iterators.md (shift_rotate): New code iterator.
7630 (shift): New code attribute.
7631 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
7632 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
7633 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
7634 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
7635 bit-wise operations.
7636 (andsi3, *andsi3): Ditto.
7637 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
7638 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
7639 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
7640 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
7641 nds32_ior_operand, nds32_xor_operand): New predicates.
7643 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7645 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
7646 (addsi3, subsi3): ... this.
7648 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7650 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
7652 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7654 * config/nds32/nds32.md: Adjust indention.
7656 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7658 * config/nds32/nds32.md (feature): New attribute.
7660 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7662 * config/nds32/nds32.md (subtype): New attribute.
7664 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
7667 * config/arm/arm-builtins.c (arm_expand_builtin): Change
7668 expansion to perform a bitwise AND of the argument followed by a
7669 boolean negation of the result.
7671 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
7673 PR rtl-optimization/84878
7674 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
7675 the basic block. Assert the use reference is not artificial and that
7676 it has an associated insn.
7678 2018-04-04 Michael Matz <matz@suse.de>
7680 * builtins.c (compute_objsize): Pass correct operand
7681 to array_at_struct_end_p.
7683 2018-04-04 Richard Biener <rguenther@suse.de>
7686 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
7687 from contexts for DINFO_LEVEL_TERSE and below.
7689 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7691 * config/nds32/nds32-doubleword.md (move_<mode>): Require
7692 resiter_operand condition.
7693 * config/nds32/nds32.md (*move<mode>): Ditto.
7695 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7696 Monk Chiang <sh.chiang04@gmail.com>
7698 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
7700 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7702 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
7704 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7705 Kito Cheng <kito.cheng@gmail.com>
7707 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
7708 nds32_cond_code_str, output_cond_branch,
7709 output_cond_branch_compare_zero, nds32_expand_cbranch,
7710 nds32_expand_cstore, nds32_expand_movcc,
7711 nds32_output_cbranchsi4_equality_zero,
7712 nds32_output_cbranchsi4_equality_reg,
7713 nds32_output_cbranchsi4_equality_reg_or_const_int,
7714 nds32_output_cbranchsi4_greater_less_zero: New functions.
7715 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
7716 nds32_expand_cstore, nds32_expand_movcc,
7717 nds32_output_cbranchsi4_equality_zero,
7718 nds32_output_cbranchsi4_equality_reg,
7719 nds32_output_cbranchsi4_equality_reg_or_const_int,
7720 nds32_output_cbranchsi4_greater_less_zero): Declare.
7721 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
7722 nds32_rimm11s_operand): New predicates.
7723 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
7724 * config/nds32/nds32.md: Rewrite all the branch and conditional move
7727 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7729 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
7730 * config/nds32/nds32.md: Ditto.
7731 * config/nds32/pipelines.md: Ditto.
7733 2018-04-04 Richard Biener <rguenther@suse.de>
7735 PR tree-optimization/85168
7736 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
7737 propagating abnormals.
7739 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7741 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
7743 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7744 Kito Cheng <kito.cheng@gmail.com>
7746 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
7747 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
7748 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
7749 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
7750 * config/nds32/nds32.md (sibcall_internal): New.
7751 (sibcall_register): Remove.
7752 (sibcall_immediate): Remove.
7753 (sibcall_value_internal): New.
7754 (sibcall_value_register): Remove.
7755 (sibcall_value_immediate): Remove.
7756 * config/nds32/predicates.md (nds32_general_register_operand): New.
7757 (nds32_call_address_operand): New.
7759 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7761 PR rtl-optimization/85167
7762 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
7763 bb_defs if *split_p, instead preinitialize it to NULL.
7765 PR tree-optimization/85156
7766 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
7767 evaluating the argument multiple times.
7769 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
7771 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
7773 (_mm_cvtpd_ps): Likewise.
7774 (_mm_cvttpd_epi32): Likewise.
7775 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
7776 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
7777 vector, pixel, and bool following altivec.h include.
7779 2018-04-03 Martin Sebor <msebor@redhat.com>
7781 * doc/extend.texi (Common Function Attributes): Clarify.
7782 (const attribute): Likewise.
7783 (pure attribute): Likewise.
7785 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7788 * config/i386/i386.c (ix86_expand_vector_set): Use
7789 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
7791 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
7793 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
7794 instructions when changing rounding bits to preserve precision bits
7795 in the x87 control word.
7797 2018-04-03 Martin Liska <mliska@suse.cz>
7799 PR tree-optimization/82491
7800 * rtl.h (strip_offset_and_add): Replace += suboffset with
7801 poly_uint64 () + suboffset.
7803 2018-03-29 Martin Liska <mliska@suse.cz>
7804 Martin Jambor <mjambor@suse.cz>
7807 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
7808 param_type is not an integral or pointer type.
7810 2018-04-03 Richard Biener <rguenther@suse.de>
7812 * sese.h (recompute_all_dominators): Remove.
7814 2018-04-02 Martin Sebor <msebor@redhat.com>
7816 * doc/invoke.texi (-Wrestrict): Fix typos.
7818 2018-04-02 Jim Wilson <jimw@sifive.com>
7820 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
7821 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
7822 (<optab>di3, <optab>si3_extend): Likewise.
7823 (<optab>si3_mask, <optab>si3_mask_1): New.
7824 (<optab>di3_mask, <optab>di3_mask_1): New.
7825 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
7826 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
7827 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
7829 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
7831 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
7834 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
7836 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
7837 (nds32_canonicalize_comparison): New function.
7839 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7840 Kito Cheng <kito.cheng@gmail.com>
7841 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7843 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
7844 * config/nds32/constants.md (unspec_volatile_element): Add
7845 UNSPEC_VOLATILE_RELAX_GROUP.
7846 * config/nds32/nds32-relax-opt.c: New file.
7847 * config/nds32/nds32-predicates.c
7848 (nds32_symbol_load_store_p): New function.
7849 * config/nds32/nds32-protos.h
7850 (nds32_symbol_load_store_p): Declare function.
7851 (make_pass_nds32_relax_opt): Declare new rtl pass function.
7852 * config/nds32/nds32.c
7853 (nds32_register_pass): New function to register pass.
7854 (nds32_register_passes): New function to register passes.
7855 * config/nds32/nds32.md (relax_group): New pattern.
7856 * config/nds32/nds32.opt (mrelax-hint): New option.
7857 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
7859 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
7861 * config/nds32/t-nds32: Modify files dependency.
7863 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7865 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
7866 (PROFILE_HOOK): Define its implementation.
7868 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7870 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
7871 type and 32-bit size.
7873 2018-04-01 Jakub Jelinek <jakub@redhat.com>
7876 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
7877 (V_128_256): New mode iterator.
7878 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
7879 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
7880 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
7882 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
7883 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
7885 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
7888 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
7889 NaN inputs correctly.
7891 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
7894 * config/rs6000/vsx.md (??r): New mode attribute.
7895 (*vsx_mov<mode>_64bit): Use it.
7896 (*vsx_mov<mode>_32bit): Likewise.
7898 2018-03-30 Martin Sebor <msebor@redhat.com>
7900 PR tree-optimization/84818
7901 * builtins.c (check_access): Use warning_n.
7903 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7906 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
7908 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
7911 2018-03-30 Julia Koval <julia.koval@intel.com>
7914 * x86-tune.def (movx, partial_reg_dependency): Enable for
7917 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
7920 * lra-constraints.c (process_alt_operands): Move setting
7921 this_alternative_matches below.
7923 2018-03-29 Martin Liska <mliska@suse.cz>
7926 * doc/invoke.texi: Document how LTO works with debug info.
7927 Describe auto-load support of binutils. Mention 'x86-64'
7928 as valid option value of -march option.
7930 2018-03-29 Jakub Jelinek <jakub@redhat.com>
7932 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
7935 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
7936 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
7937 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
7940 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7943 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
7944 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
7945 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
7946 for RS6000_BTM_POWERPC64.
7947 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
7948 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
7949 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
7954 2018-03-28 Carl Love <cel@us.ibm.com>
7957 2017-09-27 Carl Love <cel@us.ibm.com>
7959 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
7960 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
7961 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
7964 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7966 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
7967 instead of __vector bool.
7968 (_mm_max_pu8): Likewise.
7969 (_mm_min_pi16): Likewise.
7971 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7974 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
7975 (DIVWEUO): Likewise.
7977 (DIVDEUO): Likewise.
7978 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
7979 DIVWEUO and DIVDEUO.
7980 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
7981 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
7982 (div_extend): Likewise.
7983 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
7985 (__builtin_divweuo): Likewise.
7986 (__builtin_divdeo): Likewise.
7987 (__builtin_divdeuo): Likewise.
7989 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7992 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
7993 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
7995 PR tree-optimization/82004
7996 * gimple-match-head.c (optimize_pow_to_exp): New function.
7997 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
7998 Don't fold to exp if optimize_pow_to_exp is false.
8000 2018-03-28 Martin Liska <mliska@suse.cz>
8003 * calls.c (initialize_argument_information): Fix trailing space.
8004 * common.opt: Fix typo and provide better explanation for
8005 -fsanitize-coverage option.
8006 * config/i386/i386.opt: Fix typo.
8008 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8009 Martin Liska <mliska@suse.cz>
8012 * gimplify.c (asan_poison_variable): Don't do the check for
8013 gimplify_omp_ctxp here.
8014 (gimplify_decl_expr): Do it here.
8015 (gimplify_target_expr): Likewise.
8017 2018-03-28 Martin Liska <mliska@suse.cz>
8020 * config/i386/i386.c (ix86_function_arg_advance): Do not call
8021 chkp_type_bounds_count if MPX is not enabled.
8023 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
8025 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
8027 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
8030 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
8031 function to create the function decl for complex long double
8032 multiply and divide for -mabi=ieeelongdouble.
8033 (init_float128_ieee): Call it.
8035 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
8038 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
8039 -fcf-protection=branch -mibt.
8040 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
8042 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8045 * config/arm/arm.c (arm_valid_symbolic_address): Handle
8046 arm_word_relocations.
8048 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
8051 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
8052 extern array declarations.
8054 2018-03-27 Richard Biener <rguenther@suse.de>
8057 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
8058 explicit single_use checks.
8060 2018-03-27 Richard Biener <rguenther@suse.de>
8062 PR tree-optimization/85082
8063 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
8066 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8068 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
8069 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
8070 Turn on fasynchronous-unwind-tables and funwind-tables.
8072 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
8075 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
8076 (*bmi_blsr_<mode>_ccz): Ditto.
8078 2018-03-26 Tom de Vries <tom@codesourcery.com>
8080 PR tree-optimization/85063
8081 * omp-general.c (offloading_function_p): New function. Factor out
8083 * omp-offload.c (pass_omp_target_link::gate): ... here.
8084 * omp-general.h (offloading_function_p): Declare.
8085 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
8086 with attribute omp declare target for offloading functions.
8088 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
8090 PR tree-optimization/84005
8091 * tree-data-ref.h (get_base_for_alignment): Declare.
8092 * tree-data-ref.c (get_base_for_alignment_1): New function.
8093 (get_base_for_alignment): Likewise.
8094 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
8095 get_base_for_alignment to find a suitable base object, instead
8096 of always using drb->base_address.
8098 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8101 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
8102 known size by default.
8104 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
8107 * lra-constraints.c (process_alt_operands): Don't match BLKmode
8108 and non BLKmode operands.
8110 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8113 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
8114 Clean up attributes.
8116 2018-03-23 Richard Biener <rguenther@suse.de>
8119 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
8120 we are going to emit early debug for LTO.
8122 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8125 * function.c (match_asm_constraints_1): Don't optimize if input
8126 doesn't satisfy general_operand predicate for output's mode.
8129 * alias.c (write_dependence_p): Don't require for x_canonicalized
8130 non-VOIDmode if x has VOIDmode.
8133 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
8134 just don't try to optimize it rather than assert it never happens.
8136 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8138 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
8139 macro expansions for definition of ST_INTERNAL_<mode> and
8140 LD_INTERNAL_<mode> builtins.
8141 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
8143 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
8145 (altivec_expand_st_builtin): Likewise.
8146 (altivec_expand_builtin): Remove calls to deleted functions.
8147 (rs6000_address_for_altivec): Delete this function.
8148 * config/rs6000/vector.md: Remove expands for
8149 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
8151 2018-03-22 Sudakshina Das <sudi.das@arm.com>
8154 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
8155 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
8156 re-computing once computed.
8157 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
8158 (arm_init_machine_status): Initialize
8159 machine->static_chain_stack_bytes.
8161 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8164 * doc/extend.texi: Add four new prototypes for vec_ld.
8165 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
8166 definitions for more logical presentation.
8167 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
8168 entries for V1TI variants of __builtin_altivec_ld builtin.
8169 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
8170 handling of V1TI variant of LVX icode pattern.
8171 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
8172 (rs6000_gimple_fold_builtin): Likewise.
8173 (altivec_init_builtins): Add code to define
8174 __builtin_altivec_lvx_v1ti function.
8176 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8179 * function.c (match_asm_constraints_1): Don't do the optimization
8180 if input isn't a REG, SUBREG, MEM or constant.
8182 2018-03-22 Tom de Vries <tom@codesourcery.com>
8184 PR tree-optimization/84956
8185 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
8186 bb_has_abnormal_pred.
8188 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8191 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
8192 DECL_INITIAL (decl) to decl at the end.
8193 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
8196 2018-03-21 Joseph Myers <joseph@codesourcery.com>
8198 * doc/extend.texi (__builtin_tgmath): Document when complex
8199 integer types are treated as _Complex _Float64.
8201 2018-03-21 Tom de Vries <tom@codesourcery.com>
8203 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
8205 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8207 PR tree-optimization/84960
8208 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
8209 if it is ENTRY block, move them into single succ of ENTRY in that case.
8211 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
8213 PR tree-optimization/84811
8214 * poly-int.h (poly_span_traits): Remove the T3 parameter and
8215 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
8216 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
8217 (known_subrange_p): Update accordingly. Cast each value involved
8218 in the size comparison, rather than casting the result of the
8221 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8223 PR tree-optimization/84982
8224 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
8225 by flipping the least significant bit rather than all bits from
8226 bitpos to bitpos + bitsize - 1.
8228 2018-03-21 Nathan Sidwell <nathan@acm.org>
8230 * doc/extend.texi (Deprecated Features): Remove mention of
8231 long-deleted deprecations.
8233 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8236 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
8237 * configure: Regenerate.
8239 2018-03-21 Tom de Vries <tom@codesourcery.com>
8241 PR tree-optimization/83126
8242 * tree-parloops.c (num_phis): New function.
8243 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
8245 2018-03-21 Nathan Sidwell <nathan@acm.org>
8247 * doc/extend.texi (Deprecated Features): Update deprecated flags,
8248 mention anon-struct/union members and trailing attributes.
8250 2018-03-21 Bin Cheng <bin.cheng@arm.com>
8252 PR tree-optimization/84969
8253 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
8254 builtin memset partitions if they set different rhs values.
8256 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8258 PR rtl-optimization/84989
8259 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
8260 VEC_DUPLICATE with scalar result mode.
8262 2018-03-21 Martin Liska <mliska@suse.cz>
8265 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
8266 not intended return statement.
8268 2018-03-21 Martin Liska <mliska@suse.cz>
8271 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
8272 (chkp_find_bound_slots_1): Limit number of iterations.
8274 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
8277 * Minor grammar fixes for x86 options.
8279 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8282 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
8283 holding REG_CFA_RESTORE notes, instead turn them into a USE.
8285 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
8288 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
8289 (altivec_lvx_<mode>_1op): Likewise.
8290 (altivec_stvx_<mode>_2op): Likewise.
8291 (altivec_stvx_<mode>_1op): Likewise.
8292 (altivec_lvx_<VM2:mode>): New define_expand.
8293 (altivec_stvx_<VM2:mode>): Likewise.
8294 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
8295 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8296 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
8297 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8298 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
8299 (rs6000_gen_lvx): Likewise.
8300 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
8301 (altivec_expand_stv_builtin): Likewise.
8302 (altivec_expand_builtin): Likewise.
8303 * config/rs6000/vector.md: Likewise.
8305 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8308 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
8311 2018-03-20 Richard Biener <rguenther@suse.de>
8314 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
8315 sign-conversions as zero, fall back to standard scalar_stmt
8318 2018-03-20 Martin Liska <mliska@suse.cz>
8321 * predict.c (rebuild_frequencies): Handle case when we have
8322 PROFILE_ABSENT, but flag_guess_branch_prob is false.
8324 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8327 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
8328 flag_section_anchors.
8329 * varasm.c (use_blocks_for_decl_p): Remove hack for
8330 dw2_force_const_mem.
8333 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
8335 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
8336 be created, use lowpart_subreg of operands[0] rather than operands[0]
8338 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
8339 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
8340 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
8341 and n constraint instead of aarch64_shift_imm_di and Usd.
8342 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
8343 (*aarch64_<optab>_reg_minus<mode>3): ... this.
8345 2018-03-20 Sudakshina Das <sudi.das@arm.com>
8348 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
8349 to favor GPR over NEON registers.
8350 (<shift>di3_neon): Likewise.
8352 2018-03-20 Tom de Vries <tom@codesourcery.com>
8355 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
8356 (nvptx_process_pars): Emit bar.sync asap and alap.
8358 2018-03-20 Tom de Vries <tom@codesourcery.com>
8361 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
8362 seen_label if seen_label is already set.
8364 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8367 * config/i386/i386.c (fold_builtin_cpu): For features above 31
8368 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
8369 Use 1U instead of 1. Formatting fixes.
8372 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
8373 instead of TREE_TYPE (s1) for the return value.
8375 2018-03-19 Jakub Jelinek <jakub@redhat.com>
8377 PR tree-optimization/84946
8378 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
8379 bitsize + bitsize in poly_uint64 rather than poly_int64.
8382 * dwarf2asm.c: Include fold-const.c.
8383 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
8384 of decl rather than decl itself.
8386 PR rtl-optimization/84643
8387 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
8389 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
8392 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
8393 calling assemble_variable.
8395 2018-03-19 Sudakshina Das <sudi.das@arm.com>
8398 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
8399 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
8401 2018-03-19 Jim Wilson <jimw@sifive.com>
8404 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
8405 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
8406 (riscv_first_stack_step): Likewise.
8407 (riscv_option_override): Use STACK_BOUNDARY instead of
8409 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
8411 (BIGGEST_ALIGNMENT): Set to 128.
8412 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
8413 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
8416 2018-03-19 Richard Biener <rguenther@suse.de>
8418 PR tree-optimization/84933
8419 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
8420 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
8422 2018-03-19 Richard Biener <rguenther@suse.de>
8424 PR tree-optimization/84859
8425 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
8426 (cond_if_else_store_replacement): Perform sinking operation on
8427 single-store BBs regardless of MAX_STORES_TO_SINK setting.
8428 Generalize what a BB with a single eligible store is.
8430 2018-03-19 Richard Biener <rguenther@suse.de>
8432 PR tree-optimization/84929
8433 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
8434 chrec_is_positive against non-chrec arg.
8436 2018-03-19 Tamar Christina <tamar.christina@arm.com>
8439 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
8441 2018-03-18 Martin Liska <mliska@suse.cz>
8443 PR rtl-optimization/84635
8444 * regrename.c (build_def_use): Use matches_mode only when
8447 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
8449 PR tree-optimization/84913
8450 * tree-vect-loop.c (vectorizable_reduction): Don't try to
8451 vectorize chains of COND_EXPRs.
8453 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8455 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
8457 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8459 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
8461 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8463 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
8465 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8466 Kito Cheng <kito.cheng@gmail.com>
8468 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
8469 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
8470 (nds32_adjust_reg_alloc_order): New function.
8471 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
8473 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
8475 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
8476 nds32_print_operand, nds32_print_operand_address): Use
8477 HOST_WIDE_INT_PRINT_DEC instead.
8479 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8481 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
8483 2018-03-17 Jakub Jelinek <jakub@redhat.com>
8486 * config/i386/i386.c (initial_ix86_tune_features,
8487 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
8489 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
8490 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
8491 rather than 1u << ix86_tune. Formatting fix.
8492 (ix86_option_override_internal): Change ix86_arch_mask from
8493 unsigned int to unsigned HOST_WIDE_INT, initialize to
8494 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
8495 (ix86_function_specific_restore): Likewise.
8497 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8500 * postreload.c (reload_combine_recognize_pattern): Perform
8501 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
8502 truncate_int_for_mode the result for the destination's mode.
8505 * hsa-gen.c (mem_type_for_type): Fix comment typo.
8506 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
8508 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8511 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
8514 * lra-assigns.c (lra_split_hard_reg_for): Don't use
8515 regno_allocno_class_array and sorted_pseudos.
8516 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
8517 insns where regno is used.
8519 2018-03-16 Martin Liska <mliska@suse.cz>
8522 * multiple_target.c (create_dispatcher_calls): Redirect
8523 reference in the symbol table.
8525 2018-03-16 Martin Liska <mliska@suse.cz>
8528 * multiple_target.c (create_dispatcher_calls): Redirect also
8531 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8535 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
8536 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
8537 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
8539 2018-03-16 Julia Koval <julia.koval@intel.com>
8541 * doc/invoke.texi (Skylake Server): Add CLWB.
8542 Cannonlake): Remove CLWB.
8544 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8546 PR tree-optimization/84841
8547 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
8549 (FLOAT_ONE_CONST_TYPE): Define.
8550 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
8551 (sort_by_operand_rank): Put entries with higher constant_type last
8552 rather than first to match comments.
8554 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
8556 * config/nios2/nios2.md (movsi_internal): Fix thinko in
8559 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8562 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
8563 check and use address of target always.
8565 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
8568 * config/i386/i386.c (indirect_thunk_needed): Update comments.
8569 (indirect_thunk_bnd_needed): Likewise.
8570 (indirect_thunks_used): Likewise.
8571 (indirect_thunks_bnd_used): Likewise.
8572 (indirect_return_needed): New.
8573 (indirect_return_bnd_needed): Likewise.
8574 (output_indirect_thunk_function): Add a bool argument for
8576 (output_indirect_thunk_function): Don't generate alias for
8577 function return thunk.
8578 (ix86_code_end): Call output_indirect_thunk_function to generate
8579 function return thunks.
8580 (ix86_output_function_return): Set indirect_return_bnd_needed
8581 and indirect_return_needed instead of indirect_thunk_bnd_needed
8582 and indirect_thunk_needed.
8584 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
8586 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
8587 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
8588 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
8590 2018-03-15 David Malcolm <dmalcolm@redhat.com>
8591 Paul Hua <paul.hua.gm@gmail.com>
8594 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
8596 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
8598 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
8599 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
8602 2018-03-15 Tamar Christina <tamar.christina@arm.com>
8605 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8606 instead of GET_MODE_SIZE when comparing Units.
8608 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
8611 * varasm.c (hash_section): Return an unchangeble hash value
8612 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
8613 Return !aarch64_can_use_per_function_literal_pools_p ().
8615 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8618 * optabs.c (emit_conditional_move): Pass address of cmode's copy
8619 rather than address of cmode as last argument to prepare_cmp_insn.
8621 2018-03-15 Julia Koval <julia.koval@intel.com>
8623 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
8624 F_AVX512VNNI, F_AVX512BITALG): New.
8626 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
8629 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
8630 insn for floating-point loads and stores.
8632 2018-03-14 Carl Love <cel@us.ibm.com>
8634 * config/rs6000/rs6000-c.c: Add macro definitions for
8635 ALTIVEC_BUILTIN_VEC_PERMXOR.
8636 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
8637 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
8638 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
8639 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
8641 * config/doc/extend.texi: Add prototypes for vec_permxor.
8643 2018-03-14 David Malcolm <dmalcolm@redhat.com>
8646 * diagnostic-show-locus.c (class layout_point): Convert m_line
8647 from int to linenum_type.
8648 (line_span::comparator): Use linenum "compare" function when
8649 comparing line numbers.
8650 (test_line_span): New function.
8651 (layout_range::contains_point): Convert param "row" from int to
8653 (layout_range::intersects_line_p): Likewise.
8654 (layout::will_show_line_p): Likewise.
8655 (layout::print_source_line): Likewise.
8656 (layout::should_print_annotation_line_p): Likewise.
8657 (layout::print_annotation_line): Likewise.
8658 (layout::print_leading_fixits): Likewise.
8659 (layout::annotation_line_showed_range_p): Likewise.
8660 (struct line_corrections): Likewise for field m_row.
8661 (line_corrections::line_corrections): Likewise for param "row".
8662 (layout::print_trailing_fixits): Likewise.
8663 (layout::get_state_at_point): Likewise.
8664 (layout::get_x_bound_for_row): Likewise.
8665 (layout::print_line): Likewise.
8666 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
8667 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
8668 * input.c (selftest::test_linenum_comparisons): New function.
8669 (selftest::input_c_tests): Call it.
8670 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
8671 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
8672 * selftest.h (ASSERT_GT): New macro.
8673 (ASSERT_GT_AT): New macro.
8674 (ASSERT_LT): New macro.
8675 (ASSERT_LT_AT): New macro.
8677 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
8679 PR rtl-optimization/84780
8680 * combine.c (distribute_links): Don't make a link based on pc_rtx.
8682 2018-03-14 Martin Liska <mliska@suse.cz>
8684 * tree.c (record_node_allocation_statistics): Use
8685 get_stats_node_kind.
8686 (get_stats_node_kind): New function extracted from
8687 record_node_allocation_statistics.
8688 (free_node): Use get_stats_node_kind.
8690 2018-03-14 Richard Biener <rguenther@suse.de>
8692 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
8693 that the value-set of ANTIC_IN doesn't grow.
8696 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8698 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8699 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8701 2018-03-14 Julia Koval <julia.koval@intel.com>
8703 * config.gcc (icelake-client, icelake-server): New.
8705 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
8706 (initial_ix86_arch_features): Ditto.
8707 (PTA_SKYLAKE): Add SGX.
8708 (PTA_ICELAKE): Remove.
8709 (PTA_ICELAKE_CLIENT): New.
8710 (PTA_ICELAKE_SERVER): New.
8711 (ix86_option_override_internal): Split up icelake on icelake client and
8713 (get_builtin_code_for_version): Ditto.
8714 (fold_builtin_cpu): Ditto.
8715 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
8716 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
8717 * config/i386/i386.h (processor_type): Ditto.
8718 * doc/invoke.texi: Ditto.
8720 2018-03-14 Jakub Jelinek <jakub@redhat.com>
8723 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
8724 INTEGER_CST offset, add it together with bitpos / 8 and
8725 sign extend based on POINTER_SIZE.
8729 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
8732 * config/i386/constraints.md (Yc): New register constraint.
8733 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
8734 Use Yc constraint for alternative 2 of operand 0. Remove
8735 preferred_for_speed attribute.
8737 2018-03-14 Richard Biener <rguenther@suse.de>
8739 PR tree-optimization/84830
8740 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
8741 with the old one to avoid oscillations.
8743 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
8746 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
8748 (assign_by_spills): Return a flag of reload assignment failure.
8749 Do not process the reload assignment failures. Do not spill other
8750 reload pseudos if they has the same reg class. Update n if
8752 (lra_assign): Add a return arg. Set up from the result of
8753 assign_by_spills call.
8754 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8755 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8756 usage_insns if it is not NULL.
8757 (spill_hard_reg_in_range): New function.
8758 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8759 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8760 function prototypes.
8761 (lra_assign): Change prototype.
8762 * lra.c (lra): Add code to deal with fails by splitting hard reg
8765 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
8767 * config/riscv/riscv.opt (mrelax): New option.
8768 * config/riscv/riscv.c (riscv_file_start): Emit ".option
8769 "norelax" when riscv_mrelax is disabled.
8770 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
8772 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8775 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
8776 reassociation for int modes.
8778 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8780 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
8781 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
8783 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
8784 * config/aarch64/aarch64-sve.md
8785 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
8786 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
8787 (*extend<mode><Vwide>2): Rename to...
8788 (aarch64_sve_extend<mode><Vwide>2): ...this.
8789 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
8790 renaming the old pattern to...
8791 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
8793 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
8794 define_expand, renaming the old pattern to...
8795 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
8796 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
8797 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
8798 account when deciding which SVE instruction the optab should use.
8799 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
8801 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8803 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
8804 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
8805 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
8806 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
8807 (tlsdesc_small_<mode>): Turn a define_expand and use
8808 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
8809 (tlsdesc_small_advsimd_<mode>): ...this.
8810 (tlsdesc_small_sve_<mode>): New pattern.
8812 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8814 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
8815 (UNSPEC_UMUL_HIGHPART): New constants.
8816 (MUL_HIGHPART): New int iteraor.
8817 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
8818 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
8820 (*<su>mul<mode>3_highpart): New define_insn.
8822 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8825 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8828 2018-03-13 Martin Liska <mliska@suse.cz>
8831 * (sem_item_optimizer::sem_item_optimizer): Initialize new
8833 (sem_item_optimizer::~sem_item_optimizer): Release it.
8834 (sem_item_optimizer::merge_classes): Register variable aliases.
8835 (sem_item_optimizer::fixup_pt_set): New function.
8836 (sem_item_optimizer::fixup_points_to_sets): Likewise.
8837 * ipa-icf.h: Declare new variables and functions.
8839 2018-03-13 Jakub Jelinek <jakub@redhat.com>
8842 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
8843 integer_pow2p@2 and test integer_pow2p in condition.
8844 (A < 0 ? C : 0): Similarly for @1.
8847 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
8848 characters starting at p contain '\0' character, don't look beyond
8852 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
8853 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
8856 * reg-stack.c (change_stack): Change update_end var from int to
8857 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
8858 also call set_block_for_insn on the newly added insns and rescan.
8861 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
8862 on the last operand.
8865 * tree.c (stabilize_reference_1): Return save_expr (e) for
8866 STATEMENT_LIST even if it doesn't have side-effects.
8868 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
8870 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
8872 2018-03-12 Renlin Li <renlin.li@arm.com>
8874 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
8875 aarch64_output_scalar_simd_mov_immediate.
8877 2018-03-12 Martin Sebor <msebor@redhat.com>
8879 PR tree-optimization/83456
8880 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
8881 for perfectly overlapping calls to memcpy.
8882 (gimple_fold_builtin_memory_chk): Same.
8883 (gimple_fold_builtin_strcpy): Handle no-warning.
8884 (gimple_fold_builtin_stxcpy_chk): Same.
8885 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
8887 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8889 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
8890 parameter. Use it for SFmode.
8891 (rs6000_function_arg_advance_1): Adjust.
8892 (rs6000_function_arg): Adjust.
8893 (rs6000_gimplify_va_arg): Pass false for that new parameter.
8895 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8897 PR rtl-optimization/84169
8898 PR rtl-optimization/84780
8899 * combine.c (can_combine_p): Check for a 2-insn combination whether
8900 the destination register is used between the two insns, too.
8902 2018-03-12 Richard Biener <rguenther@suse.de>
8904 PR tree-optimization/84803
8905 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
8906 for refs DR analysis didn't process.
8908 2018-03-12 Richard Biener <rguenther@suse.de>
8910 PR tree-optimization/84777
8911 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
8912 force-vectorize loops ignore whether we are optimizing for size.
8914 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
8916 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
8917 (TARGET_MD_ASM_ADJUST): Define.
8919 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
8920 Kito Cheng <kito.cheng@gmail.com>
8921 Chung-Ju Wu <jasonwucj@gmail.com>
8923 * config/nds32/nds32.c (nds32_compute_stack_frame,
8924 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
8925 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
8926 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
8927 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
8928 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
8929 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
8930 * config/nds32/nds32.md (prologue, epilogue): Use macro
8931 NDS32_V3PUSH_AVAILABLE_P to do checking.
8933 2018-03-11 Jakub Jelinek <jakub@redhat.com>
8936 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
8937 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
8938 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
8939 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
8940 addition of most attributes on !orig_type_die or the attribute not
8941 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
8943 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8944 Chung-Ju Wu <jasonwucj@gmail.com>
8946 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
8948 * config/nds32/nds32.opt (mvh): New option.
8950 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8951 Chung-Ju Wu <jasonwucj@gmail.com>
8953 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
8955 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
8956 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
8959 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8960 Chung-Ju Wu <jasonwucj@gmail.com>
8962 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
8964 * config/nds32/nds32-multiple.md (strlensi): New pattern.
8965 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
8967 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
8968 Kito Cheng <kito.cheng@gmail.com>
8969 Chung-Ju Wu <jasonwucj@gmail.com>
8971 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
8972 UNSPEC_FFMISM and UNSPEC_FLMISM.
8973 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
8974 for ffb, ffmism and flmism.
8975 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
8976 (unspec_ffmism): Ditto.
8977 (unspec_flmism): Ditto.
8978 (nds32_expand_builtin_impl): Check if string extension is available.
8979 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
8980 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
8982 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
8985 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8988 * lra-assigns.c (assign_by_spills): Return a flag of reload
8989 assignment failure. Do not process the reload assignment
8990 failures. Do not spill other reload pseudos if they has the same
8992 (lra_assign): Add a return arg. Set up from the result of
8993 assign_by_spills call.
8994 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8995 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8996 usage_insns if it is not NULL.
8997 (spill_hard_reg_in_range): New function.
8998 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8999 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9000 function prototypes.
9001 (lra_assign): Change prototype.
9002 * lra.c (lra): Add code to deal with fails by splitting hard reg
9005 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
9008 * config/i386/i386.opt: Replace Enforcment with Enforcement.
9010 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
9013 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
9014 (dw_val_node): Add val_symbolic_view.
9015 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
9016 (symview_upper_bound): New.
9017 (new_line_info_table): Initialize symviews_since_reset.
9018 (dwarf2out_source_line): Count symviews_since_reset and set
9019 symview_upper_bound.
9020 (dw_val_equal_p): Handle symview.
9021 (add_AT_symview): New.
9022 (print_dw_val): Handle symview.
9023 (attr_checksum, attr_checksum_ordered): Likewise.
9024 (same_dw_val_p, size_of_die): Likewise.
9025 (value_format, output_die): Likewise.
9026 (add_high_low_attributes): Use add_AT_symview for entry_view.
9027 (dwarf2out_finish): Reset symview_upper_bound, clear
9030 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
9033 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
9034 Add strict argument and use it.
9035 (rs6000_split_multireg_move): Update for new strict argument.
9036 (mem_operand_gpr): Disallow all non-offsettable addresses.
9037 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
9039 2018-03-09 Jakub Jelinek <jakub@redhat.com>
9042 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
9043 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
9044 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
9047 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
9048 decl, use remap_type if we want to use the type.
9050 2018-03-09 Martin Sebor <msebor@redhat.com>
9052 PR tree-optimization/84526
9053 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9055 (builtin_access::generic_overlap): Be prepared to handle non-array
9058 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
9060 PR rtl-optimization/84682
9061 * lra-constraints.c (process_address_1): Check is_address flag
9062 for address constraints.
9063 (process_alt_operands): Likewise.
9064 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
9065 preprocess_constraints.
9066 * recog.h (preprocess_constraints): Add oploc parameter.
9068 * recog.c (preprocess_constraints): Test address_operand for
9069 CT_ADDRESS constraints.
9071 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9074 * lra-assigns.c (assign_by_spills): Return a flag of reload
9075 assignment failure. Do not process the reload assignment
9076 failures. Do not spill other reload pseudos if they has the same
9078 (lra_assign): Add a return arg. Set up from the result of
9079 assign_by_spills call.
9080 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9081 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9082 usage_insns if it is not NULL.
9083 (spill_hard_reg_in_range): New function.
9084 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9085 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9086 function prototypes.
9087 (lra_assign): Change prototype.
9088 * lra.c (lra): Add code to deal with fails by splitting hard reg
9091 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9094 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
9095 Accept complain bool parameter. Only emit errors if it is true.
9096 (arm_parse_cpu_option_name): Likewise.
9097 (arm_target_thumb_only): Adjust callers of the above.
9098 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
9099 prototype to take a default true bool parameter.
9100 (arm_parse_arch_option_name): Likewise.
9102 2018-03-09 David Malcolm <dmalcolm@redhat.com>
9103 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
9107 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
9108 * configure: Regenerate.
9109 * configure.ac ("linker --version-script option"): New.
9110 ("linker soname option"): New.
9112 2018-03-09 Richard Biener <rguenther@suse.de>
9114 PR tree-optimization/84775
9115 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
9116 immediate uses of predicate stmts and mark them modified.
9119 PR tree-optimization/84178
9120 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9122 (version_loop_for_if_conversion): Delay update_ssa call.
9123 (tree_if_conversion): Delay update_ssa until after predicate
9126 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
9129 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
9130 when the function accesses prior frames.
9132 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9135 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
9136 gen_llsym, otherwise call maybe_gen_llsym.
9139 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
9140 has ',' character inside of it.
9142 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9145 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
9146 as clobbering CC_REGNUM.
9148 2018-03-08 Richard Biener <rguenther@suse.de>
9151 * tree-scalar-evolution.c: Include tree-into-ssa.h.
9152 (follow_copies_to_constant): Do not follow SSA names registered
9155 2018-03-08 Richard Biener <rguenther@suse.de>
9157 PR tree-optimization/84178
9158 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9160 (version_loop_for_if_conversion): Delay update_ssa call.
9161 (tree_if_conversion): Delay update_ssa until after predicate
9164 2018-03-08 David Malcolm <dmalcolm@redhat.com>
9166 PR tree-optimization/84178
9167 * tree-if-conv.c (release_bb_predicate): Remove the
9168 the assertion that the stmts have NULL use_ops.
9169 Discard the statements, asserting that they haven't
9170 yet been added to a BB.
9172 2018-03-08 Richard Biener <rguenther@suse.de>
9174 PR tree-optimization/84746
9175 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
9176 (phi_translate): Pass in destination ANTIC_OUT set.
9177 (phi_translate_1): Likewise. For a simplified result lookup
9178 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
9179 (phi_translate_set): Adjust.
9180 (do_pre_regular_insertion): Likewise.
9181 (do_pre_partial_partial_insertion): Likewise.
9183 2018-03-08 Martin Liska <mliska@suse.cz>
9185 PR gcov-profile/84735
9186 * doc/gcov.texi: Document usage of profile files.
9187 * gcov-io.h: Document changes in the format.
9189 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
9193 * dwarf2out.c (struct dw_line_info_table): Update comments for
9195 (FORCE_RESET_NEXT_VIEW): New.
9196 (FORCE_RESETTING_VIEW_P): New.
9197 (RESETTING_VIEW_P): Check for -1 too.
9198 (ZERO_VIEW_P): Likewise.
9199 (new_line_info_table): Force-reset next view.
9200 (dwarf2out_begin_function): Likewise.
9201 (dwarf2out_source_line): Simplify zero_view_p initialization.
9202 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
9203 view directly. Omit view when omitting .loc at line 0.
9205 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9207 PR tree-optimization/84740
9208 * tree-switch-conversion.c (process_switch): Call build_constructors
9209 only if info.phi_count is non-zero.
9211 PR tree-optimization/84739
9212 * tree-tailcall.c (find_tail_calls): Check call arguments against
9213 DECL_ARGUMENTS (current_function_decl) rather than
9214 DECL_ARGUMENTS (func) when checking for tail recursion.
9216 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9218 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
9219 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
9220 Volker Reichelt's entry and add entries for people that perform
9221 GCC fuzzy testing and report numerous bugs.
9223 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
9226 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
9227 readonly data in sdata, if that is disabled.
9228 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
9229 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
9230 -mreadonly-in-sdata option.
9232 2018-03-07 Martin Sebor <msebor@redhat.com>
9234 PR tree-optimization/84468
9235 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
9236 basic block when looking for nul assignment.
9238 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
9241 * except.h (output_function_exception_table): Adjust prototype.
9242 * except.c (output_function_exception_table): Remove FNNAME parameter
9243 and add SECTION parameter. Ouput one part of the table at a time.
9244 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
9245 the first part of the exception table and emit unwind directives.
9246 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
9247 (i386_pe_seh_cold_init): Likewise.
9248 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
9249 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
9250 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
9251 (ix86_output_call_insn): Emit a nop in one more case for SEH.
9252 * config/i386/winnt.c: Include except.h.
9253 (struct seh_frame_state): Add reg_offset, after_prologue and
9254 in_cold_section fields.
9255 (i386_pe_seh_end_prologue): Set seh->after_prologue.
9256 (i386_pe_seh_cold_init): New function.
9257 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
9258 to seh->in_cold_section.
9259 (seh_emit_push): Record the offset of the push.
9260 (seh_emit_save): Record the offet of the save.
9261 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
9262 Test seh->after_prologue to disregard the epilogue.
9263 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
9264 (i386_pe_end_cold_function): New function.
9266 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9269 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
9270 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
9273 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
9275 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
9276 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
9279 * multiple_target.c: Include tree-inline.h and intl.h.
9280 (expand_target_clones): Diagnose and fail if node->definition and
9281 !tree_versionable_function_p (node->decl).
9283 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
9285 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
9287 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
9288 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
9289 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
9291 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9294 * combine.c (try_combine): Use reg_or_subregno instead of handling
9295 just paradoxical SUBREGs and REGs.
9297 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
9299 * config/arc/arc.c (arc_finalize_pic): Remove function.
9300 (arc_must_save_register): We use single base PIC register, remove
9301 checks to save/restore the PIC register.
9302 (arc_expand_prologue): Likewise.
9303 * config/arc/arc-protos.h (arc_set_default_type_attributes):
9305 (arc_verify_short): Likewise.
9306 (arc_attr_type): Likewise.
9307 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
9308 (walk_stores): Likewise.
9309 (arc_address_cost): Make it static.
9310 (arc_verify_short): Likewise.
9311 (branch_dest): Likewise.
9312 (arc_attr_type): Likewise.
9313 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
9314 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
9315 (arc_final_prescan_insn): Remove inserting the nops due to
9316 hardware hazards. It is done in reorg step.
9317 (insn_length_variant_t): Remove.
9318 (insn_length_parameters_t): Likewise.
9319 (arc_insn_length_parameters): Likewise.
9320 (arc_get_insn_variants): Likewise.
9321 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
9323 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9326 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
9329 PR tree-optimization/84687
9330 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
9332 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
9334 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9336 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
9337 Rename to ppc_speculation_barrier.
9338 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
9339 __builtin_ppc_speculation_barrier.
9341 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9344 * combine.c (combine_simplify_rtx): Don't try to simplify if
9345 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
9348 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
9350 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
9351 to 32 bytes when compiling for POWER9.
9353 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9356 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
9357 regparm >= 3 with no arg reg available also for calls with
9358 flag_force_indirect_call. Pass decl to ix86_function_regparm.
9361 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
9363 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
9365 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
9368 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
9370 2018-03-05 Richard Biener <rguenther@suse.de>
9372 PR tree-optimization/84486
9373 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
9374 When inserting a __builtin_assume_aligned call set the LHS
9375 SSA name alignment info accordingly.
9377 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
9379 PR tree-optimization/84114
9380 * config/aarch64/aarch64.c (aarch64_reassociation_width)
9381 Avoid reassociation of FLOAT_MODE addition.
9383 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
9385 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
9386 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
9387 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
9388 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
9389 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
9390 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
9391 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
9393 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
9394 __builtin_ia32_wbinvd): New builtins.
9395 (SPECIAL_ARGS2): New.
9396 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
9397 (SPECIAL_ARGS2): New.
9398 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
9399 (ix86_valid_target_attribute_inner_p): Ditto.
9400 (ix86_init_mmx_sse_builtins): Add special_args2.
9401 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
9402 TARGET_WBNOINVD_P): New.
9403 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
9404 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
9405 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
9406 * config/i386/immintrin.h (_wbinvd): New intrinsic.
9407 * config/i386/pconfigintrin.h: New file.
9408 * config/i386/wbnoinvdintrin.h: Ditto.
9409 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
9411 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
9413 2018-03-05 Richard Biener <rguenther@suse.de>
9415 PR tree-optimization/84670
9416 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9418 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9419 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9420 (compute_antic_aux): Only assert the number of values in ANTIC_IN
9421 doesn't grow if all successors (recursively) were visited at least
9424 2018-03-05 Richard Biener <rguenther@suse.de>
9426 PR tree-optimization/84650
9427 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
9428 if executed in the loop pipeline.
9430 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
9432 * doc/configfiles.texi (Configuration Files): Move info about
9433 conditionalizing $target-protos.h to...
9434 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
9435 differs from $target-protos.h.
9437 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
9438 Chung-Ju Wu <jasonwucj@gmail.com>
9440 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
9441 * config/nds32/nds32-multiple.md (setmemsi): Define.
9442 * config/nds32/nds32-memory-manipulation.c
9443 (nds32_gen_dup_4_byte_to_word_value): New.
9444 (emit_setmem_word_loop): New.
9445 (emit_setmem_byte_loop): New.
9446 (nds32_expand_setmem_loop): New.
9447 (nds32_expand_setmem_loop_v3m): New.
9448 (nds32_expand_setmem_unroll): New.
9449 (nds32_expand_setmem): New.
9451 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9452 Chung-Ju Wu <jasonwucj@gmail.com>
9454 * config/nds32/nds32-memory-manipulation.c
9455 (nds32_emit_load_store): New.
9456 (nds32_emit_post_inc_load_store): New.
9457 (nds32_emit_mem_move): New.
9458 (nds32_emit_mem_move_block): New.
9459 (nds32_expand_movmemsi_loop_unknown_size): New.
9460 (nds32_expand_movmemsi_loop_known_size): New.
9461 (nds32_expand_movmemsi_loop): New.
9462 (nds32_expand_movmemsi_unroll): New.
9463 (nds32_expand_movmemqi): Rename ...
9464 (nds32_expand_movmemsi): ... to this.
9465 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
9466 (movmemsi): ... to this.
9467 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
9468 (nds32_expand_movmemsi): ... to this.
9470 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9471 Monk Chiang <sh.chiang04@gmail.com>
9472 Chung-Ju Wu <jasonwucj@gmail.com>
9474 * config/nds32/nds32-protos.h
9475 (nds32_expand_load_multiple): New arguments.
9476 (nds32_expand_store_multiple): Ditto.
9477 (nds32_valid_multiple_load_store): Rename ...
9478 (nds32_valid_multiple_load_store_p): ... to this.
9479 * config/nds32/nds32-memory-manipulation.c
9480 (nds32_expand_load_multiple): Refine implementation.
9481 (nds32_expand_store_multiple): Ditto.
9482 * config/nds32/nds32-multiple.md
9483 (load_multiple): Update nds32_expand_load_multiple interface.
9484 (store_multiple): Update nds32_expand_store_multiple interface.
9485 * config/nds32/nds32-predicates.c
9486 (nds32_valid_multiple_load_store): Rename ...
9487 (nds32_valid_multiple_load_store_p): ... to this and refine
9489 * config/nds32/predicates.md
9490 (nds32_load_multiple_and_update_address_operation): New predicate.
9491 (nds32_store_multiple_and_update_address_operation): New predicate.
9493 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9494 Chung-Ju Wu <jasonwucj@gmail.com>
9496 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
9497 (combo): New attribute.
9498 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
9500 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
9502 * config/nds32/nds32.opt: Change -mcmodel= default value.
9504 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
9505 Monk Chiang <sh.chiang04@gmail.com>
9506 Chung-Ju Wu <jasonwucj@gmail.com>
9508 * config/nds32/constants.md (unspec_element): New enum.
9509 * config/nds32/constraints.md (Umw): New constraint.
9510 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
9511 * config/nds32/nds32-intrinsic.md: Likewise.
9512 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
9513 (nds32_valid_smw_lwm_base_p): New.
9514 (nds32_output_smw_single_word): New.
9515 (nds32_output_lmw_single_word): New.
9516 (nds32_expand_unaligned_load): New.
9517 (nds32_expand_unaligned_store): New.
9518 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
9519 (nds32_output_smw_single_word): Declare.
9520 (nds32_output_lmw_single_word): Declare.
9521 (nds32_expand_unaligned_load): Declare.
9522 (nds32_expand_unaligned_store): Declare.
9523 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
9524 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
9525 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
9526 NDS32_BUILTIN_UASTORE_DW.
9527 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
9530 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
9531 Kito Cheng <kito.cheng@gmail.com>
9532 Chung-Ju Wu <jasonwucj@gmail.com>
9534 * config/nds32/nds32-intrinsic.c
9535 (nds32_expand_builtin_null_ftype_reg): Delete.
9536 (nds32_expand_builtin_reg_ftype_imm): Ditto.
9537 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
9538 (nds32_read_argument): New.
9539 (nds32_legitimize_target): Ditto.
9540 (nds32_legitimize_argument): Ditto.
9541 (nds32_check_constant_argument): Ditto.
9542 (nds32_expand_unop_builtin): Ditto.
9543 (nds32_expand_unopimm_builtin): Ditto.
9544 (nds32_expand_binop_builtin): Ditto.
9545 (nds32_builtin_decl_impl): Ditto.
9546 (builtin_description): Ditto.
9547 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
9548 (nds32_init_builtins_impl): Ditto.
9549 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
9550 (nds32_builtin_decl): New.
9551 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
9552 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
9554 2018-03-02 Jeff Law <law@redhat.com>
9556 * reorg.c (stop_search_p): Handle DEBUG_INSN.
9557 (redundant_insn, fill_simple_delay_slots): Likewise.
9558 (fill_slots_from_thread): Likewise.
9559 * resource.c (mark_referenced_resources): Likewise.
9560 (mark_set_resources, find_dead_or_set_registers): Likewise.
9562 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9564 * substring-locations.h (format_warning_va): Formatting fix for
9566 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
9568 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
9569 * substring-locations.c: Include intl.h.
9570 (format_warning_va): Turned into small wrapper around
9571 format_warning_n_va, renamed to ...
9572 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
9573 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
9575 (format_warning_at_substring_n): New function.
9576 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
9577 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
9578 format_warning_at_substring with just a shorter name instead of
9579 const function pointer.
9580 (fmtwarn_n): New function.
9581 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
9582 appropriate, get rid of all the fmtstr temporaries, move conditionals
9583 with G_() wrapped string literals directly into fmtwarn arguments,
9584 cast dir.len to (int), formatting fixes.
9586 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
9588 * doc/invoke.texi: Remove "Cilk Plus" references.
9590 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9591 Richard Biener <rguenther@suse.de>
9594 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
9595 for error or warning attributes if CALL_FROM_THUNK_P is set.
9598 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9601 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
9602 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
9605 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
9606 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
9608 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
9609 instead of -1U in last predictors element's probability member.
9611 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9614 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9615 arguments if they are comparable.
9617 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9619 PR tree-optimization/84634
9620 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
9621 masks and masked_loop_p with a single loop_masks, making sure it's
9622 null for bb vectorization.
9624 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9626 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9627 (vect_analyze_data_ref_access): Use loop->safe_len rather than
9628 loop->force_vectorize to check whether there is no alias.
9630 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9633 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
9635 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
9637 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
9638 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
9639 instead of a loop around prev_real_insn.
9640 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
9644 * config/i386/i386.c (ix86_print_operand): Use conditional
9645 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
9648 2018-03-02 Richard Biener <rguenther@suse.de>
9650 PR tree-optimization/84427
9651 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
9652 (bitmap_set_subtract_values): Rewrite to handle multiple
9655 (prune_clobbered_mems): Likewise.
9656 (phi_translate): Take edge instead of pred/phiblock.
9657 (phi_translate_1): Likewise.
9658 (phi_translate_set): Likewise. Insert all translated
9659 exprs for a value into the set, keeping possibly multiple
9660 expressions per value.
9661 (compute_antic_aux): Adjust for phi_translate changes.
9662 When intersecting union the expressions and prune those
9663 not in the final value set, keeping possibly multiple
9664 expressions per value. Do not use value-insertion
9665 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
9666 all expressions. Add verification that the value-sets
9667 only shrink during iteration.
9668 (compute_partial_antic_aux): Adjust for the phi_translate changes.
9669 (do_pre_regular_insertion): Likewise.
9670 (do_pre_partial_partial_insertion): Likewise.
9672 2018-03-02 Richard Biener <rguenther@suse.de>
9675 * config/darwin.c (saved_debug_info_level): New static global.
9676 (darwin_asm_lto_start): Disable debug info generation for LTO out.
9677 (darwin_asm_lto_end): Restore debug info generation settings.
9679 2018-03-01 Martin Liska <mliska@suse.cz>
9682 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
9685 2018-03-01 Richard Biener <rguenther@suse.de>
9688 * dwarf2out.c (gen_variable_die): Properly handle late VLA
9689 type annotation with LTO when debug was disabled at compile-time.
9691 2018-03-01 Matthew Fortune <mfortune@gmail.com>
9693 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
9695 (mips_final_postscan_insn): Likewise.
9697 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
9699 PR rtl-optimization/84528
9700 * alias.c (init_alias_target): Add commentary.
9701 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
9702 a unique base value if the frame pointer is not eliminated
9703 to the stack pointer.
9705 2018-03-01 Tom de Vries <tom@codesourcery.com>
9707 PR rtl-optimization/83327
9708 * lra-int.h (hard_regs_spilled_into): Declare.
9709 * lra.c (hard_regs_spilled_into): Define.
9710 (init_reg_info): Init hard_regs_spilled_into.
9711 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
9712 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
9713 (process_bb_lives): Handle hard_regs_spilled_into.
9714 (lra_create_live_ranges_1): Before doing liveness propagation, clear
9715 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
9717 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
9719 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
9720 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
9721 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
9722 * config/rs6000/aix72.h: New file.
9724 2018-02-28 Jakub Jelinek <jakub@redhat.com>
9726 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
9727 instead of warning_at with conditional singular and plural messages
9731 * stor-layout.c (update_alignment_for_field): For
9732 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
9733 && !DECL_PACKED (field), do the alignment update, just use
9734 only desired_align instead of MAX (type_align, desired_align)
9736 (place_field): Don't do known_align < desired_align handling
9737 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
9738 is non-NULL, instead do it after rli->prev_field handling and
9739 only if not within a bitfield word. For DECL_PACKED (field)
9740 use type_align of BITS_PER_UNIT.
9742 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
9744 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
9745 superfluous parentheses and trailing spaces.
9747 2018-02-28 Richard Biener <rguenther@suse.de>
9749 PR tree-optimization/84584
9750 * graphite-scop-detection.c (scop_detection::add_scop): Discard
9751 SCoPs with fake exit edge.
9753 2018-02-28 Martin Liska <mliska@suse.cz>
9756 * timevar.c (timer::print): Fix format to properly print 100%
9759 2018-02-28 Richard Biener <rguenther@suse.de>
9762 * genmatch.c (capture_info::walk_match): Do not mark
9763 captured expressions without operands as expr_p given
9764 they act more like predicates and should be subject to
9765 "lost tail" side-effect preserving.
9767 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
9769 PR rtl-optimization/81611
9770 * auto-inc-dec.c (attempt_change): Move dead note from
9771 mem_insn if it's the next use of regno
9772 (find_address): Take address use of reg holding
9773 non-incremented value. Add parm to limit search to the named
9775 (merge_in_block): Attempt to use a mem insn that is the next
9776 use of the original regno.
9778 2018-02-27 Martin Sebor <msebor@redhat.com>
9781 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
9782 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
9784 2018-02-27 Martin Sebor <msebor@redhat.com>
9786 PR translation/84207
9787 * diagnostic-core.h (warning_n, error_n, inform_n): Change
9788 n argument to unsigned HOST_WIDE_INT.
9789 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
9790 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
9791 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
9792 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
9794 2018-02-27 Richard Biener <rguenther@suse.de>
9796 PR tree-optimization/84512
9797 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
9798 Do not use the estimate returned from record_stmt_cost for
9799 the scalar iteration cost but sum properly using add_stmt_cost.
9801 2018-02-27 Richard Biener <rguenther@suse.de>
9803 PR tree-optimization/84466
9804 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9805 Adjust last change to less strictly validate use operands.
9807 2018-02-27 Martin Liska <mliska@suse.cz>
9809 PR gcov-profile/84548
9810 * gcov.c (process_file): Allow partial overlap and consider it
9811 also as group functions.
9812 (output_lines): Properly calculate range of lines for a group.
9814 2018-02-27 Martin Liska <mliska@suse.cz>
9816 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
9817 'ggc' suffixes. Change first column width.
9818 (timer::print): Fix formatting of the column.
9820 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
9822 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
9823 preserve inline entry blocks for the sake of debug inline
9824 entry point markers alone.
9825 (remove_unused_locals): Suggest in comments a better place to
9826 force the preservation of inline entry blocks that are
9827 otherwise unused, but do not preserve them.
9829 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9831 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
9833 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9836 * config/i386/constraints.md (Bs): Replace
9837 ix86_indirect_branch_register with
9838 TARGET_INDIRECT_BRANCH_REGISTER.
9840 * config/i386/i386.md (indirect_jump): Likewise.
9841 (tablejump): Likewise.
9842 (*sibcall_memory): Likewise.
9843 (*sibcall_value_memory): Likewise.
9844 Peepholes of indirect call and jump via memory: Likewise.
9845 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
9846 (*sibcall_value_GOT_32): Likewise.
9847 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9848 (GOT_memory_operand): Likewise.
9849 (call_insn_operand): Likewise.
9850 (sibcall_insn_operand): Likewise.
9851 (GOT32_symbol_operand): Likewise.
9852 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
9854 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9856 PR rtl-optimization/83496
9857 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
9858 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
9859 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
9860 redundant insn, if any.
9861 (relax_delay_slots): Likewise.
9862 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
9864 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
9866 PR tree-optimization/83965
9867 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
9868 that grouped statements are part of a reduction chain. Return
9869 true if the statement is not marked as a reduction itself but
9871 (vect_recog_dot_prod_pattern): Don't check whether the statement
9872 is part of a group here.
9873 (vect_recog_sad_pattern): Likewise.
9874 (vect_recog_widen_sum_pattern): Likewise.
9876 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9879 * final.c (rest_of_clean_state): Also look for calls inside sequences.
9881 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9884 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
9886 (ix86_output_indirect_function_return): New prototype.
9887 (ix86_split_simple_return_pop_internal): Likewise.
9888 * config/i386/i386.c (indirect_return_via_cx): New.
9889 (indirect_return_via_cx_bnd): Likewise.
9890 (indirect_thunk_name): Handle return va CX_REG.
9891 (output_indirect_thunk_function): Create alias for
9892 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
9893 (ix86_output_indirect_jmp): Remove the bool argument.
9894 (ix86_output_indirect_function_return): New function.
9895 (ix86_split_simple_return_pop_internal): Likewise.
9896 * config/i386/i386.md (*indirect_jump): Don't pass false
9897 to ix86_output_indirect_jmp.
9898 (*tablejump_1): Likewise.
9899 (simple_return_pop_internal): Change it to define_insn_and_split.
9900 Call ix86_split_simple_return_pop_internal to split it for
9902 (simple_return_indirect_internal): Call
9903 ix86_output_indirect_function_return instead of
9904 ix86_output_indirect_jmp.
9906 2018-02-26 Jakub Jelinek <jakub@redhat.com>
9909 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
9910 memset and value initialization afterwards.
9912 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
9914 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
9916 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9919 * common/config/aarch64/aarch64-common.c
9920 (aarch_option_optimization_table[]): Switch
9921 off fomit-frame-pointer
9923 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9924 Chung-Ju Wu <jasonwucj@gmail.com>
9926 * config/nds32/nds32-multiple.md (load_multiple): Disallow
9928 (store_multiple): Ditto.
9930 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9932 * config.gcc: Add --with-cpu support for nds32 target.
9933 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
9934 * config/nds32/nds32.opt: Add -mcpu= option.
9936 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
9938 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
9939 isel=yes): Warn for these deprecated options.
9941 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
9943 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
9944 ISA_2_5_MASKS_EMBEDDED.
9946 2018-02-23 Jakub Jelinek <jakub@redhat.com>
9948 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
9949 p->max as pointers rather than using iterative_hash_expr.
9951 2018-02-23 Carl Love <cel@us.ibm.com>
9953 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
9954 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
9956 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
9957 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
9958 P8V_BUILTIN_VEC_VUNSIGNED2.
9960 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
9963 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
9964 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
9966 * lra-constraints.c (curr_insn_transform): Set up
9967 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
9969 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
9970 * lra-eliminations.c (spill_pseudos): Ditto.
9971 (process_insn_for_elimination): Ditto.
9972 * lra-lives.c (reg_early_clobber_p): Use the new macros.
9973 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
9974 LRA_NON_CLOBBERED_ALT.
9976 2018-02-22 Martin Sebor <msebor@redhat.com>
9978 PR tree-optimization/84480
9979 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
9980 to maybe_diag_stxncpy_trunc. Call it.
9981 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
9982 from gimple_fold_builtin_strcpy. Print inlining stack.
9983 (handle_builtin_stxncpy): Print inlining stack.
9984 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
9986 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
9989 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
9990 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
9991 and -fcheck-pointer-bounds are used together.
9992 (indirect_thunk_prefix): New enum.
9993 (indirect_thunk_need_prefix): New function.
9994 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
9995 "_nt" instead of "_bnd" for NOTRACK prefix.
9996 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
9997 (output_indirect_thunk_function): Likewise.
9999 (ix86_code_end): Update output_indirect_thunk_function calls.
10000 (ix86_output_indirect_branch_via_reg): Replace
10001 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
10002 (ix86_output_indirect_branch_via_push): Likewise.
10003 (ix86_output_function_return): Likewise.
10004 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
10005 incompatible with -fcf-protection=branch and
10006 -fcheck-pointer-bounds.
10008 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10011 * config/aarch64/aarch64.c (aarch64_print_address_internal):
10012 Change gcc_assert call to output_operand_lossage.
10014 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10016 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
10018 2018-02-22 DJ Delorie <dj@redhat.com>
10019 Sebastian Perta <sebastian.perta@renesas.com>
10020 Oleg Endo <olegendo@gcc.gnu.org>
10022 * config/rx/rx.c (rx_rtx_costs): New function.
10023 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
10025 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
10027 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
10029 2018-02-22 Martin Liska <mliska@suse.cz>
10032 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
10033 Add "native" as a possible value.
10035 2018-02-22 Martin Liska <mliska@suse.cz>
10038 * config/i386/i386.c (ix86_option_override_internal):
10039 Add "native" as a possible value for -march and -mtune.
10041 2018-02-22 Jakub Jelinek <jakub@redhat.com>
10044 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
10045 to all type variants.
10047 PR tree-optimization/84503
10048 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
10049 width as info->bitpos + info->bitsize - start.
10050 (merged_store_group::merge_overlapping): Simplify width computation.
10051 (check_no_overlap): New function.
10052 (imm_store_chain_info::try_coalesce_bswap): Compute expected
10053 start + width and last_order of the group, fail if check_no_overlap
10055 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
10056 to group if check_no_overlap fails.
10058 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10060 * config/rs6000/altivec.md: Delete contraint arguments to
10061 define_expand, define_split, and define_peephole2, and in
10062 define_insn_and_split if always unused.
10063 * config/rs6000/darwin.md: Ditto.
10064 * config/rs6000/dfp.md: Ditto.
10065 * config/rs6000/rs6000.md: Ditto.
10066 * config/rs6000/sync.md: Ditto.
10067 * config/rs6000/vector.md: Ditto.
10068 * config/rs6000/vsx.md: Ditto.
10070 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10072 * config/rs6000/altivec.md: Write output control strings as braced
10073 blocks instead of double-quoted strings.
10074 * config/rs6000/darwin.md: Ditto.
10075 * config/rs6000/rs6000.md: Ditto.
10076 * config/rs6000/vector.md: Ditto.
10077 * config/rs6000/vsx.md: Ditto.
10079 2018-02-21 Jason Merrill <jason@redhat.com>
10081 PR c++/84314 - ICE with templates and fastcall attribute.
10082 * attribs.c (build_type_attribute_qual_variant): Remove assert.
10084 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10086 * ipa-cp.c (determine_versionability): Fix comment typos.
10088 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10091 * ipa-cp.c (determine_versionability): Do not version functions caling
10094 2018-02-21 Martin Liska <mliska@suse.cz>
10097 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
10098 Add "native" as a possible value.
10099 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
10100 the macro when native cpu detection is available.
10102 2018-02-21 Martin Liska <mliska@suse.cz>
10105 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
10106 Add "native" as a possible value.
10107 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
10108 when native cpu detection is available.
10110 2018-02-21 Jakub Jelinek <jakub@redhat.com>
10111 Martin Sebor <msebor@redhat.com>
10113 PR tree-optimization/84478
10114 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
10116 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
10117 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
10118 support which is conservatively correct, for 2 only stay conservative
10119 for maxlen. Formatting and comment capitalization fixes. Add STRICT
10120 argument to the 2 argument get_range_strlen, adjust 6 arg
10121 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
10123 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
10124 (gimple_fold_builtin_strlen): Pass true as last argument to
10127 2018-02-20 Martin Sebor <msebor@redhat.com>
10129 PR middle-end/84095
10130 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
10131 (builtin_memref::set_base_and_offset): Same. Handle inner references.
10132 (builtin_memref::builtin_memref): Factor out parts into
10133 set_base_and_offset and call it.
10135 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10137 PR middle-end/84406
10138 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
10139 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
10140 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
10141 search at the associated MODE_INT.
10143 2018-02-20 Jeff Law <law@redhat.com>
10145 PR middle-end/82123
10146 PR tree-optimization/81592
10147 PR middle-end/79257
10148 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
10149 for range data rather than using global data.
10150 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
10151 range data rather than using global data.
10152 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
10153 pass it to children as needed.
10154 (struct directive::fmtresult): Similarly.
10155 (struct directive::set_width): Similarly.
10156 (struct directive::set_precision): Similarly.
10157 (format_integer, format_directive, parse_directive): Similarly.
10158 (format_none): Accept unnamed vr_values parameter.
10159 (format_percent, format_floating, format_character): Similarly.
10160 (format_string, format_plain): Similarly.
10161 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
10162 the EVRP range analyzer for range data rather than using global data.
10163 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
10164 gimple-ssa-evrp-analyze.h
10165 (class sprintf_dom_walker): Add after_dom_children member function.
10166 Add evrp_range_analyzer member.
10167 (sprintf_dom_walker::before_dom_children): Call into the EVRP
10168 range analyzer as needed.
10169 (sprintf_dom_walker::after_dom_children): New member function.
10170 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
10172 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
10173 (evrp_range_analyzer::pop_to_marker): Likewise.
10175 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10177 PR tree-optimization/84419
10178 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
10179 with the required type if its current type is compatible but
10182 2018-02-20 Jakub Jelinek <jakub@redhat.com>
10184 PR middle-end/82004
10185 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
10186 after vectorization.
10188 2018-02-20 Martin Liska <mliska@suse.cz>
10191 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
10192 possible values if we don't have a hint.
10194 2018-02-20 Martin Liska <mliska@suse.cz>
10198 * final.c (shorten_branches): Build align_tab array with one
10200 * opts.c (finish_options): Add alignment option limit check.
10201 (MAX_CODE_ALIGN): Likewise.
10202 (MAX_CODE_ALIGN_VALUE): Likewise.
10203 * doc/invoke.texi: Document maximum allowed option value for
10204 all -falign-* options.
10206 2018-02-19 Jakub Jelinek <jakub@redhat.com>
10209 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
10210 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
10211 * var-tracking.c (emit_note_insn_var_location): Remove all references
10212 to NOTE_INSN_CALL_ARG_LOCATION.
10213 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
10214 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
10215 Use copy_rtx_if_shared.
10216 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
10217 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
10218 (dwarf2out_var_location): Remove handling of
10219 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
10221 * final.c (final_scan_insn): Remove all references to
10222 NOTE_INSN_CALL_ARG_LOCATION.
10223 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
10224 before dumping final insns.
10225 * except.c (emit_note_eh_region_end): Remove all references to
10226 NOTE_INSN_CALL_ARG_LOCATION.
10227 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
10228 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
10229 * config/arc/arc.c (hwloop_optimize): Likewise.
10230 * config/arm/arm.c (create_fix_barrier): Likewise.
10231 * config/s390/s390.c (s390_chunkify_start): Likewise.
10232 * config/sh/sh.c (find_barrier): Likewise.
10233 * config/i386/i386.c (rest_of_insert_endbranch,
10234 ix86_seh_fixup_eh_fallthru): Likewise.
10235 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
10236 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
10237 * config/frv/frv.c (frv_function_prologue): Likewise.
10238 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
10240 (note_outside_basic_block_p): Remove all references to
10241 NOTE_INSN_CALL_ARG_LOCATION.
10242 * gengtype.c (adjust_field_rtx_def): Likewise.
10243 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
10245 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
10246 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
10249 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
10252 PR tree-optimization/84452
10253 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
10254 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
10257 2018-02-19 Martin Liska <mliska@suse.cz>
10260 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
10262 2018-02-19 Martin Liska <mliska@suse.cz>
10263 Richard Sandiford <richard.sandiford@linaro.org>
10265 PR tree-optimization/82491
10266 * gimple-fold.c (get_base_constructor): Make earlier bail out
10269 2018-02-19 Carl Love <cel@us.ibm.com>
10271 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
10272 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
10274 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
10275 P8V_BUILTIN_VEC_NEG.
10277 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
10279 * config/rl78/rl78.md (movdf): New define expand.
10281 2018-02-19 Martin Liska <mliska@suse.cz>
10284 * doc/invoke.texi: Fix typo.
10285 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
10287 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
10289 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
10290 handle rs6000_single_float and rs6000_double_float specially for
10293 2018-02-16 Jeff Law <law@redhat.com>
10295 * config/rx/rx.c (add_pop_cfi_notes): New function.;
10296 (pop_regs): Use it.
10298 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10301 * ipa-inline.c (inline_small_functions): Fix a typo.
10303 2018-02-16 Nathan Sidwell <nathan@acm.org>
10305 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
10307 2018-02-16 Carl Love <cel@us.ibm.com>
10309 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
10310 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
10311 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
10312 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
10313 expansion to P8V_BUILTIN_VEC_FLOAT2.
10315 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
10317 PR rtl-optimization/70023
10318 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
10319 src_regno into account.
10321 2018-02-16 Carl Love <cel@us.ibm.com>
10323 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
10324 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
10325 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
10326 * config/rs6000/rs6000.c: Remove case statements for
10327 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
10328 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
10329 and P9V_BUILTIN_VEC_VINSERT4B.
10330 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
10331 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
10332 * config/rs6000/vsx.md:
10333 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
10336 2018-02-16 Carl Love <cel@us.ibm.com>
10338 * config/rs6000/altivec.h: Add builtin names vec_extract4b
10340 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
10342 * config/rs6000/rs6000-c.c: Add the definitions for
10343 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
10344 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
10345 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
10346 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
10347 definition for insert4b and define insn *insert3b_internal.
10348 * doc/extend.texi: Add documentation for vec_extract4b.
10350 2018-02-16 Nathan Sidwell <nathan@acm.org>
10352 * doc/extend.texi (Backwards Compatibility): Mention friend
10353 injection. Note for-scope is deprecated.
10354 * doc/invoke.texi (-ffriend-injection): Deprecate.
10356 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
10358 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
10359 that moved to I2, also allow destinations that are a paradoxical
10360 subreg (instead of a normal reg).
10362 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
10365 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
10368 2018-02-16 Richard Biener <rguenther@suse.de>
10370 PR tree-optimization/84037
10371 PR tree-optimization/84016
10373 * config/i386/i386.c (ix86_builtin_vectorization_cost):
10374 Adjust vec_construct for the fact we need additional higher latency
10375 128bit inserts for AVX256 and AVX512 vector builds.
10376 (ix86_add_stmt_cost): Scale vector construction cost for
10379 2018-02-16 Richard Biener <rguenther@suse.de>
10381 PR tree-optimization/84417
10382 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
10383 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
10384 (non_rewritable_lvalue_p): Likewise, use poly-ints.
10386 2018-02-16 Martin Liska <mliska@suse.cz>
10389 * internal-fn.def (ASAN_CHECK): Set proper flags.
10390 (ASAN_MARK): Likewise.
10392 2018-02-16 Julia Koval <julia.koval@intel.com>
10394 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
10395 from PTA_CANNONLAKE.
10397 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10400 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
10401 Use ++iter rather than iter++ for std::list iterators.
10402 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
10403 defer deleting them until all nodes in the forest are processed. Do
10404 free even leaf nodes. Change to_process into auto_vec.
10407 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
10408 * vec.h (vec_default_construct): Use memset instead of placement new
10409 if BROKEN_VALUE_INITIALIZATION is defined.
10410 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
10411 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
10414 PR rtl-optimization/83723
10415 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
10416 * lra.c (lra_substitute_pseudo): Likewise. If true, use
10417 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
10419 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
10421 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
10423 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
10425 PR rtl-optimization/81443
10426 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
10427 from inner REGs to paradoxical SUBREGs.
10429 2018-02-16 Richard Biener <rguenther@suse.de>
10431 PR tree-optimization/84399
10432 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10433 For operands we can analyze at their definition make sure we can
10434 analyze them at each use as well.
10436 2018-02-16 Richard Biener <rguenther@suse.de>
10438 PR tree-optimization/84190
10439 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
10440 volatile accesses if the decl isn't volatile.
10442 2018-02-15 Jason Merrill <jason@redhat.com>
10444 PR c++/84314 - ICE with templates and fastcall attribute.
10445 * attribs.c (build_type_attribute_qual_variant): Don't clobber
10446 TYPE_CANONICAL on an existing type.
10448 2018-02-15 Jakub Jelinek <jakub@redhat.com>
10450 PR tree-optimization/84383
10451 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
10452 dstoff nor call operand_equal_p if dstbase is NULL.
10454 PR tree-optimization/84334
10455 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
10456 also a CONSTANT_CLASS_P, punt.
10458 2018-02-14 Jim Wilson <jimw@sifive.com>
10460 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
10461 first SMALL_OPERAND check. New local min_second_step. Move assert
10462 to where locals are set. Add TARGET_RVC support.
10463 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
10465 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
10467 * doc/invoke.texi: Correct -Wformat-overflow code sample.
10469 2018-02-14 Martin Sebor <msebor@redhat.com>
10471 PR tree-optimization/83698
10472 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
10473 arrays constrain the offset range to their bounds.
10474 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
10475 (builtin_access::overlap): Avoid setting the size of overlap if it's
10477 (maybe_diag_overlap): Also consider arrays when deciding what values
10478 of offsets to include in diagnostics.
10480 2018-02-14 Martin Sebor <msebor@redhat.com>
10483 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
10484 that correspond to the kind of a declaration.
10486 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
10489 * config/pa/pa.md: Load address of PIC label using the linkage table
10490 if the label is nonlocal.
10492 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
10494 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
10495 warning message if user requests -maltivec=be.
10496 * doc/invoke.texi: Document deprecation of -maltivec=be.
10498 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
10501 * config/rs6000/rs6000-c.c: Update definitions for
10502 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
10503 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
10505 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10508 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
10509 add _get_ssp intrinsics. Remove argument from
10510 __builtin_ia32_rdssp[d|q].
10511 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
10512 * config/i386/i386-builtin.def: Remove argument from
10513 __builtin_ia32_rdssp[d|q].
10514 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
10515 ix86_expand_special_args_builtin for _rdssp[d|q].
10516 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
10517 Clear register before usage.
10518 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
10519 Add documentation for new _get_ssp and _inc_ssp intrinsics.
10521 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
10523 PR tree-optimization/84357
10524 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
10525 operand 1 of an ARRAY_REF too.
10527 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
10530 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
10531 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10533 (set_of_reg): New struct.
10534 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
10535 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
10536 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10538 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
10539 Split into bitclr, bitset, bitinvert patterns if appropriate.
10540 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
10541 use rx_fuse_in_memory_bitop.
10542 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
10543 to named insn, correct maximum insn length.
10545 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
10548 * machmode.def: Define a complex mode for PARTIAL_INT.
10549 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
10551 * doc/rtl.texi: Document CSPImode.
10552 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
10554 (msp430_hard_regno_nregs_with_padding): Likewise.
10556 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
10559 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
10561 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
10563 PR rtl-optimization/84169
10564 * combine.c (try_combine): New variable split_i2i3. Set it to true if
10565 we generated a parallel as new i3 and we split that to new i2 and i3
10566 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
10567 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
10568 those to i2, not i1. Partially rewrite this scan code.
10570 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10573 * stor-layout.c (place_field): For variable length fields, adjust
10574 offset_align afterwards not just based on the field's alignment,
10575 but also on the size.
10577 PR middle-end/84309
10578 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
10579 of exps and logs in the use_exp2 case.
10581 2018-02-13 Jeff Law <law@redhat.com>
10583 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
10584 entry for "vector".
10586 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
10589 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
10593 * common.opt (gas-loc-support, gas-locview-support): New.
10594 (ginline-points, ginternal-reset-location-views): New.
10595 * doc/invoke.texi: Document them. Use @itemx where intended.
10596 (gvariable-location-views): Adjust.
10597 * target.def (reset_location_view): New.
10598 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
10599 (TARGET_RESET_LOCATION_VIEW): New.
10600 * doc/tm.texi: Rebuilt.
10601 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
10602 (dwarf2out_default_as_locview_support): New.
10603 (output_asm_line_debug_info): Use option variables.
10604 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
10605 (output_loc_list): Likewise.
10606 (add_high_low_attributes): Check option variables.
10607 Don't output entry view attribute in strict mode.
10608 (gen_inlined_subroutine_die): Check option variables.
10609 (dwarf2out_inline_entry): Likewise.
10610 (init_sections_and_labels): Likewise.
10611 (dwarf2out_early_finish): Likewise.
10612 (maybe_reset_location_view): New, from...
10613 (dwarf2out_var_location): ... here. Call it.
10614 * debug.h (dwarf2out_default_as_loc_support): Declare.
10615 (dwarf2out_default_as_locview_support): Declare.
10616 * hooks.c (hook_int_rtx_insn_0): New.
10617 * hooks.h (hook_int_rtx_insn_0): Declare.
10618 * toplev.c (process_options): Take -gas-loc-support and
10619 -gas-locview-support from dwarf2out. Enable
10620 -gvariable-location-views by default only with locview
10621 assembler support. Enable -ginternal-reset-location-views by
10622 default only if the target defines the corresponding hook.
10623 Enable -ginline-points by default if location views are
10624 enabled; force it disabled if statement frontiers are
10626 * tree-inline.c (expand_call_inline): Check option variables.
10627 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
10629 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
10631 PR tree-optimization/84321
10632 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
10633 handling. Also check whether the anti-range contains any values
10634 that satisfy the mask; switch to a VR_RANGE if not.
10636 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
10639 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
10641 2018-02-13 Martin Jambor <mjambor@suse.cz>
10644 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
10645 of call statements, also set location of a load to a temporary.
10647 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
10649 * config/rl78/rl78.c (add_vector_labels): New function.
10650 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
10651 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
10652 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
10653 which checks that no arguments are passed.
10654 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
10655 * doc/extend.texi: Documentation for the new attribute.
10657 2018-02-13 Andreas Schwab <schwab@suse.de>
10659 * config/riscv/linux.h (CPP_SPEC): Define.
10661 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10664 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
10665 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
10666 OPTION_MASK_ISA_AES as first argument to def_builtin_const
10667 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
10668 instead of OPTION_MASK_ISA_PCLMUL as first argument to
10669 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
10670 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
10671 temporarily for AES and PCLMUL builtins.
10673 PR tree-optimization/84339
10674 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
10675 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
10678 PR middle-end/84309
10679 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
10680 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
10681 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
10683 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
10685 * omp-simd-clone.h: New file.
10686 * omp-simd-clone.c: Include omp-simd-clone.h.
10687 (expand_simd_clones): No longer static.
10688 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
10689 cgraph.h and omp-simd-clone.h.
10690 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
10691 (vect_recog_widen_shift_pattern): Formatting fix.
10692 (vect_pattern_recog_1): Don't check optab for calls.
10695 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
10696 operands[2] into a REG before using gen_lowpart on it.
10698 2018-02-12 Jeff Law <law@redhat.com>
10701 * config/sh/sh.c (find_barrier): Consider a sibling call
10704 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
10705 successfully back substituting a reg.
10707 2018-02-12 Richard Biener <rguenther@suse.de>
10709 PR tree-optimization/84037
10710 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
10711 parameter, move visited init to caller.
10712 (vect_slp_analyze_operations): Separate cost from validity
10713 check, initialize visited once for all instances.
10714 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
10716 * tree-vect-stmts.c (vect_model_simple_cost): Make early
10718 (vect_model_promotion_demotion_cost): Likewise.
10719 (vectorizable_bswap): Guard cost modeling with !slp_node
10720 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
10722 (vectorizable_call): Likewise.
10723 (vectorizable_conversion): Likewise.
10724 (vectorizable_assignment): Likewise.
10725 (vectorizable_shift): Likewise.
10726 (vectorizable_operation): Likewise.
10727 (vectorizable_store): Likewise.
10728 (vectorizable_load): Likewise.
10729 (vectorizable_condition): Likewise.
10730 (vectorizable_comparison): Likewise.
10732 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
10735 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
10736 (ASAN_MARK): Fix fnspec to account for return value, change pointer
10737 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
10739 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10741 PR middle-end/83665
10742 * params.def (inline-min-speedup): Increase from 8 to 15.
10743 (max-inline-insns-auto): Decrease from 40 to 30.
10744 * ipa-split.c (consider_split): Add some buffer for function to
10745 be considered inlining candidate.
10746 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
10749 2018-02-12 Richard Biener <rguenther@suse.de>
10751 PR tree-optimization/84037
10752 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
10753 matched stmts if we cannot swap the non-matched ones.
10755 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10757 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
10758 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
10759 _mm_maskz_scalef_round_ss): New intrinsics.
10760 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
10761 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
10762 __builtin_ia32_scalefss_round): Remove.
10763 (__builtin_ia32_scalefsd_mask_round,
10764 __builtin_ia32_scalefss_mask_round): New intrinsics.
10765 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
10766 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
10767 ((match_operand:VF_128 2 "<round_nimm_predicate>"
10768 "<round_constraint>")): Changed to ...
10769 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
10770 "<round_scalar_constraint>")): ... this.
10771 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
10772 %0, %1, %2<round_op3>}"): Changed to ...
10773 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
10774 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
10775 %2<round_scalar_mask_op3>}"): ... this.
10776 * config/i386/subst.md (round_scalar_nimm_predicate): New.
10778 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10780 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
10781 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
10782 (_mm_maskz_sqrt_round_ss): New intrinsics.
10783 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
10784 (__builtin_ia32_sqrtsd_mask_round)
10785 (__builtin_ia32_sqrtss_mask_round): New builtins.
10786 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
10787 (__builtin_ia32_sqrtss_round): Remove.
10788 (__builtin_ia32_sqrtsd_mask_round)
10789 (__builtin_ia32_sqrtss_mask_round): New builtins.
10790 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
10791 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
10792 ((match_operand:VF_128 1 "vector_operand"
10793 "xBm,<round_constraint>")): Changed to ...
10794 ((match_operand:VF_128 1 "vector_operand"
10795 "xBm,<round_scalar_constraint>")): ... this.
10796 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
10797 %0, %2, %<iptr>1<round_op3>}): Changed to ...
10798 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
10799 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
10800 %<iptr>1<round_scalar_mask_op3>}): ... this.
10801 ((set_attr "prefix" "<round_prefix>")): Changed to ...
10802 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
10804 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
10807 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
10808 Cast vec_cmpeq result to correct type.
10809 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
10810 Cast vec_cmpgt result to correct type.
10812 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
10814 * final.c (final_scan_insn_1): Renamed from...
10815 (final_scan_insn): ... this. New wrapper, to recover
10816 seen from the outermost call in recursive ones.
10817 * config/sparc/sparc.c (output_return): Drop seen from call.
10818 (output_sibcall): Likewise.
10819 * config/visium/visium.c (output_branch): Likewise.
10821 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
10823 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
10826 2018-02-10 Alan Modra <amodra@gmail.com>
10829 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
10830 Specify LR as an input.
10832 2018-02-10 Jakub Jelinek <jakub@redhat.com>
10835 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
10836 remove_member_access_dummy_vars): New functions.
10837 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
10838 lower_omp_1, execute_lower_omp): Use them.
10840 PR rtl-optimization/84308
10841 * shrink-wrap.c (spread_components): Release todo vector.
10843 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
10845 PR rtl-optimization/57193
10846 * ira-color.c (struct allocno_color_data): Add member
10847 conflict_allocno_hard_prefs.
10848 (update_conflict_allocno_hard_prefs): New.
10849 (bucket_allocno_compare_func): Add a preference based on
10850 conflict_allocno_hard_prefs.
10851 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
10852 (color_allocnos): Remove a dead code. Initiate
10853 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
10855 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10858 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
10859 constraint from =wa to wa. Avoid a subreg on the output operand,
10860 instead use a pseudo and subreg it in a move.
10861 (p9_xxbrd_<mode>): Changed to ...
10862 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
10863 (p9_xxbrd_v2df): New expander.
10864 (p9_xxbrw_<mode>): Changed to ...
10865 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
10866 (p9_xxbrw_v4sf): New expander.
10868 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10870 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
10872 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
10875 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
10876 multiply in 32-bit mode.
10877 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
10878 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
10881 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10883 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
10884 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
10885 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
10886 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
10888 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
10891 * dwarf2out.c (is_trivial_indirect_ref): New function.
10892 (dwarf2out_late_global_decl): Do not generate a location
10893 attribute for variables that have a non-trivial DECL_VALUE_EXPR
10894 and that are not defined in the current unit.
10896 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10898 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
10899 instead of a libcall for UNORDERED.
10901 2018-02-09 Tamar Christina <tamar.christina@arm.com>
10904 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
10905 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
10907 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10910 * config/s390/s390.c (s390_set_current_function): Invoke
10911 s390_indirect_branch_settings also if fndecl didn't change.
10913 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10915 * config/rs6000/rs6000.md (blockage): Set length to zero.
10917 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10919 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
10921 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10924 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
10925 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
10929 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
10930 PARALLEL incoming that failed vt_get_decl_and_offset check.
10932 PR middle-end/84237
10933 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
10934 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
10936 (get_variable_section): For decls in named .bss* sections pass true as
10937 second argument to bss_initializer_p.
10939 2018-02-09 Marek Polacek <polacek@redhat.com>
10940 Jakub Jelinek <jakub@redhat.com>
10943 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
10944 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
10945 Sync some changes from cxx_fold_indirect_ref.
10947 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10949 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
10951 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
10952 (BLOCK_INLINE_ENTRY_LABEL): New.
10953 (dwarf2out_var_location): Disregard inline entry markers.
10954 (inline_entry_data): New struct.
10955 (inline_entry_data_hasher): New hashtable type.
10956 (inline_entry_data_hasher::hash): New.
10957 (inline_entry_data_hasher::equal): New.
10958 (inline_entry_data_table): New variable.
10959 (add_high_low_attributes): Add DW_AT_entry_pc and
10960 DW_AT_GNU_entry_view attributes if a pending entry is found
10961 in inline_entry_data_table. Add old entry_pc attribute only
10962 if debug nonbinding markers are disabled.
10963 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
10964 markers are enabled.
10965 (block_within_block_p, dwarf2out_inline_entry): New.
10966 (dwarf2out_finish): Check that no entries remained in
10967 inline_entry_data_table.
10968 * final.c (reemit_insn_block_notes): Handle inline entry notes.
10969 (final_scan_insn, notice_source_line): Likewise.
10970 (rest_of_clean_state): Skip inline entry markers.
10971 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
10973 * gimple.c (gimple_build_debug_inline_entry): New.
10974 * gimple.h (enum gimple_debug_subcode): Add
10975 GIMPLE_DEBUG_INLINE_ENTRY.
10976 (gimple_build_debug_inline_entry): Declare.
10977 (gimple_debug_inline_entry_p): New.
10978 (gimple_debug_nonbind_marker_p): Adjust.
10979 * insn-notes.def (INLINE_ENTRY): New.
10980 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
10981 inline entry marker notes.
10982 (print_insn): Likewise.
10983 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
10984 (INSN_DEBUG_MARKER_KIND): Likewise.
10985 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
10986 * tree-inline.c (expand_call_inline): Build and insert
10987 debug_inline_entry stmt.
10988 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
10989 inline entry blocks early, if nonbind markers are enabled.
10990 (dump_scope_block): Dump fragment info.
10991 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
10992 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
10993 (gimple_build_debug_inline_entry): New.
10994 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
10995 Enable/disable inline entry points too.
10996 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
10997 (DEBUG_INSN): Describe inline entry markers.
10999 * common.opt (gvariable-location-views): New.
11000 (gvariable-location-views=incompat5): New.
11001 * config.in: Rebuilt.
11002 * configure: Rebuilt.
11003 * configure.ac: Test assembler for view support.
11004 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
11005 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
11006 * dwarf2out.c (var_loc_view): New typedef.
11007 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
11008 (dwarf2out_locviews_in_attribute): New.
11009 (dwarf2out_locviews_in_loclist): New.
11010 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
11011 (enum dw_line_info_opcode): Add LI_adv_address.
11012 (struct dw_line_info_table): Add view.
11013 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
11014 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
11015 (zero_view_p): New variable.
11016 (ZERO_VIEW_P): New macro.
11017 (output_asm_line_debug_info): New.
11018 (struct var_loc_node): Add view.
11019 (add_AT_view_list, AT_loc_list): New.
11020 (add_var_loc_to_decl): Add view param. Test it against last.
11021 (new_loc_list): Add view params. Record them.
11022 (AT_loc_list_ptr): Handle loc and view lists.
11023 (view_list_to_loc_list_val_node): New.
11024 (print_dw_val): Handle dw_val_class_view_list.
11025 (size_of_die): Likewise.
11026 (value_format): Likewise.
11027 (loc_list_has_views): New.
11028 (gen_llsym): Set vl_symbol too.
11029 (maybe_gen_llsym, skip_loc_list_entry): New.
11030 (dwarf2out_maybe_output_loclist_view_pair): New.
11031 (output_loc_list): Output view list or entries too.
11032 (output_view_list_offset): New.
11033 (output_die): Handle dw_val_class_view_list.
11034 (output_dwarf_version): New.
11035 (output_compilation_unit_header): Use it.
11036 (output_skeleton_debug_sections): Likewise.
11037 (output_rnglists, output_line_info): Likewise.
11038 (output_pubnames, output_aranges): Update version comments.
11039 (output_one_line_info_table): Output view numbers in asm comments.
11040 (dw_loc_list): Determine current endview, pass it to new_loc_list.
11041 Call maybe_gen_llsym.
11042 (loc_list_from_tree_1): Adjust.
11043 (add_AT_location_description): Create view list attribute if
11044 needed, check it's absent otherwise.
11045 (convert_cfa_to_fb_loc_list): Adjust.
11046 (maybe_emit_file): Call output_asm_line_debug_info for test.
11047 (dwarf2out_var_location): Reset views as needed. Precompute
11048 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
11049 attribute. Set view.
11050 (new_line_info_table): Reset next view.
11051 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
11052 (dwarf2out_source_line): Likewise. Output view resets and labels to
11053 the assembler, or select appropriate line info opcodes.
11054 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
11055 (optimize_string_length): Catch it. Adjust.
11056 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
11057 dw_val_class_view_list, and remove it if no longer needed.
11058 (hash_loc_list): Hash view numbers.
11059 (loc_list_hasher::equal): Compare them.
11060 (optimize_location_lists): Check whether a view list symbol is
11061 needed, and whether the locview attribute is present, and
11062 whether they match. Remove the locview attribute if no longer
11064 (index_location_lists): Call skip_loc_list_entry for test.
11065 (dwarf2out_finish): Call output_asm_line_debug_info for test.
11066 Use output_dwarf_version.
11067 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
11068 (struct dw_val_node): Add val_view_list.
11069 * final.c (SEEN_NEXT_VIEW): New.
11070 (set_next_view_needed): New.
11071 (clear_next_view_needed): New.
11072 (maybe_output_next_view): New.
11073 (final_start_function): Rename to...
11074 (final_start_function_1): ... this. Take pointer to FIRST,
11075 add SEEN parameter. Emit param bindings in the initial view.
11076 (final_start_function): Reintroduce SEEN-less interface.
11077 (final): Rename to...
11078 (final_1): ... this. Take SEEN parameter. Output final pending
11079 next view at the end.
11080 (final): Reintroduce seen-less interface.
11081 (final_scan_insn): Output pending next view before switching
11082 sections or ending a block. Mark the next view as needed when
11083 outputting variable locations. Notify debug backend of section
11084 changes, and of location view changes.
11085 (rest_of_handle_final): Adjust.
11086 * toplev.c (process_options): Autodetect value for debug variable
11087 location views option. Warn on incompat5 without -gdwarf-5.
11088 * doc/invoke.texi (gvariable-location-views): New.
11089 (gvariable-location-views=incompat5): New.
11090 (gno-variable-location-views): New.
11092 2018-02-08 David Malcolm <dmalcolm@redhat.com>
11094 PR tree-optimization/84136
11095 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
11096 that the result of find_edge is non-NULL.
11098 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11101 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
11102 storing integer register in SImode. Fix cost of 256 and 512
11103 byte aligned SSE register store.
11105 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11107 * config/i386/i386.c (ix86_multiplication_cost): Fix
11108 multiplication cost for TARGET_AVX512DQ.
11110 2018-02-08 Marek Polacek <polacek@redhat.com>
11112 PR tree-optimization/84238
11113 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
11116 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11118 PR tree-optimization/84265
11119 * tree-vect-stmts.c (vectorizable_store): Don't treat
11120 VMAT_CONTIGUOUS accesses as grouped.
11121 (vectorizable_load): Likewise.
11123 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11125 PR tree-optimization/81635
11126 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
11127 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
11128 (test_round_for_mask): New functions.
11129 (wide_int_cc_tests): Call test_round_for_mask.
11130 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
11131 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
11132 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
11133 range returned by get_range_info.
11135 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11138 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
11139 * symtab.c: Include builtins.h
11140 (symtab_node::output_to_lto_symbol_table_p): Move here
11141 from lto-streamer-out.c:output_symbol_p.
11142 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
11143 (output_symbol_p): Move all logic to symtab.c
11144 (produce_symtab): Update.
11146 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11148 * config/s390/s390-opts.h (enum indirect_branch): Define.
11149 * config/s390/s390-protos.h (s390_return_addr_from_memory)
11150 (s390_indirect_branch_via_thunk)
11151 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
11152 (enum s390_indirect_branch_type): Define.
11153 * config/s390/s390.c (struct s390_frame_layout, struct
11154 machine_function): Remove.
11155 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
11156 (indirect_branch_table_label_no, indirect_branch_table_name):
11158 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
11159 (enum s390_indirect_branch_option): Define.
11160 (s390_return_addr_from_memory): New function.
11161 (s390_handle_string_attribute): New function.
11162 (s390_attribute_table): Add new attribute handler.
11163 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
11164 (s390_indirect_branch_via_thunk): New function.
11165 (s390_indirect_branch_via_inline_thunk): New function.
11166 (s390_function_ok_for_sibcall): When jumping via thunk disallow
11167 sibling call optimization for non z10 compiles.
11168 (s390_emit_call): Force indirect branch target to be a single
11169 register. Add r1 clobber for non-z10 compiles.
11170 (s390_emit_epilogue): Emit return jump via return_use expander.
11171 (s390_reorg): Handle JUMP_INSNs as execute targets.
11172 (s390_option_override_internal): Perform validity checks for the
11173 new command line options.
11174 (s390_indirect_branch_attrvalue): New function.
11175 (s390_indirect_branch_settings): New function.
11176 (s390_set_current_function): Invoke s390_indirect_branch_settings.
11177 (s390_output_indirect_thunk_function): New function.
11178 (s390_code_end): Implement target hook.
11179 (s390_case_values_threshold): Implement target hook.
11180 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
11182 * config/s390/s390.h (struct s390_frame_layout)
11183 (struct machine_function): Move here from s390.c.
11184 (TARGET_INDIRECT_BRANCH_NOBP_RET)
11185 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
11186 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
11187 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
11188 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
11189 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
11190 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
11191 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
11192 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
11193 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
11194 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
11195 (mnemonic attribute): Add values which aren't recognized
11197 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
11198 pattern for branch conversion. Fix mnemonic attribute.
11199 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
11200 indirect branch via thunk if requested.
11201 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
11202 ("*indirect_jump"): Disable for branch conversion using out of
11204 ("indirect_jump_via_thunk<mode>_z10")
11205 ("indirect_jump_via_thunk<mode>")
11206 ("indirect_jump_via_inlinethunk<mode>_z10")
11207 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
11208 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
11209 ("casesi_jump_via_inlinethunk<mode>_z10")
11210 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
11211 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
11212 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
11213 ("*indirect2_jump"): Disable for branch conversion.
11214 ("casesi_jump"): Turn into expander and expand patterns for branch
11216 ("return_use"): New expander.
11217 ("*return"): Emit return via thunk and rename it to ...
11218 ("*return<mode>"): ... this one.
11219 * config/s390/s390.opt: Add new options and and enum for the
11222 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11224 * lra-constraints.c (match_reload): Unconditionally use
11225 gen_lowpart_SUBREG, rather than selecting between that
11226 and equivalent gen_rtx_SUBREG code.
11228 2018-02-08 Richard Biener <rguenther@suse.de>
11230 PR tree-optimization/84233
11231 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
11232 changed flag instead of boguously re-using phi_inserted.
11234 2018-02-08 Martin Jambor <mjambor@suse.cz>
11236 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
11237 static local variables.
11239 2018-02-08 Richard Biener <rguenther@suse.de>
11241 PR tree-optimization/84278
11242 * tree-vect-stmts.c (vectorizable_store): When looking for
11243 smaller vector types to perform grouped strided loads/stores
11244 make sure the mode is supported by the target.
11245 (vectorizable_load): Likewise.
11247 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11249 * config/aarch64/aarch64.c (aarch64_components_for_bb):
11250 Increase LDP/STP opportunities by adding adjacent callee-saves.
11252 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11254 PR rtl-optimization/84068
11255 PR rtl-optimization/83459
11256 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
11258 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
11260 PR tree-optimization/84224
11261 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
11262 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
11263 non-zero arguments.
11265 2018-02-07 Iain Sandoe <iain@codesourcery.com>
11268 * config/rs6000/altivec.md (*restore_world): Remove LR use.
11269 * config/rs6000/predicates.md (restore_world_operation): Adjust op
11270 count, remove one USE.
11272 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
11274 * doc/install.texi (Configuration): Document the
11275 --with-long-double-format={ibm,ieee} PowerPC configuration
11279 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
11280 Convert from define_expand to be define_insn_and_split. Rework
11281 float/double/_Float128 conversions to QI/HI/SImode to work with
11282 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
11283 conversions to QI/HImode types did a store and then a load to
11284 truncate the value. For conversions to VSX registers, don't split
11285 the insn, instead emit the code directly. Use the code iterator
11286 any_fix to combine signed and unsigned conversions.
11287 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
11288 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
11289 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11290 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
11291 (fix_<mode>di2_hw): Likewise.
11292 (fixuns_<mode>di2_hw): Likewise.
11293 (fix_<mode>si2_hw): Likewise.
11294 (fixuns_<mode>si2_hw): Likewise.
11295 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
11296 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11297 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
11298 fix<uns>_trunc<SFDF:mode>si2_p8.
11299 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
11300 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
11301 (fix<uns>_<mode>_mem): Likewise.
11302 (fctiw<u>z_<mode>_mem): Likewise.
11303 (fix<uns>_<mode>_mem): Likewise.
11304 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
11305 the register allocator from doing a direct move to the GPRs to do
11306 a store, and instead use the ISA 3.0 store byte/half-word from
11307 vector register instruction. For IEEE 128-bit floating point,
11308 also optimize stores of 32-bit ints.
11309 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
11311 2018-02-07 Alan Hayward <alan.hayward@arm.com>
11313 * genextract.c (push_pathstr_operand): New function to support
11315 (walk_rtx): Call push_pathstr_operand.
11316 (print_path): Support [a-zA-Z].
11318 2018-02-07 Richard Biener <rguenther@suse.de>
11320 PR tree-optimization/84037
11321 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
11322 (cse_and_gimplify_to_preheader): Declare.
11323 (vect_get_place_in_interleaving_chain): Likewise.
11324 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11326 (_loop_vec_info::~_loop_vec_info): Delete it.
11327 (cse_and_gimplify_to_preheader): New function.
11328 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
11329 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
11330 (vectorizable_load): Likewise. For grouped stores always base
11331 the IV on the first element.
11332 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
11333 condition before gimplifying.
11335 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11337 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
11338 *DIV_EXPR and *MOD_EXPR.
11340 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
11343 * config/i386/i386.c (ix86_option_override_internal): Mask out
11344 the CF_SET bit when checking -fcf-protection.
11346 2018-02-07 Tom de Vries <tom@codesourcery.com>
11349 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
11352 2018-02-07 Richard Biener <rguenther@suse.de>
11354 PR tree-optimization/84204
11355 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
11358 PR tree-optimization/84205
11359 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
11360 special-case isl_ast_op_zdiv_r.
11362 PR tree-optimization/84223
11363 * graphite-scop-detection.c (gather_bbs::before_dom_children):
11364 Only add conditions from within the region.
11365 (gather_bbs::after_dom_children): Adjust.
11367 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
11370 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
11371 * config/avr/avr.md: Only post-reload split REG-REG moves if
11372 either register is GENERAL_REG_P.
11374 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11376 PR tree-optimization/84235
11377 * tree-ssa-scopedtables.c
11378 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
11379 if the subtraction is performed in floating point type where NaNs are
11380 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
11381 build 1. Formatting fix.
11383 2018-02-06 Jakub Jelinek <jakub@redhat.com>
11386 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
11387 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
11388 and skip it regardless of bb boundaries. Use CALL_P macro,
11389 don't test INSN_P (insn) together with CALL_P or JUMP_P check
11390 unnecessarily, formatting fix.
11392 2018-02-06 Michael Collison <michael.collison@arm.com>
11394 * config/arm/thumb2.md:
11395 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
11396 (*thumb_mov_notscc): Ditto.
11398 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
11401 * config/rs6000/rs6000.md (su code attribute): Use "u" for
11402 unsigned_fix, not "s".
11404 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11406 * configure.ac (gcc_fn_eh_frame_ro): New function.
11407 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
11408 correct .eh_frame permissions.
11409 * configure: Regenerate.
11411 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
11413 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
11414 irrelevant options.
11416 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11418 * config/rs6000/rs6000.c (rs6000_option_override_internal):
11419 Display warning message for -mno-speculate-indirect-jumps.
11421 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
11423 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
11425 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
11427 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
11429 PR tree-optimization/84225
11430 * tree-eh.c (find_trapping_overflow): Only call
11431 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
11433 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11436 * config/i386/i386.c: Reimplement the check of possible options
11437 -mibt/-mshstk conbination. Change error messages.
11438 * doc/invoke.texi: Fix a typo: remove extra '='.
11440 2018-02-06 Marek Polacek <polacek@redhat.com>
11442 PR tree-optimization/84228
11443 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
11445 2018-02-06 Tamar Christina <tamar.christina@arm.com>
11448 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
11449 emitted arch directives.
11450 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
11451 __ARM_FEATURE_COPROC before changing architectures.
11453 2018-02-06 Richard Biener <rguenther@suse.de>
11455 * config/i386/i386.c (print_reg): Fix typo.
11456 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
11458 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
11460 * configure: Regenerate.
11462 2018-02-05 Martin Sebor <msebor@redhat.com>
11464 PR tree-optimization/83369
11465 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
11468 2018-02-05 Martin Liska <mliska@suse.cz>
11470 * doc/invoke.texi: Cherry-pick upstream r323995.
11472 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
11474 * ira.c (ira_init_register_move_cost): Adjust comment.
11476 2018-02-05 Martin Liska <mliska@suse.cz>
11478 PR gcov-profile/84137
11479 * doc/gcov.texi: Fix typo in documentation.
11481 2018-02-05 Martin Liska <mliska@suse.cz>
11483 PR gcov-profile/83879
11484 * doc/gcov.texi: Document necessity of --dynamic-list-data when
11485 using dlopen functionality.
11487 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
11489 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
11490 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
11491 _mm_maskz_range_ss, _mm_mask_range_round_ss,
11492 _mm_maskz_range_round_ss): New intrinsics.
11493 (__builtin_ia32_rangesd128_round)
11494 (__builtin_ia32_rangess128_round): Remove.
11495 (__builtin_ia32_rangesd128_mask_round,
11496 __builtin_ia32_rangess128_mask_round): New builtins.
11497 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
11498 __builtin_ia32_rangess128_round): Remove.
11499 (__builtin_ia32_rangesd128_mask_round,
11500 __builtin_ia32_rangess128_mask_round): New builtins.
11501 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
11502 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
11503 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
11504 "<round_saeonly_constraint>")): Changed to ...
11505 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
11506 "<round_saeonly_scalar_constraint>")): ... this.
11507 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
11508 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
11509 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
11510 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
11511 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
11513 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
11515 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
11517 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
11518 Remove all values except native, 8540 and 8548.
11520 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
11522 * config/i386/i386.c (ix86_output_function_return): Pass
11523 INVALID_REGNUM, instead of -1, as invalid register number to
11524 indirect_thunk_name and output_indirect_thunk.
11526 2018-02-02 Julia Koval <julia.koval@intel.com>
11528 * config.gcc: Add -march=icelake.
11529 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
11530 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
11531 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
11532 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
11533 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
11534 (processor_target_table): Add icelake.
11535 (ix86_option_override_internal): Handle new PTAs.
11536 (get_builtin_code_for_version): Handle icelake.
11537 (M_INTEL_COREI7_ICELAKE): New.
11538 (fold_builtin_cpu): Handle icelake.
11539 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
11540 * doc/invoke.texi: Add -march=icelake.
11542 2018-02-02 Julia Koval <julia.koval@intel.com>
11544 * config/i386/i386.c (ix86_option_override_internal): Change flags type
11545 to wide_int_bitmask.
11546 * wide-int-bitmask.h: New.
11548 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11551 * config/i386/i386.md: Replace Pmode with word_mode in
11552 builtin_setjmp_setup and builtin_longjmp to support x32.
11554 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
11558 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
11560 (rs6000_supported_cpu_names): New static variable.
11561 (linux_cpu_translation_table): Likewise.
11562 (elf_platform) <cpu>: Define new static variable and use it.
11563 Translate kernel AT_PLATFORM name to canonical name if needed.
11564 Error if platform name is unknown.
11566 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
11569 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
11571 2018-02-01 Jeff Law <law@redhat.com>
11574 * config/i386/i386.c (release_scratch_register_on_entry): Add new
11575 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
11576 the scratch if RELEASE_VIA_POP is false.
11577 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
11578 If we have to save a temporary register, decrement SIZE appropriately.
11579 Pass new arguments to release_scratch_register_on_entry.
11580 (ix86_adjust_stack_and_probe): Likewise.
11581 (ix86_emit_probe_stack_range): Pass new arguments to
11582 release_scratch_register_on_entry.
11584 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
11586 PR rtl-optimization/84157
11587 * combine.c (change_zero_ext): Use REG_P predicate in
11588 front of HARD_REGISTER_P predicate.
11590 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
11592 * config/avr/avr.c (avr_option_override): Move disabling of
11593 -fdelete-null-pointer-checks to...
11594 * common/config/avr/avr-common.c (avr_option_optimization_table):
11597 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11599 PR tree-optimization/81635
11600 * tree-data-ref.c (split_constant_offset_1): For types that
11601 wrap on overflow, try to use range info to prove that wrapping
11604 2018-02-01 Renlin Li <renlin.li@arm.com>
11607 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
11608 TAILCALL_ADDR_REGS.
11609 (aarch64_register_move_cost): Likewise.
11610 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
11611 TAILCALL_ADDR_REGS.
11612 (REG_CLASS_NAMES): Likewise.
11613 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
11614 TAILCALL_ADDR_REGS. Remove IP registers.
11615 * config/aarch64/aarch64.md (Ucs): Update register constraint.
11617 2018-02-01 Richard Biener <rguenther@suse.de>
11619 * domwalk.h (dom_walker::dom_walker): Add additional constructor
11620 for specifying RPO order and allow NULL for that.
11621 * domwalk.c (dom_walker::dom_walker): Likewise.
11622 (dom_walker::walk): Handle NULL RPO order.
11623 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
11625 (rewrite_update_dom_walker): Likewise.
11626 (mark_def_dom_walker): Likewise.
11628 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11630 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
11631 (aarch64_maybe_expand_sve_subreg_move): Declare.
11632 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
11633 * config/aarch64/predicates.md (aarch64_any_register_operand): New
11635 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
11636 that are semantically a reverse operation.
11637 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
11638 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
11639 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
11641 (aarch64_can_change_mode_class): For big-endian, forbid changes
11642 between two SVE modes if they have different element sizes.
11644 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11646 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
11647 the TImode handling for big-endian targets.
11649 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11651 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
11652 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
11654 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
11655 Remove BSWAP handing for big-endian targets and use the form of
11656 LD1RQ appropariate for the mode.
11658 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11660 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
11661 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
11662 duplicated element.
11664 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11667 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
11668 check for operands that need to go through aarch64_sve_reload_be.
11670 2018-02-01 Jakub Jelinek <jakub@redhat.com>
11672 PR tree-optimization/81661
11673 PR tree-optimization/84117
11674 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
11675 * tree-eh.c: Include gimplify.h.
11676 (find_trapping_overflow, replace_trapping_overflow,
11677 rewrite_to_non_trapping_overflow): New functions.
11678 * tree-vect-loop.c: Include tree-eh.h.
11679 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
11680 * tree-data-ref.c: Include tree-eh.h.
11681 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
11683 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
11685 PR rtl-optimization/84123
11686 * combine.c (change_zero_ext): Check if hard register satisfies
11687 can_change_dest_mode before calling gen_lowpart_SUBREG.
11689 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
11692 * ira.c (ira_init_register_move_cost): Remove assert.
11694 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11696 PR rtl-optimization/84071
11697 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
11698 * doc/tm.texi: Regenerate.
11700 2018-01-31 Richard Biener <rguenther@suse.de>
11702 PR tree-optimization/84132
11703 * tree-data-ref.c (analyze_miv_subscript): Properly
11704 check whether evolution_function_is_affine_multivariate_p
11705 before calling gcd_of_steps_may_divide_p.
11707 2018-01-31 Julia Koval <julia.koval@intel.com>
11710 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
11711 * config/i386/i386.md (rdpid_rex64) New.
11712 (rdpid): Make 32bit only.
11714 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
11717 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
11718 an IDENTIFIER_NODE for FUNCTION_TYPE's.
11720 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11723 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11725 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11727 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11729 PR rtl-optimization/84071
11730 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
11731 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
11733 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11735 * config/arc/arc.c (arc_handle_aux_attribute): New function.
11736 (arc_attribute_table): Add 'aux' attribute.
11737 (arc_in_small_data_p): Consider aux like variables.
11738 (arc_is_aux_reg_p): New function.
11739 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
11740 (arc_get_aux_arg): New function.
11741 (prepare_move_operands): Handle aux-register access.
11742 (arc_handle_aux_attribute): New function.
11743 * doc/extend.texi (ARC Variable attributes): Add subsection.
11745 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11747 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
11748 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
11749 (arc_attribute_table): Add 'uncached' attribute.
11750 (arc_print_operand): Print '.di' flag for uncached memory
11752 (arc_in_small_data_p): Do not consider for small data the uncached
11754 (arc_is_uncached_mem_p): New function.
11755 * config/arc/predicates.md (compact_store_memory_operand): Check
11756 for uncached memory accesses.
11757 (nonvol_nonimm_operand): Likewise.
11758 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
11760 2018-01-31 Jakub Jelinek <jakub@redhat.com>
11763 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
11764 falign-loops=): Add Optimization flag.
11766 2018-01-30 Jeff Law <law@redhat.com>
11769 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
11770 INT_REGISTERS_SAVED. Check it prior to calling
11771 get_scratch_register_on_entry.
11772 (ix86_adjust_stack_and_probe): Similarly.
11773 (ix86_emit_probe_stack_range): Similarly.
11774 (ix86_expand_prologue): Corresponding changes.
11776 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11779 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
11780 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
11782 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
11785 * lra-constraints.c (curr_insn_transform): Process AND in the
11788 2018-01-30 Jakub Jelinek <jakub@redhat.com>
11790 PR rtl-optimization/83986
11791 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
11792 dependence against last_pending_memory_flush in addition to
11793 pending_jump_insns.
11795 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
11797 PR tree-optimization/81611
11798 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
11801 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11804 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
11807 2018-01-30 Richard Biener <rguenther@suse.de>
11808 Jakub Jelinek <jakub@redhat.com>
11810 PR tree-optimization/84111
11811 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
11812 inner loops added during recursion, as they don't have up-to-date
11815 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11818 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
11819 (can_inline_edge_by_limits_p): ... here.
11820 (can_early_inline_edge_p, check_callers,
11821 update_caller_keys, update_callee_keys, recursive_inlining,
11822 add_new_edges_to_heap, speculation_useful_p,
11823 inline_small_functions,
11824 inline_small_functions, flatten_function,
11825 inline_to_all_callers_1): Update.
11827 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11829 * profile-count.c (profile_count::combine_with_ipa_count): Handle
11832 2018-01-30 Richard Biener <rguenther@suse.de>
11834 PR tree-optimization/83008
11835 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
11836 invariant and constant vector uses in stmts when they need
11837 more than one stmt.
11839 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11842 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
11843 * configure: Regenerate.
11845 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11847 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
11849 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
11850 Use gen_rtx_REG rather than gen_lowpart.
11852 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11854 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
11855 rather than 0 when creating partial subregs.
11857 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11859 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
11862 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
11865 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
11866 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
11867 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
11868 flags. This restores the settings used before the 2017-07-24.
11869 Turning off pre increment/decrement/modify allows IVOPTS to
11870 optimize DF/SF loops where the index is an int.
11872 2018-01-29 Richard Biener <rguenther@suse.de>
11873 Kelvin Nilsen <kelvin@gcc.gnu.org>
11876 * tree-vect-stmts.c (vectorizable_call): Don't call
11877 targetm.vectorize_builtin_md_vectorized_function if callee is
11880 2018-01-22 Carl Love <cel@us.ibm.com>
11882 * doc/extend.tex: Fix typo in second arg in
11883 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
11885 2018-01-29 Richard Biener <rguenther@suse.de>
11887 PR tree-optimization/84086
11888 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
11889 (flush_ssaname_freelist): When SSA names were released reset
11890 the SCEV hash table.
11892 2018-01-29 Richard Biener <rguenther@suse.de>
11894 PR tree-optimization/84057
11895 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
11896 removed paths when removing edges.
11898 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
11900 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
11901 -mfunction-return=@var{choice}.
11903 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
11905 PR diagnostic/84034
11906 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
11907 Handle CR like TAB.
11908 (layout::print_source_line): Likewise.
11909 (test_get_line_width_without_trailing_whitespace): Add test cases.
11911 2018-01-27 Jakub Jelinek <jakub@redhat.com>
11913 PR middle-end/84040
11914 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
11917 2018-01-26 Jim Wilson <jimw@sifive.com>
11919 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
11921 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
11924 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11926 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
11927 and CMP + SUB-immediate -> SUBS.
11929 2018-01-26 Martin Sebor <msebor@redhat.com>
11931 PR tree-optimization/83896
11932 * tree-ssa-strlen.c (get_string_len): Rename...
11933 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
11934 Avoid assuming length is constant.
11935 (handle_char_store): Use HOST_WIDE_INT for string length.
11937 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
11940 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
11941 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
11943 2018-01-26 Richard Biener <rguenther@suse.de>
11945 PR rtl-optimization/84003
11946 * dse.c (record_store): Only record redundant stores when
11947 the earlier store aliases at least all accesses the later one does.
11949 2018-01-26 Jakub Jelinek <jakub@redhat.com>
11951 PR rtl-optimization/83985
11952 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
11953 REG_CFA_RESTORE insns.
11954 (delete_unmarked_insns): Don't ignore separate shrink wrapping
11955 REG_CFA_RESTORE insns here.
11958 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
11959 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
11961 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11963 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
11964 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
11965 (arc_init): Likewise.
11966 (arc_override_options): Likewise.
11967 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
11969 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
11971 * config/arc/arc.h (TARGET_DBNZ): Define.
11972 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
11973 properly set the tune attribute.
11974 (dbnz): Use TARGET_DBNZ guard.
11975 * config/arc/arc.opt (mtune): Add core3 option.
11977 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11979 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
11980 recognize new pic like addresses.
11981 (arc_delegitimize_address): Clean up.
11983 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11985 * config/arc/arc-arches.def: Option mrf16 valid for all
11987 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
11988 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
11989 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
11990 * config/arc/arc-tables.opt: Regenerate.
11991 * config/arc/arc.c (arc_conditional_register_usage): Handle
11992 reduced register file case.
11993 (arc_file_start): Set must have build attributes.
11994 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
11995 mrf16 option value.
11996 * config/arc/arc.opt (mrf16): Add new option.
11997 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
11998 * config/arc/genmultilib.awk: Handle new mrf16 option.
11999 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
12000 * config/arc/t-multilib: Regenerate.
12001 * doc/invoke.texi (ARC Options): Document mrf16 option.
12003 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12005 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
12006 * config/arc/arc.c (arc_handle_secure_attribute): New function.
12007 (arc_attribute_table): Add 'secure_call' attribute.
12008 (arc_print_operand): Print secure call operand.
12009 (arc_function_ok_for_sibcall): Don't optimize tail calls when
12011 (arc_is_secure_call_p): New function. * config/arc/arc.md
12012 (call_i): Add support for sjli instruction.
12013 (call_value_i): Likewise.
12014 * config/arc/constraints.md (Csc): New constraint.
12016 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12017 John Eric Martin <John.Martin@emmicro-us.com>
12019 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
12020 * config/arc/arc.c (_arc_jli_section): New struct.
12021 (arc_jli_section): New type.
12022 (rc_jli_sections): New static variable.
12023 (arc_handle_jli_attribute): New function.
12024 (arc_attribute_table): Add jli_always and jli_fixed attribute.
12025 (arc_file_end): New function.
12026 (TARGET_ASM_FILE_END): Define.
12027 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
12028 (arc_add_jli_section): New function.
12029 (jli_call_scan): Likewise.
12030 (arc_reorg): Call jli_call_scan.
12031 (arc_output_addsi): Remove 'S' from printing asm operand.
12032 (arc_is_jli_call_p): New function.
12033 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
12035 (movhi_insn): Likewise.
12036 (movsi_insn): Likewise.
12037 (movsi_set_cc_insn): Likewise.
12038 (loadqi_update): Likewise.
12039 (load_zeroextendqisi_update): Likewise.
12040 (load_signextendqisi_update): Likewise.
12041 (loadhi_update): Likewise.
12042 (load_zeroextendhisi_update): Likewise.
12043 (load_signextendhisi_update): Likewise.
12044 (loadsi_update): Likewise.
12045 (loadsf_update): Likewise.
12046 (movsicc_insn): Likewise.
12047 (bset_insn): Likewise.
12048 (bxor_insn): Likewise.
12049 (bclr_insn): Likewise.
12050 (bmsk_insn): Likewise.
12051 (bicsi3_insn): Likewise.
12052 (cmpsi_cc_c_insn): Likewise.
12053 (movsi_ne): Likewise.
12054 (movsi_cond_exec): Likewise.
12055 (clrsbsi2): Likewise.
12056 (norm_f): Likewise.
12066 (call_i): Remove 'S' asm letter, add jli instruction.
12067 (call_value_i): Likewise.
12068 * config/arc/arc.op (mjli-always): New option.
12069 * config/arc/constraints.md (Cji): New constraint.
12070 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
12072 (subsf3_fpx): Likewise.
12073 (mulsf3_fpx): Likewise.
12074 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
12076 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
12077 function attrbutes.
12078 * doc/invoke.texi (ARC): Document mjli-always option.
12080 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
12082 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
12083 avoid addition with 0 and use incw and decw where possible.
12085 2018-01-26 Richard Biener <rguenther@suse.de>
12087 PR tree-optimization/81082
12088 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
12089 association if it requires casting to unsigned.
12090 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
12091 from fold_plusminus_mult_expr to catch important cases late when
12092 range info is available.
12094 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12096 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
12097 * configure.ac (hidden_linkonce): New test.
12098 * configure: Regenerate.
12099 * config.in: Regenerate.
12101 2018-01-26 Julia Koval <julia.koval@intel.com>
12103 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
12104 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
12105 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
12106 _mm_mask_bitshuffle_epi64_mask): Fix type.
12107 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
12108 USI_FTYPE_V4DI_V4DI_USI): Remove.
12109 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
12110 __builtin_ia32_vpshufbitqmb256_mask,
12111 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
12112 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
12113 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
12115 2018-01-26 Alan Modra <amodra@gmail.com>
12118 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
12119 UNSPEC_VBPERMQ. Sort other unspecs.
12121 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
12123 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
12125 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
12127 PR middle-end/83055
12128 * predict.c (drop_profile): Do not push/pop cfun; update also
12130 (handle_missing_profiles): Fix logic looking for zero profiles.
12132 2018-01-25 Jakub Jelinek <jakub@redhat.com>
12134 PR middle-end/83977
12135 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
12136 on functions with #pragma omp declare simd or functions with simd
12138 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
12139 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
12140 Remove trailing \n from warning_at calls.
12142 2018-01-25 Tom de Vries <tom@codesourcery.com>
12145 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12146 for neutered workers.
12148 2018-01-24 Joseph Myers <joseph@codesourcery.com>
12151 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
12152 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
12154 2018-01-24 Jeff Law <law@redhat.com>
12157 * i386.c (get_probe_interval): Move to earlier point.
12158 (ix86_compute_frame_layout): If -fstack-clash-protection and
12159 the frame is larger than the probe interval, then use pushes
12160 to save registers rather than reg->mem moves.
12161 (ix86_expand_prologue): Remove conditional for int_registers_saved
12164 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
12167 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
12168 min/max for never referenced object.
12170 2018-01-24 Jakub Jelinek <jakub@redhat.com>
12172 PR middle-end/83977
12173 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
12175 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
12176 attributes from DECL_ATTRIBUTES (decl) without affecting
12177 DECL_ATTRIBUTES (current_function_decl).
12178 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
12179 functions with non-NULL DECL_ABSTRACT_ORIGIN.
12181 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
12183 PR tree-optimization/83979
12184 * fold-const.c (fold_comparison): Use constant_boolean_node
12185 instead of boolean_{true,false}_node.
12187 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
12189 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
12192 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12194 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12195 Simplify the clause that sets the length attribute.
12196 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12197 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
12198 clause that sets the length attribute.
12199 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12201 2018-01-24 Tom de Vries <tom@codesourcery.com>
12204 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
12205 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
12206 Add strict parameter.
12207 (prevent_branch_around_nothing): Insert dummy insn between branch to
12208 label and label with no ptx insn inbetween.
12209 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
12211 2018-01-24 Tom de Vries <tom@codesourcery.com>
12214 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12215 for neutered threads in warp.
12216 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
12218 2018-01-24 Richard Biener <rguenther@suse.de>
12220 PR tree-optimization/83176
12221 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
12224 2018-01-24 Richard Biener <rguenther@suse.de>
12226 PR tree-optimization/82819
12227 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
12228 code generating pluses that are no-ops in the target precision.
12230 2018-01-24 Richard Biener <rguenther@suse.de>
12232 PR middle-end/84000
12233 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
12235 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12237 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
12238 to merge probabilities.
12239 * predict.c (probably_never_executed): Also mark as cold functions
12240 with global 0 profile and guessed local profile.
12241 * profile-count.c (profile_probability::combine_with_count): New
12243 * profile-count.h (profile_probability::operator*,
12244 profile_probability::operator*=, profile_probability::operator/,
12245 profile_probability::operator/=): Reduce precision to adjusted
12246 and set value to guessed on contradictory divisions.
12247 (profile_probability::combine_with_freq): Remove.
12248 (profile_probability::combine_wiht_count): Declare.
12249 (profile_count::force_nonzero):: Set to adjusted.
12250 (profile_count::probability_in):: Set quality to adjusted.
12251 * tree-ssa-tail-merge.c (replace_block_by): Use
12252 combine_with_count.
12254 2018-01-23 Andrew Waterman <andrew@sifive.com>
12255 Jim Wilson <jimw@sifive.com>
12257 * config/riscv/riscv.c (riscv_stack_boundary): New.
12258 (riscv_option_override): Set riscv_stack_boundary. Handle
12259 riscv_preferred_stack_boundary_arg.
12260 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
12261 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
12262 (STACK_BOUNDARY): Set to riscv_stack_boundary.
12263 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
12264 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
12265 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
12267 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
12270 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
12271 of struct ix86_frame.
12272 (ix86_expand_epilogue): Likewise. Add a local variable for
12273 the reg_save_offset field in struct ix86_frame.
12275 2018-01-23 Bin Cheng <bin.cheng@arm.com>
12277 PR tree-optimization/82604
12278 * tree-loop-distribution.c (enum partition_kind): New enum item
12279 PKIND_PARTIAL_MEMSET.
12280 (partition_builtin_p): Support above new enum item.
12281 (generate_code_for_partition): Ditto.
12282 (compute_access_range): Differentiate cases that equality can be
12283 proven at all loops, the innermost loops or no loops.
12284 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
12285 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
12286 (finalize_partitions, distribute_loop): Don't fuse partition of
12287 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
12288 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
12289 parloop is enabled.
12291 2018-01-23 Martin Liska <mliska@suse.cz>
12293 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
12294 order to ignore the predictor.
12295 (PRED_POLYMORPHIC_CALL): Likewise.
12296 (PRED_RECURSIVE_CALL): Likewise.
12298 2018-01-23 Martin Liska <mliska@suse.cz>
12300 * tree-profile.c (tree_profiling): Print function header to
12301 aware reader which function we are working on.
12302 * value-prof.c (gimple_find_values_to_profile): Do not print
12303 not interesting value histograms.
12305 2018-01-23 Martin Liska <mliska@suse.cz>
12307 * profile-count.h (enum profile_quality): Add
12308 profile_uninitialized as the first value. Do not number values
12309 as they are zero based.
12310 (profile_count::verify): Update sanity check.
12311 (profile_probability::verify): Likewise.
12313 2018-01-23 Nathan Sidwell <nathan@acm.org>
12315 * doc/invoke.texi (ffor-scope): Deprecate.
12317 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12319 PR tree-optimization/83510
12320 * domwalk.c (set_all_edges_as_executable): New function.
12321 (dom_walker::dom_walker): Convert bool param
12322 "skip_unreachable_blocks" to enum reachability. Move setup of
12323 edge flags to set_all_edges_as_executable and only do it when
12324 reachability is REACHABLE_BLOCKS.
12325 * domwalk.h (enum dom_walker::reachability): New enum.
12326 (dom_walker::dom_walker): Convert bool param
12327 "skip_unreachable_blocks" to enum reachability.
12328 (set_all_edges_as_executable): New decl.
12329 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
12330 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
12332 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
12333 but converting true to REACHABLE_BLOCKS.
12334 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
12336 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
12337 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
12338 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
12340 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
12341 if check_all_array_refs will be called.
12343 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12345 * tree.c (selftest::test_location_wrappers): Add more test
12348 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12350 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
12351 (selftest::test_bit_in_range): Likewise.
12353 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12356 * doc/sourcebuild.texi (vect_float): Say that the selector
12357 only describes the situation when -funsafe-math-optimizations is on.
12358 (vect_float_strict): Document.
12360 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12362 PR tree-optimization/83965
12363 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
12364 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
12365 instead of checking only for a reduction.
12366 (vect_recog_widen_sum_pattern): Likewise.
12368 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12370 * predict.c (probably_never_executed): Only use precise profile info.
12371 (compute_function_frequency): Skip after inlining hack since we now
12372 have quality checking.
12374 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12376 * profile-count.h (profile_probability::very_unlikely,
12377 profile_probability::unlikely, profile_probability::even): Set
12378 precision to guessed.
12380 2018-01-23 Richard Biener <rguenther@suse.de>
12382 PR tree-optimization/83963
12383 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
12384 Properly terminate dominator walk when crossing the exit edge not
12385 when visiting its source block.
12387 2018-01-23 Jakub Jelinek <jakub@redhat.com>
12390 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
12391 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
12393 2018-01-22 Jakub Jelinek <jakub@redhat.com>
12395 PR tree-optimization/83957
12396 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
12397 semicolon after for body surrounded by braces.
12399 PR tree-optimization/83081
12400 * profile-count.h (profile_probability::split): New method.
12401 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
12402 Use profile_probability::split.
12403 (do_compare_rtx_and_jump): Fix adjustment of probabilities
12404 when splitting a single conditional jump into 2.
12406 2018-01-22 David Malcolm <dmalcolm@redhat.com>
12408 PR tree-optimization/69452
12409 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
12412 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12414 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
12415 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
12416 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
12418 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12420 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
12422 * config/rl78/rl78.md (movdi): New define_expand.
12423 * config/rl78/rl78.c (rl78_split_movdi): New function.
12425 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
12428 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
12430 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
12431 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
12432 128-bit to produce an UNSPEC move to get the double word with the
12433 signbit and then a shift directly to do signbit.
12434 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
12435 implementation with a new version that just does either a direct
12436 move or a regular move. Move memory interface to separate insns.
12437 Move insns so they are next to the expander.
12438 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
12439 with signbit move. Split big and little endian case.
12440 (signbit<mode>2_dm_mem_le): Likewise.
12441 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
12442 (signbit<mode>2_dm2): Likewise.
12444 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12446 * config/rl78/rl78.md (anddi3): New define_expand.
12448 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12450 * config/rl78/rl78.md (umindi3): New define_expand.
12452 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12454 * config/rl78/rl78.md (smindi3): New define_expand.
12456 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12458 * config/rl78/rl78.md (smaxdi3): New define_expand.
12460 2018-01-22 Carl Love <cel@us.ibm.com>
12462 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
12463 LVX_V1TI): Add macro expansion.
12464 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
12465 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
12466 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
12467 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
12468 Change check to determine if the instruction is a byte reversing
12469 entry. Fix typo in comment.
12470 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
12471 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
12472 Add def_builtin calls for new builtins.
12473 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
12474 Add define_insn expansion.
12476 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12478 * config/rl78/rl78.md (umaxdi3): New define_expand.
12480 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12482 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
12483 for non-QImode registers.
12485 2018-01-22 Richard Biener <rguenther@suse.de>
12487 PR tree-optimization/83963
12488 * graphite-scop-detection.c (scop_detection::get_sese): Delay
12489 including the loop exit block.
12490 (scop_detection::merge_sese): Likewise.
12491 (scop_detection::add_scop): Do it here instead.
12493 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12495 * doc/sourcebuild.texi (arm_softfloat): Document.
12497 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
12500 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
12501 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
12502 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
12504 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12505 David Edelsohn <dje.gcc@gmail.com>
12508 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12509 Change "crset eq" to "crset 2".
12510 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12511 (*call_indirect_aix<mode>_nospec): Likewise.
12512 (*call_value_indirect_aix<mode>_nospec): Likewise.
12513 (*call_indirect_elfv2<mode>_nospec): Likewise.
12514 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
12515 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
12516 change assembly output from . to $.
12517 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12518 (indirect_jump<mode>_nospec): Change assembly output from . to $.
12519 (*tablejump<mode>_internal1_nospec): Likewise.
12521 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
12524 * config/sh/sh_optimize_sett_clrt.cc:
12525 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
12527 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
12529 PR tree-optimization/83940
12530 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
12531 offset_dt to vect_constant_def rather than vect_unknown_def_type.
12532 (vect_check_load_store_mask): Add a mask_dt_out parameter and
12533 use it to pass back the definition type.
12534 (vect_check_store_rhs): Likewise rhs_dt_out.
12535 (vect_build_gather_load_calls): Add a mask_dt argument and use
12536 it instead of a call to vect_is_simple_use.
12537 (vectorizable_store): Update calls to vect_check_load_store_mask
12538 and vect_check_store_rhs. Use the dt returned by the latter instead
12539 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
12540 instead of calls to vect_is_simple_use. Pass the scalar rather
12541 than the vector operand to vect_is_simple_use when handling
12542 second and subsequent copies of an rhs value.
12543 (vectorizable_load): Update calls to vect_check_load_store_mask
12544 and vect_build_gather_load_calls. Use the cached mask_dt and
12545 gs_info.offset_dt instead of calls to vect_is_simple_use.
12547 2018-01-20 Jakub Jelinek <jakub@redhat.com>
12549 PR middle-end/83945
12550 * tree-emutls.c: Include gimplify.h.
12551 (lower_emutls_2): New function.
12552 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
12553 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
12554 it before further processing.
12557 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
12558 UINTVAL (trueop1) instead of INTVAL (op1).
12560 2018-01-19 Jakub Jelinek <jakub@redhat.com>
12564 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
12565 INCOMING_FRAME_SP_OFFSET if not defined.
12566 (scan_trace): Add ENTRY argument. If true and
12567 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
12568 emit a note to adjust the CFA offset.
12569 (create_cfi_notes): Adjust scan_trace callers.
12570 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
12571 INCOMING_FRAME_SP_OFFSET in the CIE.
12572 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
12573 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
12575 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
12576 * doc/tm.texi: Regenerated.
12578 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12580 PR rtl-optimization/83147
12581 * lra-constraints.c (remove_inheritance_pseudos): Use
12582 lra_substitute_pseudo_within_insn.
12584 2018-01-19 Tom de Vries <tom@codesourcery.com>
12585 Cesar Philippidis <cesar@codesourcery.com>
12588 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
12590 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
12593 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
12594 spaces for function labels.
12596 2018-01-19 Martin Liska <mliska@suse.cz>
12598 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
12599 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
12600 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
12601 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
12602 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
12603 (PRED_CONST_RETURN): Change from 69 to 65.
12604 (PRED_NULL_RETURN): Change from 91 to 71.
12605 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
12606 (PRED_LOOP_GUARD): Change from 66 to 73.
12608 2018-01-19 Martin Liska <mliska@suse.cz>
12610 * predict.c (predict_insn_def): Add new assert.
12611 (struct branch_predictor): Change type to signed integer.
12612 (test_prediction_value_range): Amend test to cover
12613 PROB_UNINITIALIZED.
12614 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
12615 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
12616 (PRED_LOOP_ITERATIONS_MAX): Likewise.
12617 (PRED_LOOP_IV_COMPARE): Likewise.
12618 * predict.h (PROB_UNINITIALIZED): Define new constant.
12620 2018-01-19 Martin Liska <mliska@suse.cz>
12622 * predict.c (dump_prediction): Add new format for
12623 analyze_brprob.py script which is enabled with -details
12625 * profile-count.h (precise_p): New function.
12627 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12629 PR tree-optimization/83922
12630 * tree-vect-loop.c (vect_verify_full_masking): Return false if
12631 there are no statements that need masking.
12632 (vect_active_double_reduction_p): New function.
12633 (vect_analyze_loop_operations): Use it when handling phis that
12634 are not in the loop header.
12636 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12638 PR tree-optimization/83914
12639 * tree-vect-loop.c (vectorizable_induction): Don't convert
12640 init_expr or apply the peeling adjustment for inductions
12641 that are nested within the vectorized loop.
12643 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12645 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
12648 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12652 * function.h (gimplify_parameters): Add gimple_seq * argument.
12653 * function.c: Include gimple.h and options.h.
12654 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
12655 for the added local temporaries if needed.
12656 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
12657 if there are any parameter cleanups, wrap whole body into a
12658 try/finally with the cleanups.
12660 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
12663 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
12664 Use GET_MODE_CLASS for scalar floating point.
12666 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12670 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
12671 Fix call of call_cgraph_insertion_hooks.
12673 2018-01-18 Martin Sebor <msebor@redhat.com>
12675 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
12677 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12680 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
12683 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
12686 * common.opt: (-ffile-prefix-map): New option.
12687 * opts.c (common_handle_option): Defer it.
12688 * opts-global.c (handle_common_deferred_options): Handle it.
12689 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
12690 * file-prefix-map.h: New file.
12691 (remap_debug_filename, add_debug_prefix_map): ...here.
12692 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
12693 * final.c (debug_prefix_map, add_debug_prefix_map
12694 remap_debug_filename): Move to...
12695 * file-prefix-map.c: New file.
12696 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
12697 generalize, get rid of alloca(), use strrchr() instead of strchr().
12698 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
12699 Implement in terms of add_prefix_map().
12700 (remap_macro_filename, remap_debug_filename): Implement in term of
12702 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
12703 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
12704 * dbxout.c: Include file-prefix-map.h.
12705 * varasm.c: Likewise.
12706 * vmsdbgout.c: Likewise.
12707 * xcoffout.c: Likewise.
12708 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
12709 * doc/cppopts.texi (-fmacro-prefix-map): Document.
12710 * doc/invoke.texi (-ffile-prefix-map): Document.
12711 (-fdebug-prefix-map): Update description.
12713 2018-01-18 Martin Liska <mliska@suse.cz>
12715 * config/i386/i386.c (indirect_thunk_name): Document that also
12717 (output_indirect_thunk): Document why both instructions
12718 (pause and lfence) are generated.
12720 2018-01-18 Richard Biener <rguenther@suse.de>
12722 PR tree-optimization/83887
12723 * graphite-scop-detection.c
12724 (scop_detection::get_nearest_dom_with_single_entry): Remove.
12725 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
12726 (scop_detection::merge_sese): Re-implement with a flood-fill
12727 algorithm that properly finds a SESE region if it exists.
12729 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12732 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
12733 pointer_diff optimizations use view_convert instead of convert.
12735 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12737 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12738 Generate different code for -mno-speculate-indirect-jumps.
12739 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12740 (*call_indirect_aix<mode>): Disable for
12741 -mno-speculate-indirect-jumps.
12742 (*call_indirect_aix<mode>_nospec): New define_insn.
12743 (*call_value_indirect_aix<mode>): Disable for
12744 -mno-speculate-indirect-jumps.
12745 (*call_value_indirect_aix<mode>_nospec): New define_insn.
12746 (*sibcall_nonlocal_sysv<mode>): Generate different code for
12747 -mno-speculate-indirect-jumps.
12748 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12750 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12752 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
12753 long double type, set the flags for noting the default long double
12754 type, even if we don't pass or return a long double type.
12756 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
12759 * ipa-inline.c (flatten_function): Do not overwrite final inlining
12762 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
12764 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
12765 support for merge[hl].
12766 (fold_mergehl_helper): New helper function.
12767 (tree-vector-builder.h): New #include for tree_vector_builder usage.
12768 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
12769 (altivec_vmrglw_direct): Add xxmrglw insn.
12771 2018-01-17 Andrew Waterman <andrew@sifive.com>
12773 * config/riscv/riscv.c (riscv_conditional_register_usage): If
12774 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
12776 2018-01-17 David Malcolm <dmalcolm@redhat.com>
12779 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
12780 call the lto_location_cache before reading the
12781 DECL_SOURCE_LOCATION of the types.
12783 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
12784 Richard Sandiford <richard.sandiford@linaro.org>
12786 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
12787 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
12788 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
12789 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
12790 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
12792 * config/aarch64/constraints.md (aarch64_movti_operand):
12794 * config/aarch64/predicates.md (Uti): Add new constraint.
12796 2018-01-17 Carl Love <cel@us.ibm.com>
12798 * config/rs6000/vsx.md (define_expand xl_len_r,
12799 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
12800 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
12802 (define_expand, define_insn): Move the shift left from the
12803 define_insn to the define_expand for lxvl and stxvl instructions.
12804 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
12805 and XL_LEN_R definitions to PURE.
12807 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
12809 * config/i386/i386.c (indirect_thunk_name): Declare regno
12810 as unsigned int. Compare regno with INVALID_REGNUM.
12811 (output_indirect_thunk): Ditto.
12812 (output_indirect_thunk_function): Ditto.
12813 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
12814 in the call to output_indirect_thunk_function.
12816 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
12818 PR middle-end/83884
12819 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
12820 rather than the size of inner_type to determine the stack slot size
12821 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
12823 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
12826 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
12829 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
12831 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
12832 endian Linux systems to optionally enable multilibs for selecting
12833 the long double type if the user configured an explicit type.
12834 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
12835 have no long double multilibs if not defined.
12836 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
12837 warn if the user used -mabi={ieee,ibm}longdouble and we built
12838 multilibs for long double.
12839 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
12840 appropriate multilib option.
12841 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
12843 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
12844 for building long double multilibs.
12845 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
12847 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
12849 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
12852 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
12854 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
12857 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
12860 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
12863 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12865 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
12866 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
12868 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
12870 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
12871 different rtl trees depending on TARGET_64BIT.
12872 (rs6000_gen_lvx): Likewise.
12874 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12876 * config/visium/visium.md (nop): Tweak comment.
12877 (hazard_nop): Likewise.
12879 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12881 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
12882 -mspeculate-indirect-jumps.
12883 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
12884 for -mno-speculate-indirect-jumps.
12885 (*call_indirect_elfv2<mode>_nospec): New define_insn.
12886 (*call_value_indirect_elfv2<mode>): Disable for
12887 -mno-speculate-indirect-jumps.
12888 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
12889 (indirect_jump): Emit different RTL for
12890 -mno-speculate-indirect-jumps.
12891 (*indirect_jump<mode>): Disable for
12892 -mno-speculate-indirect-jumps.
12893 (*indirect_jump<mode>_nospec): New define_insn.
12894 (tablejump): Emit different RTL for
12895 -mno-speculate-indirect-jumps.
12896 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
12897 (tablejumpsi_nospec): New define_expand.
12898 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
12899 (tablejumpdi_nospec): New define_expand.
12900 (*tablejump<mode>_internal1): Disable for
12901 -mno-speculate-indirect-jumps.
12902 (*tablejump<mode>_internal1_nospec): New define_insn.
12903 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
12906 2018-01-16 Artyom Skrobov tyomitch@gmail.com
12908 * caller-save.c (insert_save): Drop unnecessary parameter. All
12911 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12912 Richard Biener <rguenth@suse.de>
12915 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
12916 return early, inline manually is_gimple_sizepos. Make sure if we
12917 call gimplify_expr we don't end up with a gimple constant.
12918 * tree.c (variably_modified_type_p): Don't return true for
12919 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
12920 * gimplify.h (is_gimple_sizepos): Remove.
12922 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12924 PR tree-optimization/83857
12925 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
12926 vectorizable_live_operation for pure SLP statements.
12927 (vectorizable_live_operation): Handle PHIs.
12929 2018-01-16 Richard Biener <rguenther@suse.de>
12931 PR tree-optimization/83867
12932 * tree-vect-stmts.c (vect_transform_stmt): Precompute
12933 nested_in_vect_loop_p since the scalar stmt may get invalidated.
12935 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12938 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
12939 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
12940 If off is not INTEGER_CST, issue a may not be aligned warning
12941 rather than isn't aligned. Use isn%'t rather than isn't.
12942 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
12944 <case MULT_EXPR>: Improve the case when bottom and one of the
12945 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
12946 operand, in that case check if the other operand is multiple of
12947 bottom divided by the INTEGER_CST operand.
12949 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12952 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
12953 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
12954 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
12955 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
12956 * config/pa/pa.c (pa_function_arg_advance): Likewise.
12957 (pa_function_arg, pa_arg_partial_bytes): Likewise.
12958 (pa_function_arg_size): New function.
12960 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12962 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
12963 in a separate statement.
12965 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12967 PR tree-optimization/83847
12968 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
12969 group gathers and scatters.
12971 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12973 PR rtl-optimization/86620
12974 * params.def (max-sched-ready-insns): Bump minimum value to 1.
12976 PR rtl-optimization/83213
12977 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
12978 to last if both are JUMP_INSNs.
12980 PR tree-optimization/83843
12981 * gimple-ssa-store-merging.c
12982 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
12983 store_immediate_info for bswap/nop orig_stores.
12985 2018-01-15 Andrew Waterman <andrew@sifive.com>
12987 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
12989 <UDIV>: Increase cost if !TARGET_DIV.
12991 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
12993 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
12994 (define_attr "cr_logical_3op"): New.
12995 (cceq_ior_compare): Adjust.
12996 (cceq_ior_compare_complement): Adjust.
12997 (*cceq_rev_compare): Adjust.
12998 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
12999 (is_cracked_insn): Adjust.
13000 (insn_must_be_first_in_group): Adjust.
13001 * config/rs6000/40x.md: Adjust.
13002 * config/rs6000/440.md: Adjust.
13003 * config/rs6000/476.md: Adjust.
13004 * config/rs6000/601.md: Adjust.
13005 * config/rs6000/603.md: Adjust.
13006 * config/rs6000/6xx.md: Adjust.
13007 * config/rs6000/7450.md: Adjust.
13008 * config/rs6000/7xx.md: Adjust.
13009 * config/rs6000/8540.md: Adjust.
13010 * config/rs6000/cell.md: Adjust.
13011 * config/rs6000/e300c2c3.md: Adjust.
13012 * config/rs6000/e500mc.md: Adjust.
13013 * config/rs6000/e500mc64.md: Adjust.
13014 * config/rs6000/e5500.md: Adjust.
13015 * config/rs6000/e6500.md: Adjust.
13016 * config/rs6000/mpc.md: Adjust.
13017 * config/rs6000/power4.md: Adjust.
13018 * config/rs6000/power5.md: Adjust.
13019 * config/rs6000/power6.md: Adjust.
13020 * config/rs6000/power7.md: Adjust.
13021 * config/rs6000/power8.md: Adjust.
13022 * config/rs6000/power9.md: Adjust.
13023 * config/rs6000/rs64.md: Adjust.
13024 * config/rs6000/titan.md: Adjust.
13026 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13028 * config/i386/predicates.md (indirect_branch_operand): Rewrite
13029 ix86_indirect_branch_register logic.
13031 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13033 * config/i386/constraints.md (Bs): Update
13034 ix86_indirect_branch_register check. Don't check
13035 ix86_indirect_branch_register with GOT_memory_operand.
13037 * config/i386/predicates.md (GOT_memory_operand): Don't check
13038 ix86_indirect_branch_register here.
13039 (GOT32_symbol_operand): Likewise.
13041 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13043 * config/i386/predicates.md (constant_call_address_operand):
13044 Rewrite ix86_indirect_branch_register logic.
13045 (sibcall_insn_operand): Likewise.
13047 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13049 * config/i386/constraints.md (Bs): Replace
13050 ix86_indirect_branch_thunk_register with
13051 ix86_indirect_branch_register.
13053 * config/i386/i386.md (indirect_jump): Likewise.
13054 (tablejump): Likewise.
13055 (*sibcall_memory): Likewise.
13056 (*sibcall_value_memory): Likewise.
13057 Peepholes of indirect call and jump via memory: Likewise.
13058 * config/i386/i386.opt: Likewise.
13059 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13060 (GOT_memory_operand): Likewise.
13061 (call_insn_operand): Likewise.
13062 (sibcall_insn_operand): Likewise.
13063 (GOT32_symbol_operand): Likewise.
13065 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13067 PR middle-end/83837
13068 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
13069 type rather than type addr's type points to.
13070 (expand_omp_atomic_mutex): Likewise.
13071 (expand_omp_atomic): Likewise.
13073 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13076 * config/i386/i386.c (output_indirect_thunk_function): Use
13077 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
13078 for __x86_return_thunk.
13080 2018-01-15 Richard Biener <rguenther@suse.de>
13082 PR middle-end/83850
13083 * expmed.c (extract_bit_field_1): Fix typo.
13085 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13088 * config/arm/iterators.md (VF): New mode iterator.
13089 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
13090 Remove integer-related logic from pattern.
13091 (neon_vabd<mode>_3): Likewise.
13093 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13095 PR middle-end/82694
13096 * common.opt (fstrict-overflow): No longer an alias.
13097 (fwrapv-pointer): New option.
13098 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
13099 also for pointer types based on flag_wrapv_pointer.
13100 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
13101 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
13102 opts->x_flag_wrapv got set.
13103 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
13104 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
13105 POINTER_TYPE_OVERFLOW_UNDEFINED.
13106 * match.pd: Likewise in address comparison pattern.
13107 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
13109 2018-01-15 Richard Biener <rguenther@suse.de>
13112 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
13113 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
13114 Reset type names to their identifier if their TYPE_DECL doesn't
13115 have linkage (and thus is used for ODR and devirt).
13116 (save_debug_info_for_decl): Remove.
13117 (save_debug_info_for_type): Likewise.
13118 (add_tree_to_fld_list): Adjust.
13119 * tree-pretty-print.c (dump_generic_node): Make dumping of
13120 type names more robust.
13122 2018-01-15 Richard Biener <rguenther@suse.de>
13124 * BASE-VER: Bump to 8.0.1.
13126 2018-01-14 Martin Sebor <msebor@redhat.com>
13129 * builtins.c (check_access): Avoid warning when the no-warning bit
13132 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
13134 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
13135 * ira-color (allocno_hard_regs_compare): Likewise.
13137 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
13140 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
13141 Use .pushsection/.popsection.
13143 2018-01-14 Martin Sebor <msebor@redhat.com>
13146 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
13148 2018-01-14 Jakub Jelinek <jakub@redhat.com>
13150 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
13151 entry from extra_headers.
13152 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
13153 extra_headers, make the list bitwise identical to the i?86-*-* one.
13155 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13157 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
13158 -mcmodel=large with -mindirect-branch=thunk,
13159 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
13160 -mfunction-return=thunk-extern.
13161 * doc/invoke.texi: Document -mcmodel=large is incompatible with
13162 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
13163 -mfunction-return=thunk and -mfunction-return=thunk-extern.
13165 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13167 * config/i386/i386.c (print_reg): Print the name of the full
13168 integer register without '%'.
13169 (ix86_print_operand): Handle 'V'.
13170 * doc/extend.texi: Document 'V' modifier.
13172 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13174 * config/i386/constraints.md (Bs): Disallow memory operand for
13175 -mindirect-branch-register.
13177 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13178 (GOT_memory_operand): Likewise.
13179 (call_insn_operand): Likewise.
13180 (sibcall_insn_operand): Likewise.
13181 (GOT32_symbol_operand): Likewise.
13182 * config/i386/i386.md (indirect_jump): Call convert_memory_address
13183 for -mindirect-branch-register.
13184 (tablejump): Likewise.
13185 (*sibcall_memory): Likewise.
13186 (*sibcall_value_memory): Likewise.
13187 Disallow peepholes of indirect call and jump via memory for
13188 -mindirect-branch-register.
13189 (*call_pop): Replace m with Bw.
13190 (*call_value_pop): Likewise.
13191 (*sibcall_pop_memory): Replace m with Bs.
13192 * config/i386/i386.opt (mindirect-branch-register): New option.
13193 * doc/invoke.texi: Document -mindirect-branch-register option.
13195 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13197 * config/i386/i386-protos.h (ix86_output_function_return): New.
13198 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
13199 set function_return_type.
13200 (indirect_thunk_name): Add ret_p to indicate thunk for function
13202 (output_indirect_thunk_function): Pass false to
13203 indirect_thunk_name.
13204 (ix86_output_indirect_branch_via_reg): Likewise.
13205 (ix86_output_indirect_branch_via_push): Likewise.
13206 (output_indirect_thunk_function): Create alias for function
13207 return thunk if regno < 0.
13208 (ix86_output_function_return): New function.
13209 (ix86_handle_fndecl_attribute): Handle function_return.
13210 (ix86_attribute_table): Add function_return.
13211 * config/i386/i386.h (machine_function): Add
13212 function_return_type.
13213 * config/i386/i386.md (simple_return_internal): Use
13214 ix86_output_function_return.
13215 (simple_return_internal_long): Likewise.
13216 * config/i386/i386.opt (mfunction-return=): New option.
13217 (indirect_branch): Mention -mfunction-return=.
13218 * doc/extend.texi: Document function_return function attribute.
13219 * doc/invoke.texi: Document -mfunction-return= option.
13221 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13223 * config/i386/i386-opts.h (indirect_branch): New.
13224 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
13225 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
13226 with local indirect jump when converting indirect call and jump.
13227 (ix86_set_indirect_branch_type): New.
13228 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
13229 (indirectlabelno): New.
13230 (indirect_thunk_needed): Likewise.
13231 (indirect_thunk_bnd_needed): Likewise.
13232 (indirect_thunks_used): Likewise.
13233 (indirect_thunks_bnd_used): Likewise.
13234 (INDIRECT_LABEL): Likewise.
13235 (indirect_thunk_name): Likewise.
13236 (output_indirect_thunk): Likewise.
13237 (output_indirect_thunk_function): Likewise.
13238 (ix86_output_indirect_branch_via_reg): Likewise.
13239 (ix86_output_indirect_branch_via_push): Likewise.
13240 (ix86_output_indirect_branch): Likewise.
13241 (ix86_output_indirect_jmp): Likewise.
13242 (ix86_code_end): Call output_indirect_thunk_function if needed.
13243 (ix86_output_call_insn): Call ix86_output_indirect_branch if
13245 (ix86_handle_fndecl_attribute): Handle indirect_branch.
13246 (ix86_attribute_table): Add indirect_branch.
13247 * config/i386/i386.h (machine_function): Add indirect_branch_type
13248 and has_local_indirect_jump.
13249 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
13251 (tablejump): Likewise.
13252 (*indirect_jump): Use ix86_output_indirect_jmp.
13253 (*tablejump_1): Likewise.
13254 (simple_return_indirect_internal): Likewise.
13255 * config/i386/i386.opt (mindirect-branch=): New option.
13256 (indirect_branch): New.
13259 (thunk-inline): Likewise.
13260 (thunk-extern): Likewise.
13261 * doc/extend.texi: Document indirect_branch function attribute.
13262 * doc/invoke.texi: Document -mindirect-branch= option.
13264 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
13267 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
13269 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
13271 * ipa-inline.c (want_inline_small_function_p): Return false if
13272 inlining has already failed with CIF_FINAL_ERROR.
13273 (update_caller_keys): Call want_inline_small_function_p before
13275 (update_callee_keys): Likewise.
13277 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
13279 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
13281 (rs6000_quadword_masked_address_p): Likewise.
13282 (quad_aligned_load_p): Likewise.
13283 (quad_aligned_store_p): Likewise.
13284 (const_load_sequence_p): Add comment to describe the outer-most loop.
13285 (mimic_memory_attributes_and_flags): New function.
13286 (rs6000_gen_stvx): Likewise.
13287 (replace_swapped_aligned_store): Likewise.
13288 (rs6000_gen_lvx): Likewise.
13289 (replace_swapped_aligned_load): Likewise.
13290 (replace_swapped_load_constant): Capitalize argument name in
13291 comment describing this function.
13292 (rs6000_analyze_swaps): Add a third pass to search for vector loads
13293 and stores that access quad-word aligned addresses and replace
13294 with stvx or lvx instructions when appropriate.
13295 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
13296 New function prototype.
13297 (rs6000_quadword_masked_address_p): Likewise.
13298 (rs6000_gen_lvx): Likewise.
13299 (rs6000_gen_stvx): Likewise.
13300 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
13301 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
13302 when memory address is aligned.
13303 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
13304 this split to select lvx instruction when memory address is aligned.
13305 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
13306 instruction when memory address is aligned.
13307 (*vsx_le_perm_load_v16qi): Likewise.
13308 (four unnamed splitters): Modify to select the stvx instruction
13309 when memory is aligned.
13311 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
13313 * predict.c (determine_unlikely_bbs): Handle correctly BBs
13314 which appears in the queue multiple times.
13316 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13317 Alan Hayward <alan.hayward@arm.com>
13318 David Sherwood <david.sherwood@arm.com>
13320 * tree-vectorizer.h (vec_lower_bound): New structure.
13321 (_loop_vec_info): Add check_nonzero and lower_bounds.
13322 (LOOP_VINFO_CHECK_NONZERO): New macro.
13323 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
13324 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
13325 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
13326 fields. Make seg_len the distance travelled, not including the
13328 (dr_direction_indicator): Declare.
13329 (dr_zero_step_indicator): Likewise.
13330 (dr_known_forward_stride_p): Likewise.
13331 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
13333 (runtime_alias_check_p): Allow runtime alias checks with
13335 (operator ==): Compare access_size and align.
13336 (prune_runtime_alias_test_list): Rework for new distinction between
13337 the access_size and seg_len.
13338 (create_intersect_range_checks_index): Likewise. Cope with polynomial
13340 (get_segment_min_max): New function.
13341 (create_intersect_range_checks): Use it.
13342 (dr_step_indicator): New function.
13343 (dr_direction_indicator): Likewise.
13344 (dr_zero_step_indicator): Likewise.
13345 (dr_known_forward_stride_p): Likewise.
13346 * tree-loop-distribution.c (data_ref_segment_size): Return
13347 DR_STEP * (niters - 1).
13348 (compute_alias_check_pairs): Update call to the dr_with_seg_len
13350 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
13351 (vect_preserves_scalar_order_p): New function, split out from...
13352 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
13353 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
13354 (vect_vfa_access_size): New function.
13355 (vect_vfa_align): Likewise.
13356 (vect_compile_time_alias): Take access_size_a and access_b arguments.
13357 (dump_lower_bound): New function.
13358 (vect_check_lower_bound): Likewise.
13359 (vect_small_gap_p): Likewise.
13360 (vectorizable_with_step_bound_p): Likewise.
13361 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
13362 depencies if the vectorization factor is 1. Convert the checks
13363 for nonzero steps into checks on the bounds of DR_STEP. Try using
13364 a bunds check for variable steps if the minimum required step is
13365 relatively small. Update calls to the dr_with_seg_len
13366 constructor and to vect_compile_time_alias.
13367 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
13369 (vect_loop_versioning): Call it.
13370 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
13372 (vect_estimate_min_profitable_iters): Account for any bounds checks.
13374 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13375 Alan Hayward <alan.hayward@arm.com>
13376 David Sherwood <david.sherwood@arm.com>
13378 * doc/sourcebuild.texi (vect_scatter_store): Document.
13379 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
13381 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
13383 * genopinit.c (main): Add supports_vec_scatter_store and
13384 supports_vec_scatter_store_cached to target_optabs.
13385 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
13386 IFN_MASK_SCATTER_STORE.
13387 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
13389 * internal-fn.h (internal_store_fn_p): Declare.
13390 (internal_fn_stored_value_index): Likewise.
13391 * internal-fn.c (scatter_store_direct): New macro.
13392 (expand_scatter_store_optab_fn): New function.
13393 (direct_scatter_store_optab_supported_p): New macro.
13394 (internal_store_fn_p): New function.
13395 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
13396 IFN_MASK_SCATTER_STORE.
13397 (internal_fn_mask_index): Likewise.
13398 (internal_fn_stored_value_index): New function.
13399 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
13400 for scatter stores.
13401 * optabs-query.h (supports_vec_scatter_store_p): Declare.
13402 * optabs-query.c (supports_vec_scatter_store_p): New function.
13403 * tree-vectorizer.h (vect_get_store_rhs): Declare.
13404 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
13405 true for scatter stores.
13406 (vect_gather_scatter_fn_p): Handle scatter stores too.
13407 (vect_check_gather_scatter): Consider using scatter stores if
13408 supports_vec_scatter_store_p.
13409 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
13410 scatter stores too.
13411 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13412 internal_fn_stored_value_index.
13413 (check_load_store_masking): Handle scatter stores too.
13414 (vect_get_store_rhs): Make public.
13415 (vectorizable_call): Use internal_store_fn_p.
13416 (vectorizable_store): Handle scatter store internal functions.
13417 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
13418 when deciding whether the end of the group has been reached.
13419 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
13420 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
13421 (mask_scatter_store<mode>): New insns.
13423 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13424 Alan Hayward <alan.hayward@arm.com>
13425 David Sherwood <david.sherwood@arm.com>
13427 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
13428 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
13429 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
13431 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
13432 Use vect_truncate_gather_scatter_offset if we can't treat the
13433 operation as a normal gather load or scatter store.
13434 (get_group_load_store_type): Take the gather_scatter_info
13435 as argument. Try using a gather load or scatter store for
13436 single-element groups.
13437 (get_load_store_type): Update calls to get_group_load_store_type
13438 and vect_use_strided_gather_scatters_p.
13440 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13441 Alan Hayward <alan.hayward@arm.com>
13442 David Sherwood <david.sherwood@arm.com>
13444 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
13445 optional tree argument.
13446 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
13448 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
13449 but continue to use the current value as a fallback.
13450 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
13451 to compare the updates.
13452 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
13453 (get_load_store_type): Use it when handling a strided access.
13454 (vect_get_strided_load_store_ops): New function.
13455 (vect_get_data_ptr_increment): Likewise.
13456 (vectorizable_load): Handle strided gather loads. Always pass
13457 a step to vect_create_data_ref_ptr and bump_vector_ptr.
13459 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13460 Alan Hayward <alan.hayward@arm.com>
13461 David Sherwood <david.sherwood@arm.com>
13463 * doc/md.texi (gather_load@var{m}): Document.
13464 (mask_gather_load@var{m}): Likewise.
13465 * genopinit.c (main): Add supports_vec_gather_load and
13466 supports_vec_gather_load_cached to target_optabs.
13467 * optabs-tree.c (init_tree_optimization_optabs): Use
13468 ggc_cleared_alloc to allocate target_optabs.
13469 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
13470 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
13472 * internal-fn.h (internal_load_fn_p): Declare.
13473 (internal_gather_scatter_fn_p): Likewise.
13474 (internal_fn_mask_index): Likewise.
13475 (internal_gather_scatter_fn_supported_p): Likewise.
13476 * internal-fn.c (gather_load_direct): New macro.
13477 (expand_gather_load_optab_fn): New function.
13478 (direct_gather_load_optab_supported_p): New macro.
13479 (direct_internal_fn_optab): New function.
13480 (internal_load_fn_p): Likewise.
13481 (internal_gather_scatter_fn_p): Likewise.
13482 (internal_fn_mask_index): Likewise.
13483 (internal_gather_scatter_fn_supported_p): Likewise.
13484 * optabs-query.c (supports_at_least_one_mode_p): New function.
13485 (supports_vec_gather_load_p): Likewise.
13486 * optabs-query.h (supports_vec_gather_load_p): Declare.
13487 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
13488 and memory_type field.
13489 (NUM_PATTERNS): Bump to 15.
13490 * tree-vect-data-refs.c: Include internal-fn.h.
13491 (vect_gather_scatter_fn_p): New function.
13492 (vect_describe_gather_scatter_call): Likewise.
13493 (vect_check_gather_scatter): Try using internal functions for
13494 gather loads. Recognize existing calls to a gather load function.
13495 (vect_analyze_data_refs): Consider using gather loads if
13496 supports_vec_gather_load_p.
13497 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
13498 (vect_get_gather_scatter_offset_type): Likewise.
13499 (vect_convert_mask_for_vectype): Likewise.
13500 (vect_add_conversion_to_patterm): Likewise.
13501 (vect_try_gather_scatter_pattern): Likewise.
13502 (vect_recog_gather_scatter_pattern): New pattern recognizer.
13503 (vect_vect_recog_func_ptrs): Add it.
13504 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13505 internal_fn_mask_index and internal_gather_scatter_fn_p.
13506 (check_load_store_masking): Take the gather_scatter_info as an
13507 argument and handle gather loads.
13508 (vect_get_gather_scatter_ops): New function.
13509 (vectorizable_call): Check internal_load_fn_p.
13510 (vectorizable_load): Likewise. Handle gather load internal
13512 (vectorizable_store): Update call to check_load_store_masking.
13513 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
13514 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
13515 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
13516 (aarch64_gather_scale_operand_d): New predicates.
13517 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
13518 (mask_gather_load<mode>): New insns.
13520 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13521 Alan Hayward <alan.hayward@arm.com>
13522 David Sherwood <david.sherwood@arm.com>
13524 * optabs.def (fold_left_plus_optab): New optab.
13525 * doc/md.texi (fold_left_plus_@var{m}): Document.
13526 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
13527 * internal-fn.c (fold_left_direct): Define.
13528 (expand_fold_left_optab_fn): Likewise.
13529 (direct_fold_left_optab_supported_p): Likewise.
13530 * fold-const-call.c (fold_const_fold_left): New function.
13531 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
13532 * tree-parloops.c (valid_reduction_p): New function.
13533 (gather_scalar_reductions): Use it.
13534 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
13535 (vect_finish_replace_stmt): Declare.
13536 * tree-vect-loop.c (fold_left_reduction_fn): New function.
13537 (needs_fold_left_reduction_p): New function, split out from...
13538 (vect_is_simple_reduction): ...here. Accept reductions that
13539 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
13540 (vect_force_simple_reduction): Also store the reduction type in
13541 the assignment's STMT_VINFO_REDUC_TYPE.
13542 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
13543 (merge_with_identity): New function.
13544 (vect_expand_fold_left): Likewise.
13545 (vectorize_fold_left_reduction): Likewise.
13546 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
13547 scalar phi in place for it. Check for target support and reject
13548 cases that would reassociate the operation. Defer the transform
13549 phase to vectorize_fold_left_reduction.
13550 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
13551 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
13552 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
13554 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13556 * tree-if-conv.c (predicate_mem_writes): Remove redundant
13557 call to ifc_temp_var.
13559 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13560 Alan Hayward <alan.hayward@arm.com>
13561 David Sherwood <david.sherwood@arm.com>
13563 * target.def (legitimize_address_displacement): Take the original
13564 offset as a poly_int.
13565 * targhooks.h (default_legitimize_address_displacement): Update
13567 * targhooks.c (default_legitimize_address_displacement): Likewise.
13568 * doc/tm.texi: Regenerate.
13569 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
13570 as an argument, moving assert of ad->disp == ad->disp_term to...
13571 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
13572 Try calling targetm.legitimize_address_displacement before expanding
13573 the address rather than afterwards, and adjust for the new interface.
13574 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
13575 Match the new hook interface. Handle SVE addresses.
13576 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
13577 new hook interface.
13579 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13581 * Makefile.in (OBJS): Add early-remat.o.
13582 * target.def (select_early_remat_modes): New hook.
13583 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
13584 * doc/tm.texi: Regenerate.
13585 * targhooks.h (default_select_early_remat_modes): Declare.
13586 * targhooks.c (default_select_early_remat_modes): New function.
13587 * timevar.def (TV_EARLY_REMAT): New timevar.
13588 * passes.def (pass_early_remat): New pass.
13589 * tree-pass.h (make_pass_early_remat): Declare.
13590 * early-remat.c: New file.
13591 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
13593 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
13595 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13596 Alan Hayward <alan.hayward@arm.com>
13597 David Sherwood <david.sherwood@arm.com>
13599 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
13600 vfm1 with a bound_epilog parameter.
13601 (vect_do_peeling): Update calls accordingly, and move the prologue
13602 call earlier in the function. Treat the base bound_epilog as 0 for
13603 fully-masked loops and retain vf - 1 for other loops. Add 1 to
13604 this base when peeling for gaps.
13605 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
13606 with fully-masked loops.
13607 (vect_estimate_min_profitable_iters): Handle the single peeled
13608 iteration in that case.
13610 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13611 Alan Hayward <alan.hayward@arm.com>
13612 David Sherwood <david.sherwood@arm.com>
13614 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
13615 single-element interleaving even if the size is not a power of 2.
13616 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
13617 accesses for single-element interleaving if the group size is
13620 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13621 Alan Hayward <alan.hayward@arm.com>
13622 David Sherwood <david.sherwood@arm.com>
13624 * doc/md.texi (fold_extract_last_@var{m}): Document.
13625 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
13626 * optabs.def (fold_extract_last_optab): New optab.
13627 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
13628 * internal-fn.c (fold_extract_direct): New macro.
13629 (expand_fold_extract_optab_fn): Likewise.
13630 (direct_fold_extract_optab_supported_p): Likewise.
13631 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
13632 * tree-vect-loop.c (vect_model_reduction_cost): Handle
13633 EXTRACT_LAST_REDUCTION.
13634 (get_initial_def_for_reduction): Do not create an initial vector
13635 for EXTRACT_LAST_REDUCTION reductions.
13636 (vectorizable_reduction): Leave the scalar phi in place for
13637 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
13638 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
13639 epilogue code for EXTRACT_LAST_REDUCTION and defer the
13640 transform phase to vectorizable_condition.
13641 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
13643 (vect_finish_stmt_generation): ...here.
13644 (vect_finish_replace_stmt): New function.
13645 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
13646 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
13648 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
13650 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13651 Alan Hayward <alan.hayward@arm.com>
13652 David Sherwood <david.sherwood@arm.com>
13654 * doc/md.texi (extract_last_@var{m}): Document.
13655 * optabs.def (extract_last_optab): New optab.
13656 * internal-fn.def (EXTRACT_LAST): New internal function.
13657 * internal-fn.c (cond_unary_direct): New macro.
13658 (expand_cond_unary_optab_fn): Likewise.
13659 (direct_cond_unary_optab_supported_p): Likewise.
13660 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
13661 loops using EXTRACT_LAST.
13662 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
13663 (extract_last_<mode>): ...this optab.
13664 (vec_extract<mode><Vel>): Update accordingly.
13666 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13667 Alan Hayward <alan.hayward@arm.com>
13668 David Sherwood <david.sherwood@arm.com>
13670 * target.def (empty_mask_is_expensive): New hook.
13671 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
13672 * doc/tm.texi: Regenerate.
13673 * targhooks.h (default_empty_mask_is_expensive): Declare.
13674 * targhooks.c (default_empty_mask_is_expensive): New function.
13675 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
13676 if the target says that empty masks are expensive.
13677 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
13679 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
13681 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13682 Alan Hayward <alan.hayward@arm.com>
13683 David Sherwood <david.sherwood@arm.com>
13685 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
13686 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
13687 (vect_use_loop_mask_for_alignment_p): New function.
13688 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
13689 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
13690 niters_skip argument. Make sure that the first niters_skip elements
13691 of the first iteration are inactive.
13692 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
13693 Update call to vect_set_loop_masks_directly.
13694 (get_misalign_in_elems): New function, split out from...
13695 (vect_gen_prolog_loop_niters): ...here.
13696 (vect_update_init_of_dr): Take a code argument that specifies whether
13697 the adjustment should be added or subtracted.
13698 (vect_update_init_of_drs): Likewise.
13699 (vect_prepare_for_masked_peels): New function.
13700 (vect_do_peeling): Skip prologue peeling if we're using a mask
13701 instead. Update call to vect_update_inits_of_drs.
13702 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13704 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
13705 alignment. Do not include the number of peeled iterations in
13706 the minimum threshold in that case.
13707 (vectorizable_induction): Adjust the start value down by
13708 LOOP_VINFO_MASK_SKIP_NITERS iterations.
13709 (vect_transform_loop): Call vect_prepare_for_masked_peels.
13710 Take the number of skipped iterations into account when calculating
13712 * tree-vect-stmts.c (vect_gen_while_not): New function.
13714 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13715 Alan Hayward <alan.hayward@arm.com>
13716 David Sherwood <david.sherwood@arm.com>
13718 * doc/sourcebuild.texi (vect_fully_masked): Document.
13719 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
13720 default value to 0.
13721 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
13723 (vect_analyze_loop_2): ...here. Don't check the vectorization
13724 factor against the number of loop iterations if the loop is
13727 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13728 Alan Hayward <alan.hayward@arm.com>
13729 David Sherwood <david.sherwood@arm.com>
13731 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
13732 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
13733 (dump_groups): Update accordingly.
13734 (iv_use::mem_type): New member variable.
13735 (address_p): New function.
13736 (record_use): Add a mem_type argument and initialize the new
13738 (record_group_use): Add a mem_type argument. Use address_p.
13739 Remove obsolete null checks of base_object. Update call to record_use.
13740 (find_interesting_uses_op): Update call to record_group_use.
13741 (find_interesting_uses_cond): Likewise.
13742 (find_interesting_uses_address): Likewise.
13743 (get_mem_type_for_internal_fn): New function.
13744 (find_address_like_use): Likewise.
13745 (find_interesting_uses_stmt): Try find_address_like_use before
13746 calling find_interesting_uses_op.
13747 (addr_offset_valid_p): Use the iv mem_type field as the type
13748 of the addressed memory.
13749 (add_autoinc_candidates): Likewise.
13750 (get_address_cost): Likewise.
13751 (split_small_address_groups_p): Use address_p.
13752 (split_address_groups): Likewise.
13753 (add_iv_candidate_for_use): Likewise.
13754 (autoinc_possible_for_pair): Likewise.
13755 (rewrite_groups): Likewise.
13756 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
13757 (determine_group_iv_cost): Update after split of USE_ADDRESS.
13758 (get_alias_ptr_type_for_ptr_address): New function.
13759 (rewrite_use_address): Rewrite address uses in calls that were
13760 identified by find_address_like_use.
13762 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13763 Alan Hayward <alan.hayward@arm.com>
13764 David Sherwood <david.sherwood@arm.com>
13766 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
13768 * gimple-expr.h (is_gimple_addressable: Likewise.
13769 * gimple-expr.c (is_gimple_address): Likewise.
13770 * internal-fn.c (expand_call_mem_ref): New function.
13771 (expand_mask_load_optab_fn): Use it.
13772 (expand_mask_store_optab_fn): Likewise.
13774 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13775 Alan Hayward <alan.hayward@arm.com>
13776 David Sherwood <david.sherwood@arm.com>
13778 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
13779 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
13780 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
13781 (cond_umax@var{mode}): Document.
13782 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
13783 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
13784 (cond_umin_optab, cond_umax_optab): New optabs.
13785 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
13786 (COND_IOR, COND_XOR): New internal functions.
13787 * internal-fn.h (get_conditional_internal_fn): Declare.
13788 * internal-fn.c (cond_binary_direct): New macro.
13789 (expand_cond_binary_optab_fn): Likewise.
13790 (direct_cond_binary_optab_supported_p): Likewise.
13791 (get_conditional_internal_fn): New function.
13792 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
13793 Cope with reduction statements that are vectorized as calls rather
13795 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
13796 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
13797 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
13798 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13799 (UNSPEC_COND_EOR): New unspecs.
13800 (optab): Add mappings for them.
13801 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
13802 (sve_int_op, sve_fp_op): New int attributes.
13804 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13805 Alan Hayward <alan.hayward@arm.com>
13806 David Sherwood <david.sherwood@arm.com>
13808 * optabs.def (while_ult_optab): New optab.
13809 * doc/md.texi (while_ult@var{m}@var{n}): Document.
13810 * internal-fn.def (WHILE_ULT): New internal function.
13811 * internal-fn.h (direct_internal_fn_supported_p): New override
13812 that takes two types as argument.
13813 * internal-fn.c (while_direct): New macro.
13814 (expand_while_optab_fn): New function.
13815 (convert_optab_supported_p): Likewise.
13816 (direct_while_optab_supported_p): New macro.
13817 * wide-int.h (wi::udiv_ceil): New function.
13818 * tree-vectorizer.h (rgroup_masks): New structure.
13819 (vec_loop_masks): New typedef.
13820 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
13821 and fully_masked_p.
13822 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
13823 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
13824 (vect_max_vf): New function.
13825 (slpeel_make_loop_iterate_ntimes): Delete.
13826 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
13827 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
13828 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
13829 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
13830 internal-fn.h, stor-layout.h and optabs-query.h.
13831 (vect_set_loop_mask): New function.
13832 (add_preheader_seq): Likewise.
13833 (add_header_seq): Likewise.
13834 (interleave_supported_p): Likewise.
13835 (vect_maybe_permute_loop_masks): Likewise.
13836 (vect_set_loop_masks_directly): Likewise.
13837 (vect_set_loop_condition_masked): Likewise.
13838 (vect_set_loop_condition_unmasked): New function, split out from
13839 slpeel_make_loop_iterate_ntimes.
13840 (slpeel_make_loop_iterate_ntimes): Rename to..
13841 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
13842 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
13843 (vect_do_peeling): Update call accordingly.
13844 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
13846 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13847 mask_compare_type, can_fully_mask_p and fully_masked_p.
13848 (release_vec_loop_masks): New function.
13849 (_loop_vec_info): Use it to free the loop masks.
13850 (can_produce_all_loop_masks_p): New function.
13851 (vect_get_max_nscalars_per_iter): Likewise.
13852 (vect_verify_full_masking): Likewise.
13853 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
13854 retries, and free the mask rgroups before retrying. Check loop-wide
13855 reasons for disallowing fully-masked loops. Make the final decision
13856 about whether use a fully-masked loop or not.
13857 (vect_estimate_min_profitable_iters): Do not assume that peeling
13858 for the number of iterations will be needed for fully-masked loops.
13859 (vectorizable_reduction): Disable fully-masked loops.
13860 (vectorizable_live_operation): Likewise.
13861 (vect_halve_mask_nunits): New function.
13862 (vect_double_mask_nunits): Likewise.
13863 (vect_record_loop_mask): Likewise.
13864 (vect_get_loop_mask): Likewise.
13865 (vect_transform_loop): Handle the case in which the final loop
13866 iteration might handle a partial vector. Call vect_set_loop_condition
13867 instead of slpeel_make_loop_iterate_ntimes.
13868 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
13869 (check_load_store_masking): New function.
13870 (prepare_load_store_mask): Likewise.
13871 (vectorizable_store): Handle fully-masked loops.
13872 (vectorizable_load): Likewise.
13873 (supportable_widening_operation): Use vect_halve_mask_nunits for
13875 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
13876 (vect_gen_while): New function.
13877 * config/aarch64/aarch64.md (umax<mode>3): New expander.
13878 (aarch64_uqdec<mode>): New insn.
13880 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13881 Alan Hayward <alan.hayward@arm.com>
13882 David Sherwood <david.sherwood@arm.com>
13884 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
13885 (reduc_xor_scal_optab): New optabs.
13886 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
13887 (reduc_xor_scal_@var{m}): Document.
13888 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
13889 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
13890 internal functions.
13891 * fold-const-call.c (fold_const_call): Handle them.
13892 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
13893 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
13894 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
13895 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
13896 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
13897 (UNSPEC_XORV): New unspecs.
13898 (optab): Add entries for them.
13899 (BITWISEV): New int iterator.
13900 (bit_reduc_op): New int attributes.
13902 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13903 Alan Hayward <alan.hayward@arm.com>
13904 David Sherwood <david.sherwood@arm.com>
13906 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
13907 * internal-fn.def (VEC_SHL_INSERT): New internal function.
13908 * optabs.def (vec_shl_insert_optab): New optab.
13909 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
13910 (duplicate_and_interleave): Likewise.
13911 * tree-vect-loop.c: Include internal-fn.h.
13912 (neutral_op_for_slp_reduction): New function, split out from
13913 get_initial_defs_for_reduction.
13914 (get_initial_def_for_reduction): Handle option 2 for variable-length
13915 vectors by loading the neutral value into a vector and then shifting
13916 the initial value into element 0.
13917 (get_initial_defs_for_reduction): Replace the code argument with
13918 the neutral value calculated by neutral_op_for_slp_reduction.
13919 Use gimple_build_vector for constant-length vectors.
13920 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
13921 but the first group_size elements have a neutral value.
13922 Use duplicate_and_interleave otherwise.
13923 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
13924 Update call to get_initial_defs_for_reduction. Handle SLP
13925 reductions for variable-length vectors by creating one vector
13926 result for each scalar result, with the elements associated
13927 with other scalar results stubbed out with the neutral value.
13928 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
13929 Require IFN_VEC_SHL_INSERT for double reductions on
13930 variable-length vectors, or SLP reductions that have
13931 a neutral value. Require can_duplicate_and_interleave_p
13932 support for variable-length unchained SLP reductions if there
13933 is no neutral value, such as for MIN/MAX reductions. Also require
13934 the number of vector elements to be a multiple of the number of
13935 SLP statements when doing variable-length unchained SLP reductions.
13936 Update call to vect_create_epilog_for_reduction.
13937 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
13938 and remove initial values.
13939 (duplicate_and_interleave): Make public.
13940 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
13941 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
13943 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13944 Alan Hayward <alan.hayward@arm.com>
13945 David Sherwood <david.sherwood@arm.com>
13947 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
13948 (can_duplicate_and_interleave_p): New function.
13949 (vect_get_and_check_slp_defs): Take the vector of statements
13950 rather than just the current one. Remove excess parentheses.
13951 Restriction rejectinon of vect_constant_def and vect_external_def
13952 for variable-length vectors to boolean types, or types for which
13953 can_duplicate_and_interleave_p is false.
13954 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
13955 (duplicate_and_interleave): New function.
13956 (vect_get_constant_vectors): Use gimple_build_vector for
13957 constant-length vectors and suitable variable-length constant
13958 vectors. Use duplicate_and_interleave for other variable-length
13959 vectors. Don't defer the update when inserting new statements.
13961 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13962 Alan Hayward <alan.hayward@arm.com>
13963 David Sherwood <david.sherwood@arm.com>
13965 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
13966 min_profitable_iters doesn't go negative.
13968 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13969 Alan Hayward <alan.hayward@arm.com>
13970 David Sherwood <david.sherwood@arm.com>
13972 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
13973 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
13974 * optabs.def (vec_mask_load_lanes_optab): New optab.
13975 (vec_mask_store_lanes_optab): Likewise.
13976 * internal-fn.def (MASK_LOAD_LANES): New internal function.
13977 (MASK_STORE_LANES): Likewise.
13978 * internal-fn.c (mask_load_lanes_direct): New macro.
13979 (mask_store_lanes_direct): Likewise.
13980 (expand_mask_load_optab_fn): Handle masked operations.
13981 (expand_mask_load_lanes_optab_fn): New macro.
13982 (expand_mask_store_optab_fn): Handle masked operations.
13983 (expand_mask_store_lanes_optab_fn): New macro.
13984 (direct_mask_load_lanes_optab_supported_p): Likewise.
13985 (direct_mask_store_lanes_optab_supported_p): Likewise.
13986 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
13988 (vect_load_lanes_supported): Likewise.
13989 * tree-vect-data-refs.c (strip_conversion): New function.
13990 (can_group_stmts_p): Likewise.
13991 (vect_analyze_data_ref_accesses): Use it instead of checking
13992 for a pair of assignments.
13993 (vect_store_lanes_supported): Take a masked_p parameter.
13994 (vect_load_lanes_supported): Likewise.
13995 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
13996 vect_store_lanes_supported and vect_load_lanes_supported.
13997 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13998 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
13999 parameter. Don't allow gaps for masked accesses.
14000 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
14001 and vect_load_lanes_supported.
14002 (get_load_store_type): Take a masked_p parameter and update
14003 call to get_group_load_store_type.
14004 (vectorizable_store): Update call to get_load_store_type.
14005 Handle IFN_MASK_STORE_LANES.
14006 (vectorizable_load): Update call to get_load_store_type.
14007 Handle IFN_MASK_LOAD_LANES.
14009 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14010 Alan Hayward <alan.hayward@arm.com>
14011 David Sherwood <david.sherwood@arm.com>
14013 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
14015 * config/aarch64/aarch64-protos.h
14016 (aarch64_sve_struct_memory_operand_p): Declare.
14017 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
14018 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
14019 (VPRED, vpred): Handle SVE structure modes.
14020 * config/aarch64/constraints.md (Utx): New constraint.
14021 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
14022 (aarch64_sve_struct_nonimmediate_operand): New predicates.
14023 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
14024 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
14025 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
14026 structure modes. Split into pieces after RA.
14027 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
14028 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
14030 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
14031 SVE structure modes.
14032 (aarch64_classify_address): Likewise.
14033 (sizetochar): Move earlier in file.
14034 (aarch64_print_operand): Handle SVE register lists.
14035 (aarch64_array_mode): New function.
14036 (aarch64_sve_struct_memory_operand_p): Likewise.
14037 (TARGET_ARRAY_MODE): Redefine.
14039 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14040 Alan Hayward <alan.hayward@arm.com>
14041 David Sherwood <david.sherwood@arm.com>
14043 * target.def (array_mode): New target hook.
14044 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
14045 * doc/tm.texi: Regenerate.
14046 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
14047 * hooks.c (hook_optmode_mode_uhwi_none): New function.
14048 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
14049 targetm.array_mode.
14050 * stor-layout.c (mode_for_array): Likewise. Support polynomial
14053 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14054 Alan Hayward <alan.hayward@arm.com>
14055 David Sherwood <david.sherwood@arm.com>
14057 * fold-const.c (fold_binary_loc): Check the argument types
14058 rather than the result type when testing for a vector operation.
14060 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14062 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
14063 * doc/tm.texi: Regenerate.
14065 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14066 Alan Hayward <alan.hayward@arm.com>
14067 David Sherwood <david.sherwood@arm.com>
14069 * doc/invoke.texi (-msve-vector-bits=): Document new option.
14070 (sve): Document new AArch64 extension.
14071 * doc/md.texi (w): Extend the description of the AArch64
14072 constraint to include SVE vectors.
14073 (Upl, Upa): Document new AArch64 predicate constraints.
14074 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
14076 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
14077 (msve-vector-bits=): New option.
14078 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
14079 SVE when these are disabled.
14080 (sve): New extension.
14081 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
14082 modes. Adjust their number of units based on aarch64_sve_vg.
14083 (MAX_BITSIZE_MODE_ANY_MODE): Define.
14084 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
14085 aarch64_addr_query_type.
14086 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
14087 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
14088 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
14089 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
14090 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
14091 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
14092 (aarch64_simd_imm_zero_p): Delete.
14093 (aarch64_check_zero_based_sve_index_immediate): Declare.
14094 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14095 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14096 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14097 (aarch64_sve_float_mul_immediate_p): Likewise.
14098 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14099 rather than an rtx.
14100 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
14101 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
14102 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
14103 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
14104 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
14105 (aarch64_regmode_natural_size): Likewise.
14106 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
14107 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
14109 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
14110 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
14111 for VG and the SVE predicate registers.
14112 (V_ALIASES): Add a "z"-prefixed alias.
14113 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
14114 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
14115 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
14116 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
14117 (REG_CLASS_NAMES): Add entries for them.
14118 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
14119 and the predicate registers.
14120 (aarch64_sve_vg): Declare.
14121 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
14122 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
14123 (REGMODE_NATURAL_SIZE): Define.
14124 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
14126 * config/aarch64/aarch64.c: Include cfgrtl.h.
14127 (simd_immediate_info): Add a constructor for series vectors,
14128 and an associated step field.
14129 (aarch64_sve_vg): New variable.
14130 (aarch64_dbx_register_number): Handle VG and the predicate registers.
14131 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
14132 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
14133 (VEC_ANY_DATA, VEC_STRUCT): New constants.
14134 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
14135 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
14136 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
14137 (aarch64_get_mask_mode): New functions.
14138 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
14139 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
14140 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
14141 predicate modes and predicate registers. Explicitly restrict
14142 GPRs to modes of 16 bytes or smaller. Only allow FP registers
14143 to store a vector mode if it is recognized by
14144 aarch64_classify_vector_mode.
14145 (aarch64_regmode_natural_size): New function.
14146 (aarch64_hard_regno_caller_save_mode): Return the original mode
14148 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
14149 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
14150 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
14151 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
14153 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
14154 does not overlap dest if the function is frame-related. Handle
14156 (aarch64_split_add_offset): New function.
14157 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
14158 them aarch64_add_offset.
14159 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
14160 and update call to aarch64_sub_sp.
14161 (aarch64_add_cfa_expression): New function.
14162 (aarch64_expand_prologue): Pass extra temporary registers to the
14163 functions above. Handle the case in which we need to emit new
14164 DW_CFA_expressions for registers that were originally saved
14165 relative to the stack pointer, but now have to be expressed
14166 relative to the frame pointer.
14167 (aarch64_output_mi_thunk): Pass extra temporary registers to the
14169 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
14170 IP0 and IP1 values for SVE frames.
14171 (aarch64_expand_vec_series): New function.
14172 (aarch64_expand_sve_widened_duplicate): Likewise.
14173 (aarch64_expand_sve_const_vector): Likewise.
14174 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
14175 Handle SVE constants. Use emit_move_insn to move a force_const_mem
14176 into the register, rather than emitting a SET directly.
14177 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
14178 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
14179 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
14180 (offset_9bit_signed_scaled_p): New functions.
14181 (aarch64_replicate_bitmask_imm): New function.
14182 (aarch64_bitmask_imm): Use it.
14183 (aarch64_cannot_force_const_mem): Reject expressions involving
14184 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
14185 (aarch64_classify_index): Handle SVE indices, by requiring
14186 a plain register index with a scale that matches the element size.
14187 (aarch64_classify_address): Handle SVE addresses. Assert that
14188 the mode of the address is VOIDmode or an integer mode.
14189 Update call to aarch64_classify_symbol.
14190 (aarch64_classify_symbolic_expression): Update call to
14191 aarch64_classify_symbol.
14192 (aarch64_const_vec_all_in_range_p): New function.
14193 (aarch64_print_vector_float_operand): Likewise.
14194 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
14195 "vN" for FP registers with SVE modes. Handle (const ...) vectors
14196 and the FP immediates 1.0 and 0.5.
14197 (aarch64_print_address_internal): Handle SVE addresses.
14198 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
14199 (aarch64_regno_regclass): Handle predicate registers.
14200 (aarch64_secondary_reload): Handle big-endian reloads of SVE
14202 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
14203 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
14204 (aarch64_convert_sve_vector_bits): New function.
14205 (aarch64_override_options): Use it to handle -msve-vector-bits=.
14206 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14207 rather than an rtx.
14208 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
14209 Handle SVE vector and predicate modes. Accept VL-based constants
14210 that need only one temporary register, and VL offsets that require
14211 no temporary registers.
14212 (aarch64_conditional_register_usage): Mark the predicate registers
14213 as fixed if SVE isn't available.
14214 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
14215 Return true for SVE vector and predicate modes.
14216 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
14217 rather than an unsigned int. Handle SVE modes.
14218 (aarch64_preferred_simd_mode): Update call accordingly. Handle
14220 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
14222 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14223 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14224 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14225 (aarch64_sve_float_mul_immediate_p): New functions.
14226 (aarch64_sve_valid_immediate): New function.
14227 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
14228 Explicitly reject structure modes. Check for INDEX constants.
14229 Handle PTRUE and PFALSE constants.
14230 (aarch64_check_zero_based_sve_index_immediate): New function.
14231 (aarch64_simd_imm_zero_p): Delete.
14232 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
14233 vector modes. Accept constants in the range of CNT[BHWD].
14234 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
14235 ask for an Advanced SIMD mode.
14236 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
14237 (aarch64_simd_vector_alignment): Handle SVE predicates.
14238 (aarch64_vectorize_preferred_vector_alignment): New function.
14239 (aarch64_simd_vector_alignment_reachable): Use it instead of
14241 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
14242 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
14244 (MAX_VECT_LEN): Delete.
14245 (expand_vec_perm_d): Add a vec_flags field.
14246 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
14247 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14248 (aarch64_evpc_ext): Don't apply a big-endian lane correction
14250 (aarch64_evpc_rev): Rename to...
14251 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
14252 (aarch64_evpc_rev_global): New function.
14253 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
14254 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
14256 (aarch64_evpc_sve_tbl): New function.
14257 (aarch64_expand_vec_perm_const_1): Update after rename of
14258 aarch64_evpc_rev. Handle SVE permutes too, trying
14259 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
14260 than aarch64_evpc_tbl.
14261 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
14262 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
14263 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
14264 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
14265 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
14266 (aarch64_expand_sve_vcond): New functions.
14267 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
14268 of aarch64_vector_mode_p.
14269 (aarch64_dwarf_poly_indeterminate_value): New function.
14270 (aarch64_compute_pressure_classes): Likewise.
14271 (aarch64_can_change_mode_class): Likewise.
14272 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
14273 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
14274 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
14275 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
14276 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
14277 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
14278 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
14279 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
14281 (Dn, Dl, Dr): Accept const as well as const_vector.
14282 (Dz): Likewise. Compare against CONST0_RTX.
14283 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
14284 of "vector" where appropriate.
14285 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
14286 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
14287 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
14288 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
14289 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
14290 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
14291 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
14292 (v_int_equiv): Extend to SVE modes.
14293 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
14295 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
14296 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
14297 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
14298 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
14299 (SVE_COND_FP_CMP): New int iterators.
14300 (perm_hilo): Handle the new unpack unspecs.
14301 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
14303 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
14304 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
14305 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
14306 (aarch64_equality_operator, aarch64_constant_vector_operand)
14307 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
14308 (aarch64_sve_nonimmediate_operand): Likewise.
14309 (aarch64_sve_general_operand): Likewise.
14310 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
14311 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
14312 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
14313 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
14314 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
14315 (aarch64_sve_float_arith_immediate): Likewise.
14316 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
14317 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
14318 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
14319 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
14320 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
14321 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
14322 (aarch64_sve_float_arith_operand): Likewise.
14323 (aarch64_sve_float_arith_with_sub_operand): Likewise.
14324 (aarch64_sve_float_mul_operand): Likewise.
14325 (aarch64_sve_vec_perm_operand): Likewise.
14326 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
14327 (aarch64_mov_operand): Accept const_poly_int and const_vector.
14328 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
14329 as well as const_vector.
14330 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
14331 in file. Use CONST0_RTX and CONSTM1_RTX.
14332 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
14333 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
14334 Use aarch64_simd_imm_zero.
14335 * config/aarch64/aarch64-sve.md: New file.
14336 * config/aarch64/aarch64.md: Include it.
14337 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
14338 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
14339 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
14340 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
14341 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
14342 (sve): New attribute.
14343 (enabled): Disable instructions with the sve attribute unless
14345 (movqi, movhi): Pass CONST_POLY_INT operaneds through
14346 aarch64_expand_mov_immediate.
14347 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
14348 CNT[BHSD] immediates.
14349 (movti): Split CONST_POLY_INT moves into two halves.
14350 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
14351 Split additions that need a temporary here if the destination
14352 is the stack pointer.
14353 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
14354 (*add<mode>3_poly_1): New instruction.
14355 (set_clobber_cc): New expander.
14357 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14359 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
14360 parameter and use it instead of GET_MODE_SIZE (innermode). Use
14361 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
14362 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
14363 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
14364 Change innermode from fixed_mode_size to machine_mode.
14365 (simplify_subreg): Update call accordingly. Handle a constant-sized
14366 subreg of a variable-length CONST_VECTOR.
14368 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14369 Alan Hayward <alan.hayward@arm.com>
14370 David Sherwood <david.sherwood@arm.com>
14372 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
14373 (add_offset_to_base): New function, split out from...
14374 (create_mem_ref): ...here. When handling a scale other than 1,
14375 check first whether the address is valid without the offset.
14376 Add it into the base if so, leaving the index and scale as-is.
14378 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14381 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
14382 fold_for_warn before checking if arg2 is INTEGER_CST.
14384 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
14386 * config/rs6000/predicates.md (load_multiple_operation): Delete.
14387 (store_multiple_operation): Delete.
14388 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
14389 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
14390 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
14391 guarded by TARGET_STRING.
14392 (rs6000_output_load_multiple): Delete.
14393 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
14394 OPTION_MASK_STRING / TARGET_STRING handling.
14395 (print_operand) <'N', 'O'>: Add comment that these are unused now.
14396 (const rs6000_opt_masks) <"string">: Change mask to 0.
14397 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
14398 (MASK_STRING): Delete.
14399 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
14401 (load_multiple): Delete.
14408 (store_multiple): Delete.
14415 (movmemsi_8reg): Delete.
14416 (corresponding unnamed define_insn): Delete.
14417 (movmemsi_6reg): Delete.
14418 (corresponding unnamed define_insn): Delete.
14419 (movmemsi_4reg): Delete.
14420 (corresponding unnamed define_insn): Delete.
14421 (movmemsi_2reg): Delete.
14422 (corresponding unnamed define_insn): Delete.
14423 (movmemsi_1reg): Delete.
14424 (corresponding unnamed define_insn): Delete.
14425 * config/rs6000/rs6000.opt (mno-string): New.
14426 (mstring): Replace by deprecation warning stub.
14427 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
14429 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14431 * regrename.c (regrename_do_replace): If replacing the same
14432 reg multiple times, try to reuse last created gen_raw_REG.
14435 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
14436 main to workaround a bug in GDB.
14438 2018-01-12 Tom de Vries <tom@codesourcery.com>
14441 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
14443 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
14445 PR rtl-optimization/80481
14446 * ira-color.c (get_cap_member): New function.
14447 (allocnos_conflict_by_live_ranges_p): Use it.
14448 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
14449 (setup_slot_coalesced_allocno_live_ranges): Ditto.
14451 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
14454 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
14455 (*saddl_se_1): Ditto.
14456 (*ssubsi_1): Ditto.
14457 (*ssubl_se_1): Ditto.
14459 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14461 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
14462 rather than wi::to_widest for DR_INITs.
14463 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
14464 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
14465 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
14467 (vect_analyze_group_access_1): Note that here.
14469 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14471 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
14472 polynomial type sizes.
14474 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14476 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
14477 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
14478 (gimple_add_tmp_var): Likewise.
14480 2018-01-12 Martin Liska <mliska@suse.cz>
14482 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
14483 (gimple_alloc_sizes): Likewise.
14484 (dump_gimple_statistics): Use PRIu64 in printf format.
14485 * gimple.h: Change uint64_t to int.
14487 2018-01-12 Martin Liska <mliska@suse.cz>
14489 * tree-core.h: Use uint64_t instead of int.
14490 * tree.c (tree_node_counts): Likewise.
14491 (tree_node_sizes): Likewise.
14492 (dump_tree_statistics): Use PRIu64 in printf format.
14494 2018-01-12 Martin Liska <mliska@suse.cz>
14496 * Makefile.in: As qsort_chk is implemented in vec.c, add
14497 vec.o to linkage of gencfn-macros.
14498 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
14499 passing the info to record_node_allocation_statistics.
14500 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
14502 * ggc-common.c (struct ggc_usage): Add operator== and use
14503 it in operator< and compare function.
14504 * mem-stats.h (struct mem_usage): Likewise.
14505 * vec.c (struct vec_usage): Remove operator< and compare
14506 function. Can be simply inherited.
14508 2018-01-12 Martin Jambor <mjambor@suse.cz>
14511 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
14512 * tree-ssa-math-opts.c: Include domwalk.h.
14513 (convert_mult_to_fma_1): New function.
14514 (fma_transformation_info): New type.
14515 (fma_deferring_state): Likewise.
14516 (cancel_fma_deferring): New function.
14517 (result_of_phi): Likewise.
14518 (last_fma_candidate_feeds_initial_phi): Likewise.
14519 (convert_mult_to_fma): Added deferring logic, split actual
14520 transformation to convert_mult_to_fma_1.
14521 (math_opts_dom_walker): New type.
14522 (math_opts_dom_walker::after_dom_children): New method, body moved
14523 here from pass_optimize_widening_mul::execute, added deferring logic
14525 (pass_optimize_widening_mul::execute): Moved most of code to
14526 math_opts_dom_walker::after_dom_children.
14527 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
14528 * config/i386/i386.c (ix86_option_override_internal): Added
14529 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
14531 2018-01-12 Richard Biener <rguenther@suse.de>
14534 * dwarf2out.c (gen_variable_die): Do not reset old_die for
14535 inline instance vars.
14537 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
14540 * config/rx/rx.c (rx_is_restricted_memory_address):
14541 Handle SUBREG case.
14543 2018-01-12 Richard Biener <rguenther@suse.de>
14545 PR tree-optimization/80846
14546 * target.def (split_reduction): New target hook.
14547 * targhooks.c (default_split_reduction): New function.
14548 * targhooks.h (default_split_reduction): Declare.
14549 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
14550 target requests first reduce vectors by combining low and high
14552 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
14553 (get_vectype_for_scalar_type_and_size): Export.
14554 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
14555 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
14556 * doc/tm.texi: Regenerate.
14557 * config/i386/i386.c (ix86_split_reduction): Implement
14558 TARGET_VECTORIZE_SPLIT_REDUCTION.
14560 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14563 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
14564 in PIC mode except for TARGET_VXWORKS_RTP.
14565 * config/sparc/sparc.c: Include cfgrtl.h.
14566 (TARGET_INIT_PIC_REG): Define.
14567 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
14568 (sparc_pic_register_p): New predicate.
14569 (sparc_legitimate_address_p): Use it.
14570 (sparc_legitimize_pic_address): Likewise.
14571 (sparc_delegitimize_address): Likewise.
14572 (sparc_mode_dependent_address_p): Likewise.
14573 (gen_load_pcrel_sym): Remove 4th parameter.
14574 (load_got_register): Adjust call to above. Remove obsolete stuff.
14575 (sparc_expand_prologue): Do not call load_got_register here.
14576 (sparc_flat_expand_prologue): Likewise.
14577 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
14578 (sparc_use_pseudo_pic_reg): New function.
14579 (sparc_init_pic_reg): Likewise.
14580 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14581 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
14583 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
14585 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
14586 Add item for branch_cost.
14588 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14590 PR rtl-optimization/83565
14591 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
14592 not extend the result to a larger mode for rotate operations.
14593 (num_sign_bit_copies1): Likewise.
14595 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14598 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
14600 Use values-Xc.o for -pedantic.
14601 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
14603 2018-01-12 Martin Liska <mliska@suse.cz>
14606 * ipa-devirt.c (final_warning_record::grow_type_warnings):
14608 (possible_polymorphic_call_targets): Use it.
14609 (ipa_devirt): Likewise.
14611 2018-01-12 Martin Liska <mliska@suse.cz>
14613 * profile-count.h (enum profile_quality): Use 0 as invalid
14614 enum value of profile_quality.
14616 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
14618 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
14619 -mext-string options.
14621 2018-01-12 Richard Biener <rguenther@suse.de>
14623 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
14624 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
14625 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
14627 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
14629 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
14631 * configure.ac (--with-long-double-format): Add support for the
14632 configuration option to change the default long double format on
14634 * config.gcc (powerpc*-linux*-*): Likewise.
14635 * configure: Regenerate.
14636 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
14637 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
14638 used without modification.
14640 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14642 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
14643 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
14644 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
14645 MISC_BUILTIN_SPEC_BARRIER.
14646 (rs6000_init_builtins): Likewise.
14647 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
14649 (speculation_barrier): New define_insn.
14650 * doc/extend.texi: Document __builtin_speculation_barrier.
14652 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14655 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
14656 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
14657 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
14659 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
14660 integral modes instead of "ss" and "sd".
14661 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
14662 vectors with 32-bit and 64-bit elements.
14663 (vecdupssescalarmodesuffix): New mode attribute.
14664 (vec_dup<mode>): Use it.
14666 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
14669 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
14670 frame if argument is passed on stack.
14672 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14675 * ree.c (combine_reaching_defs): Optimize also
14676 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
14677 reg2=any_extend(exp); reg1=reg2;, formatting fix.
14679 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14681 PR middle-end/83189
14682 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
14684 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14686 PR middle-end/83718
14687 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
14688 after they are computed.
14690 2018-01-11 Bin Cheng <bin.cheng@arm.com>
14692 PR tree-optimization/83695
14693 * gimple-loop-linterchange.cc
14694 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
14695 reset cached scev information after interchange.
14696 (pass_linterchange::execute): Remove call to scev_reset_htab.
14698 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14700 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
14701 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
14702 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
14703 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
14704 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
14705 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
14706 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
14707 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
14708 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
14709 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
14710 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
14711 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
14712 (V_lane_reg): Likewise.
14713 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
14715 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
14716 (vfmal_lane_low<mode>_intrinsic,
14717 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
14718 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
14719 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
14720 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
14721 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
14722 vfmsl_lane_high<mode>_intrinsic): New define_insns.
14724 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14726 * config/arm/arm-cpus.in (fp16fml): New feature.
14727 (ALL_SIMD): Add fp16fml.
14728 (armv8.2-a): Add fp16fml as an option.
14729 (armv8.3-a): Likewise.
14730 (armv8.4-a): Add fp16fml as part of fp16.
14731 * config/arm/arm.h (TARGET_FP16FML): Define.
14732 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
14734 * config/arm/arm-modes.def (V2HF): Define.
14735 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
14736 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
14737 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
14738 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
14739 vfmsl_low, vfmsl_high): New set of builtins.
14740 * config/arm/iterators.md (PLUSMINUS): New code iterator.
14741 (vfml_op): New code attribute.
14742 (VFMLHALVES): New int iterator.
14743 (VFML, VFMLSEL): New mode attributes.
14744 (V_reg): Define mapping for V2HF.
14745 (V_hi, V_lo): New mode attributes.
14746 (VF_constraint): Likewise.
14747 (vfml_half, vfml_half_selector): New int attributes.
14748 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
14750 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
14751 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
14753 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
14754 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
14755 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
14756 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
14758 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
14759 Document new effective target and option set.
14761 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14763 * config/arm/arm-cpus.in (armv8_4): New feature.
14764 (ARMv8_4a): New fgroup.
14765 (armv8.4-a): New arch.
14766 * config/arm/arm-tables.opt: Regenerate.
14767 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
14768 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
14769 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
14770 Add matching rules for -march=armv8.4-a and extensions.
14771 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
14773 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
14776 * config/rx/rx.md (BW): New mode attribute.
14777 (sync_lock_test_and_setsi): Add mode suffix to insn output.
14779 2018-01-11 Richard Biener <rguenther@suse.de>
14781 PR tree-optimization/83435
14782 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
14783 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
14784 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
14786 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14787 Alan Hayward <alan.hayward@arm.com>
14788 David Sherwood <david.sherwood@arm.com>
14790 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
14792 (aarch64_classify_address): Initialize it. Track polynomial offsets.
14793 (aarch64_print_address_internal): Use it to check for a zero offset.
14795 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14796 Alan Hayward <alan.hayward@arm.com>
14797 David Sherwood <david.sherwood@arm.com>
14799 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
14800 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
14801 Return a poly_int64 rather than a HOST_WIDE_INT.
14802 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
14803 rather than a HOST_WIDE_INT.
14804 * config/aarch64/aarch64.h (aarch64_frame): Protect with
14805 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
14806 hard_fp_offset, frame_size, initial_adjust, callee_offset and
14807 final_offset from HOST_WIDE_INT to poly_int64.
14808 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
14809 to_constant when getting the number of units in an Advanced SIMD
14811 (aarch64_builtin_vectorized_function): Check for a constant number
14813 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
14815 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
14816 attribute instead of GET_MODE_NUNITS.
14817 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
14818 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
14819 GET_MODE_SIZE for fixed-size registers.
14820 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
14821 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
14822 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
14823 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
14824 (aarch64_print_operand, aarch64_print_address_internal)
14825 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
14826 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
14827 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
14828 Handle polynomial GET_MODE_SIZE.
14829 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
14830 wider than SImode without modification.
14831 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
14832 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
14833 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
14834 passing and returning SVE modes.
14835 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
14836 rather than GEN_INT.
14837 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
14838 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
14839 (aarch64_allocate_and_probe_stack_space): Likewise.
14840 (aarch64_layout_frame): Cope with polynomial offsets.
14841 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
14842 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
14843 polynomial offsets.
14844 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
14845 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
14846 poly_int64 rather than a HOST_WIDE_INT.
14847 (aarch64_get_separate_components, aarch64_process_components)
14848 (aarch64_expand_prologue, aarch64_expand_epilogue)
14849 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
14850 (aarch64_anchor_offset): New function, split out from...
14851 (aarch64_legitimize_address): ...here.
14852 (aarch64_builtin_vectorization_cost): Handle polynomial
14853 TYPE_VECTOR_SUBPARTS.
14854 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
14856 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
14857 number of elements from the PARALLEL rather than the mode.
14858 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
14859 rather than GET_MODE_BITSIZE.
14860 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
14861 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
14862 (aarch64_expand_vec_perm_const_1): Handle polynomial
14863 d->perm.length () and d->perm elements.
14864 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
14865 Apply to_constant to d->perm elements.
14866 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
14867 polynomial CONST_VECTOR_NUNITS.
14868 (aarch64_move_pointer): Take amount as a poly_int64 rather
14870 (aarch64_progress_pointer): Avoid temporary variable.
14871 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
14872 the mode attribute instead of GET_MODE.
14874 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14875 Alan Hayward <alan.hayward@arm.com>
14876 David Sherwood <david.sherwood@arm.com>
14878 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
14879 x exists before using it.
14880 (aarch64_add_constant_internal): Rename to...
14881 (aarch64_add_offset_1): ...this. Replace regnum with separate
14882 src and dest rtxes. Handle the case in which they're different,
14883 including when the offset is zero. Replace scratchreg with an rtx.
14884 Use 2 additions if there is no spare register into which we can
14885 move a 16-bit constant.
14886 (aarch64_add_constant): Delete.
14887 (aarch64_add_offset): Replace reg with separate src and dest
14888 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
14889 Use aarch64_add_offset_1.
14890 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
14891 an rtx rather than an int. Take the delta as a poly_int64
14892 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
14893 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
14894 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
14895 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
14896 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
14897 and aarch64_add_sp.
14898 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
14899 aarch64_add_constant.
14901 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14903 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
14904 Use scalar_float_mode.
14906 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14908 * config/aarch64/aarch64-simd.md
14909 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
14910 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
14911 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
14912 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
14913 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
14914 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
14915 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
14916 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
14917 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
14918 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
14920 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14923 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
14924 targ_options->x_arm_arch_string is non NULL.
14926 2018-01-11 Tamar Christina <tamar.christina@arm.com>
14928 * config/aarch64/aarch64.h
14929 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
14931 2018-01-11 Sudakshina Das <sudi.das@arm.com>
14934 * expmed.c (emit_store_flag_force): Swap if const op0
14935 and change VOIDmode to mode of op0.
14937 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14939 PR rtl-optimization/83761
14940 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
14941 than bytes to mode_for_size.
14943 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14945 PR middle-end/83189
14946 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
14947 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
14950 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14952 PR middle-end/83575
14953 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
14954 when in layout mode.
14955 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
14956 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
14959 2018-01-10 Michael Collison <michael.collison@arm.com>
14961 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
14962 * config/aarch64/aarch64-option-extension.def: Add
14963 AARCH64_OPT_EXTENSION of 'fp16fml'.
14964 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14965 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
14966 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
14967 * config/aarch64/constraints.md (Ui7): New constraint.
14968 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
14969 (VFMLA_SEL_W): Ditto.
14972 (VFMLA16_LOW): New int iterator.
14973 (VFMLA16_HIGH): Ditto.
14974 (UNSPEC_FMLAL): New unspec.
14975 (UNSPEC_FMLSL): Ditto.
14976 (UNSPEC_FMLAL2): Ditto.
14977 (UNSPEC_FMLSL2): Ditto.
14978 (f16mac): New code attribute.
14979 * config/aarch64/aarch64-simd-builtins.def
14980 (aarch64_fmlal_lowv2sf): Ditto.
14981 (aarch64_fmlsl_lowv2sf): Ditto.
14982 (aarch64_fmlalq_lowv4sf): Ditto.
14983 (aarch64_fmlslq_lowv4sf): Ditto.
14984 (aarch64_fmlal_highv2sf): Ditto.
14985 (aarch64_fmlsl_highv2sf): Ditto.
14986 (aarch64_fmlalq_highv4sf): Ditto.
14987 (aarch64_fmlslq_highv4sf): Ditto.
14988 (aarch64_fmlal_lane_lowv2sf): Ditto.
14989 (aarch64_fmlsl_lane_lowv2sf): Ditto.
14990 (aarch64_fmlal_laneq_lowv2sf): Ditto.
14991 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
14992 (aarch64_fmlalq_lane_lowv4sf): Ditto.
14993 (aarch64_fmlsl_lane_lowv4sf): Ditto.
14994 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
14995 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
14996 (aarch64_fmlal_lane_highv2sf): Ditto.
14997 (aarch64_fmlsl_lane_highv2sf): Ditto.
14998 (aarch64_fmlal_laneq_highv2sf): Ditto.
14999 (aarch64_fmlsl_laneq_highv2sf): Ditto.
15000 (aarch64_fmlalq_lane_highv4sf): Ditto.
15001 (aarch64_fmlsl_lane_highv4sf): Ditto.
15002 (aarch64_fmlalq_laneq_highv4sf): Ditto.
15003 (aarch64_fmlsl_laneq_highv4sf): Ditto.
15004 * config/aarch64/aarch64-simd.md:
15005 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
15006 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15007 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
15008 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15009 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
15010 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
15011 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
15012 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
15013 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
15014 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
15015 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
15016 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
15017 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
15018 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
15019 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
15020 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
15021 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
15022 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
15023 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
15024 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
15025 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
15026 (vfmlsl_low_u32): Ditto.
15027 (vfmlalq_low_u32): Ditto.
15028 (vfmlslq_low_u32): Ditto.
15029 (vfmlal_high_u32): Ditto.
15030 (vfmlsl_high_u32): Ditto.
15031 (vfmlalq_high_u32): Ditto.
15032 (vfmlslq_high_u32): Ditto.
15033 (vfmlal_lane_low_u32): Ditto.
15034 (vfmlsl_lane_low_u32): Ditto.
15035 (vfmlal_laneq_low_u32): Ditto.
15036 (vfmlsl_laneq_low_u32): Ditto.
15037 (vfmlalq_lane_low_u32): Ditto.
15038 (vfmlslq_lane_low_u32): Ditto.
15039 (vfmlalq_laneq_low_u32): Ditto.
15040 (vfmlslq_laneq_low_u32): Ditto.
15041 (vfmlal_lane_high_u32): Ditto.
15042 (vfmlsl_lane_high_u32): Ditto.
15043 (vfmlal_laneq_high_u32): Ditto.
15044 (vfmlsl_laneq_high_u32): Ditto.
15045 (vfmlalq_lane_high_u32): Ditto.
15046 (vfmlslq_lane_high_u32): Ditto.
15047 (vfmlalq_laneq_high_u32): Ditto.
15048 (vfmlslq_laneq_high_u32): Ditto.
15049 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
15050 (AARCH64_FL_FOR_ARCH8_4): New.
15051 (AARCH64_ISA_F16FML): New ISA flag.
15052 (TARGET_F16FML): New feature flag for fp16fml.
15053 (doc/invoke.texi): Document new fp16fml option.
15055 2018-01-10 Michael Collison <michael.collison@arm.com>
15057 * config/aarch64/aarch64-builtins.c:
15058 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
15059 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15060 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
15061 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
15062 (AARCH64_ISA_SHA3): New ISA flag.
15063 (TARGET_SHA3): New feature flag for sha3.
15064 * config/aarch64/iterators.md (sha512_op): New int attribute.
15065 (CRYPTO_SHA512): New int iterator.
15066 (UNSPEC_SHA512H): New unspec.
15067 (UNSPEC_SHA512H2): Ditto.
15068 (UNSPEC_SHA512SU0): Ditto.
15069 (UNSPEC_SHA512SU1): Ditto.
15070 * config/aarch64/aarch64-simd-builtins.def
15071 (aarch64_crypto_sha512hqv2di): New builtin.
15072 (aarch64_crypto_sha512h2qv2di): Ditto.
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/aarch64-simd.md:
15080 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
15081 (aarch64_crypto_sha512su0qv2di): Ditto.
15082 (aarch64_crypto_sha512su1qv2di): Ditto.
15083 (aarch64_eor3qv8hi): Ditto.
15084 (aarch64_rax1qv2di): Ditto.
15085 (aarch64_xarqv2di): Ditto.
15086 (aarch64_bcaxqv8hi): Ditto.
15087 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
15088 (vsha512h2q_u64): Ditto.
15089 (vsha512su0q_u64): Ditto.
15090 (vsha512su1q_u64): Ditto.
15091 (veor3q_u16): Ditto.
15092 (vrax1q_u64): Ditto.
15093 (vxarq_u64): Ditto.
15094 (vbcaxq_u16): Ditto.
15095 * config/arm/types.md (crypto_sha512): New type attribute.
15096 (crypto_sha3): Ditto.
15097 (doc/invoke.texi): Document new sha3 option.
15099 2018-01-10 Michael Collison <michael.collison@arm.com>
15101 * config/aarch64/aarch64-builtins.c:
15102 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
15103 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15104 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
15105 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
15106 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
15107 (AARCH64_ISA_SM4): New ISA flag.
15108 (TARGET_SM4): New feature flag for sm4.
15109 * config/aarch64/aarch64-simd-builtins.def
15110 (aarch64_sm3ss1qv4si): Ditto.
15111 (aarch64_sm3tt1aq4si): Ditto.
15112 (aarch64_sm3tt1bq4si): Ditto.
15113 (aarch64_sm3tt2aq4si): Ditto.
15114 (aarch64_sm3tt2bq4si): Ditto.
15115 (aarch64_sm3partw1qv4si): Ditto.
15116 (aarch64_sm3partw2qv4si): Ditto.
15117 (aarch64_sm4eqv4si): Ditto.
15118 (aarch64_sm4ekeyqv4si): Ditto.
15119 * config/aarch64/aarch64-simd.md:
15120 (aarch64_sm3ss1qv4si): Ditto.
15121 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
15122 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
15123 (aarch64_sm4eqv4si): Ditto.
15124 (aarch64_sm4ekeyqv4si): Ditto.
15125 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
15126 (sm3part_op): Ditto.
15127 (CRYPTO_SM3TT): Ditto.
15128 (CRYPTO_SM3PART): Ditto.
15129 (UNSPEC_SM3SS1): New unspec.
15130 (UNSPEC_SM3TT1A): Ditto.
15131 (UNSPEC_SM3TT1B): Ditto.
15132 (UNSPEC_SM3TT2A): Ditto.
15133 (UNSPEC_SM3TT2B): Ditto.
15134 (UNSPEC_SM3PARTW1): Ditto.
15135 (UNSPEC_SM3PARTW2): Ditto.
15136 (UNSPEC_SM4E): Ditto.
15137 (UNSPEC_SM4EKEY): Ditto.
15138 * config/aarch64/constraints.md (Ui2): New constraint.
15139 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
15140 * config/arm/types.md (crypto_sm3): New type attribute.
15141 (crypto_sm4): Ditto.
15142 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
15143 (vsm3tt1aq_u32): Ditto.
15144 (vsm3tt1bq_u32): Ditto.
15145 (vsm3tt2aq_u32): Ditto.
15146 (vsm3tt2bq_u32): Ditto.
15147 (vsm3partw1q_u32): Ditto.
15148 (vsm3partw2q_u32): Ditto.
15149 (vsm4eq_u32): Ditto.
15150 (vsm4ekeyq_u32): Ditto.
15151 (doc/invoke.texi): Document new sm4 option.
15153 2018-01-10 Michael Collison <michael.collison@arm.com>
15155 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
15156 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
15157 (AARCH64_FL_FOR_ARCH8_4): New.
15158 (AARCH64_FL_V8_4): New flag.
15159 (doc/invoke.texi): Document new armv8.4-a option.
15161 2018-01-10 Michael Collison <michael.collison@arm.com>
15163 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15164 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
15165 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
15166 * config/aarch64/aarch64-option-extension.def: Add
15167 AARCH64_OPT_EXTENSION of 'sha2'.
15168 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
15169 (crypto): Disable sha2 and aes if crypto disabled.
15170 (crypto): Enable aes and sha2 if enabled.
15171 (simd): Disable sha2 and aes if simd disabled.
15172 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
15174 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
15175 (TARGET_SHA2): New feature flag for sha2.
15176 (TARGET_AES): New feature flag for aes.
15177 * config/aarch64/aarch64-simd.md:
15178 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
15179 conditional on TARGET_AES.
15180 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
15181 (aarch64_crypto_sha1hsi): Make pattern conditional
15183 (aarch64_crypto_sha1hv4si): Ditto.
15184 (aarch64_be_crypto_sha1hv4si): Ditto.
15185 (aarch64_crypto_sha1su1v4si): Ditto.
15186 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
15187 (aarch64_crypto_sha1su0v4si): Ditto.
15188 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
15189 (aarch64_crypto_sha256su0v4si): Ditto.
15190 (aarch64_crypto_sha256su1v4si): Ditto.
15191 (doc/invoke.texi): Document new aes and sha2 options.
15193 2018-01-10 Martin Sebor <msebor@redhat.com>
15195 PR tree-optimization/83781
15196 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
15199 2018-01-11 Martin Sebor <msebor@gmail.com>
15200 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15202 PR tree-optimization/83501
15203 PR tree-optimization/81703
15205 * tree-ssa-strlen.c (get_string_cst): Rename...
15206 (get_string_len): ...to this. Handle global constants.
15207 (handle_char_store): Adjust.
15209 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
15210 Jim Wilson <jimw@sifive.com>
15212 * config/riscv/riscv-protos.h (riscv_output_return): New.
15213 * config/riscv/riscv.c (struct machine_function): New naked_p field.
15214 (riscv_attribute_table, riscv_output_return),
15215 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
15216 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
15217 (riscv_compute_frame_info): Only compute frame->mask if not a naked
15219 (riscv_expand_prologue): Add early return for naked function.
15220 (riscv_expand_epilogue): Likewise.
15221 (riscv_function_ok_for_sibcall): Return false for naked function.
15222 (riscv_set_current_function): New.
15223 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
15224 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
15225 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
15226 * doc/extend.texi (RISC-V Function Attributes): New.
15228 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
15230 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
15231 check for 128-bit long double before checking TCmode.
15232 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
15233 128-bit long doubles before checking TFmode or TCmode.
15234 (FLOAT128_IBM_P): Likewise.
15236 2018-01-10 Martin Sebor <msebor@redhat.com>
15238 PR tree-optimization/83671
15239 * builtins.c (c_strlen): Unconditionally return zero for the empty
15241 Use -Warray-bounds for warnings.
15242 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
15243 for non-constant array indices with COMPONENT_REF, arrays of
15244 arrays, and pointers to arrays.
15245 (gimple_fold_builtin_strlen): Determine and set length range for
15246 non-constant character arrays.
15248 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
15250 PR middle-end/81897
15251 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
15254 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
15256 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
15258 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15261 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
15262 VECTOR_MEM_ALTIVEC_OR_VSX_P.
15263 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
15264 indexed_or_indirect_operand predicate.
15265 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
15266 (*vsx_le_perm_load_v8hi): Likewise.
15267 (*vsx_le_perm_load_v16qi): Likewise.
15268 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
15269 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
15270 (*vsx_le_perm_store_v8hi): Likewise.
15271 (*vsx_le_perm_store_v16qi): Likewise.
15272 (eight unnamed splitters): Likewise.
15274 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15276 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
15277 * config/rs6000/emmintrin.h: Likewise.
15278 * config/rs6000/mmintrin.h: Likewise.
15279 * config/rs6000/xmmintrin.h: Likewise.
15281 2018-01-10 David Malcolm <dmalcolm@redhat.com>
15284 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
15286 * tree.c (tree_nop_conversion): Return true for location wrapper
15288 (maybe_wrap_with_location): New function.
15289 (selftest::check_strip_nops): New function.
15290 (selftest::test_location_wrappers): New function.
15291 (selftest::tree_c_tests): Call it.
15292 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
15293 (maybe_wrap_with_location): New decl.
15294 (EXPR_LOCATION_WRAPPER_P): New macro.
15295 (location_wrapper_p): New inline function.
15296 (tree_strip_any_location_wrapper): New inline function.
15298 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
15301 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
15302 stack_realign_offset for the largest alignment of stack slot
15304 (ix86_find_max_used_stack_alignment): New function.
15305 (ix86_finalize_stack_frame_flags): Use it. Set
15306 max_used_stack_alignment if we don't realign stack.
15307 * config/i386/i386.h (machine_function): Add
15308 max_used_stack_alignment.
15310 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
15312 * config/arm/arm.opt (-mbranch-cost): New option.
15313 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
15316 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
15319 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
15320 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
15322 2018-01-10 Richard Biener <rguenther@suse.de>
15325 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
15326 early out so it also covers the case where we have a non-NULL
15329 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15331 PR tree-optimization/83753
15332 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
15333 for non-strided grouped accesses if the number of elements is 1.
15335 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15338 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
15339 * i386.h (TARGET_USE_GATHER): Define.
15340 * x86-tune.def (X86_TUNE_USE_GATHER): New.
15342 2018-01-10 Martin Liska <mliska@suse.cz>
15345 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
15346 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
15348 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
15349 CLEANUP_NO_PARTITIONING is not set.
15351 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15353 * doc/rtl.texi: Remove documentation of (const ...) wrappers
15354 for vectors, as a partial revert of r254296.
15355 * rtl.h (const_vec_p): Delete.
15356 (const_vec_duplicate_p): Don't test for vector CONSTs.
15357 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
15358 * expmed.c (make_tree): Likewise.
15361 * common.md (E, F): Use CONSTANT_P instead of checking for
15363 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
15364 checking for CONST_VECTOR.
15366 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15368 PR middle-end/83575
15369 * predict.c (force_edge_cold): Handle in more sane way edges
15370 with no prediction.
15372 2018-01-09 Carl Love <cel@us.ibm.com>
15374 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
15376 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
15377 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
15378 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
15379 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
15380 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
15381 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
15382 * config/rs6000/rs6000-protos.h: Add extern defition for
15383 rs6000_generate_float2_double_code.
15384 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
15386 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
15387 (float2_v2df): Add define_expand.
15389 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
15392 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
15393 op_mode in the force_to_mode call.
15395 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15397 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
15398 instead of checking each element individually.
15399 (aarch64_evpc_uzp): Likewise.
15400 (aarch64_evpc_zip): Likewise.
15401 (aarch64_evpc_ext): Likewise.
15402 (aarch64_evpc_rev): Likewise.
15403 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
15404 instead of checking each element individually. Return true without
15406 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
15407 whether all selected elements come from the same input, instead of
15408 checking each element individually. Remove calls to gen_rtx_REG,
15409 start_sequence and end_sequence and instead assert that no rtl is
15412 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15414 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
15415 order of HIGH and CONST checks.
15417 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15419 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
15420 if the destination isn't an SSA_NAME.
15422 2018-01-09 Richard Biener <rguenther@suse.de>
15424 PR tree-optimization/83668
15425 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
15427 (canonicalize_loop_form): ... here, renamed from ...
15428 (canonicalize_loop_closed_ssa_form): ... this and amended to
15429 swap successor edges for loop exit blocks to make us use
15430 the RPO order we need for initial schedule generation.
15432 2018-01-09 Joseph Myers <joseph@codesourcery.com>
15434 PR tree-optimization/64811
15435 * match.pd: When optimizing comparisons with Inf, avoid
15436 introducing or losing exceptions from comparisons with NaN.
15438 2018-01-09 Martin Liska <mliska@suse.cz>
15441 * asan.c (shadow_mem_size): Add gcc_assert.
15443 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
15445 Don't save registers in main().
15448 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
15449 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
15450 * config/avr/avr.c (avr_set_current_function): Don't error if
15451 naked, OS_task or OS_main are specified at the same time.
15452 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
15454 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
15456 * common/config/avr/avr-common.c (avr_option_optimization_table):
15457 Switch on -mmain-is-OS_task for optimizing compilations.
15459 2018-01-09 Richard Biener <rguenther@suse.de>
15461 PR tree-optimization/83572
15462 * graphite.c: Include cfganal.h.
15463 (graphite_transform_loops): Connect infinite loops to exit
15464 and remove fake edges at the end.
15466 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15468 * ipa-inline.c (edge_badness): Revert accidental checkin.
15470 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15473 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
15474 symbols; not inline clones.
15476 2018-01-09 Jakub Jelinek <jakub@redhat.com>
15479 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
15480 hard registers. Formatting fixes.
15482 PR preprocessor/83722
15483 * gcc.c (try_generate_repro): Pass
15484 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
15485 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
15488 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
15489 Kito Cheng <kito.cheng@gmail.com>
15491 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
15492 (riscv_leaf_function_p): Delete.
15493 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
15495 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15497 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
15499 (do_ifelse): New function.
15500 (do_isel): New function.
15501 (do_sub3): New function.
15502 (do_add3): New function.
15503 (do_load_mask_compare): New function.
15504 (do_overlap_load_compare): New function.
15505 (expand_compare_loop): New function.
15506 (expand_block_compare): Call expand_compare_loop() when appropriate.
15507 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
15508 option description.
15509 (-mblock-compare-inline-loop-limit): New option.
15511 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15514 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
15515 Reverse order of second and third operands in first alternative.
15516 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
15517 of first and second elements in UNSPEC_VPERMR vector.
15518 (altivec_expand_vec_perm_le): Likewise.
15520 2018-01-08 Jeff Law <law@redhat.com>
15522 PR rtl-optimizatin/81308
15523 * tree-switch-conversion.c (cfg_altered): New file scoped static.
15524 (process_switch): If group_case_labels makes a change, then set
15526 (pass_convert_switch::execute): If a switch is converted, then
15527 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
15529 PR rtl-optimization/81308
15530 * recog.c (split_all_insns): Conditionally cleanup the CFG after
15533 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
15535 PR target/83663 - Revert r255946
15536 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
15537 generation for cases where splatting a value is not useful.
15538 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
15539 across a vec_duplicate and a paradoxical subreg forming a vector
15540 mode to a vec_concat.
15542 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15544 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
15545 -march=armv8.3-a variants.
15546 * config/arm/t-multilib: Likewise.
15547 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
15549 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15551 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
15553 (cceq_ior_compare_complement): Give it a name so I can use it, and
15554 change boolean_or_operator predicate to boolean_operator so it can
15555 be used to generate a crand.
15556 (eqne): New code iterator.
15557 (bd/bd_neg): New code_attrs.
15558 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
15559 a single define_insn.
15560 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
15561 decrement (bdnzt/bdnzf/bdzt/bdzf).
15562 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
15563 with the new names of the branch decrement patterns, and added the
15564 names of the branch decrement conditional patterns.
15566 2018-01-08 Richard Biener <rguenther@suse.de>
15568 PR tree-optimization/83563
15569 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
15572 2018-01-08 Richard Biener <rguenther@suse.de>
15574 PR middle-end/83713
15575 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
15577 2018-01-08 Richard Biener <rguenther@suse.de>
15579 PR tree-optimization/83685
15580 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
15581 references to abnormals.
15583 2018-01-08 Richard Biener <rguenther@suse.de>
15586 * dwarf2out.c (output_indirect_strings): Handle empty
15587 skeleton_debug_str_hash.
15588 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
15590 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15592 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
15593 (emit_store_direct): Likewise.
15594 (arc_trampoline_adjust_address): Likewise.
15595 (arc_asm_trampoline_template): New function.
15596 (arc_initialize_trampoline): Use asm_trampoline_template.
15597 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
15598 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
15599 * config/arc/arc.md (flush_icache): Delete pattern.
15601 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15603 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
15604 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
15607 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15610 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
15611 by not USED_FOR_TARGET.
15612 (make_pass_resolve_sw_modes): Likewise.
15614 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15616 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
15619 2018-01-08 Richard Biener <rguenther@suse.de>
15621 PR middle-end/83580
15622 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
15624 2018-01-08 Richard Biener <rguenther@suse.de>
15626 PR middle-end/83517
15627 * match.pd ((t * 2) / 2) -> t): Add missing :c.
15629 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
15631 PR middle-end/81897
15632 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
15633 basic blocks with a small number of successors.
15634 (convert_control_dep_chain_into_preds): Improve handling of
15636 (dump_predicates): Split apart into...
15637 (dump_pred_chain): ...here...
15638 (dump_pred_info): ...and here.
15639 (can_one_predicate_be_invalidated_p): Add debugging printfs.
15640 (can_chain_union_be_invalidated_p): Improve check for invalidation
15642 (uninit_uses_cannot_happen): Avoid unnecessary if
15643 convert_control_dep_chain_into_preds yielded nothing.
15645 2018-01-06 Martin Sebor <msebor@redhat.com>
15647 PR tree-optimization/83640
15648 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
15649 subtracting negative offset from size.
15650 (builtin_access::overlap): Adjust offset bounds of the access to fall
15651 within the size of the object if possible.
15653 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
15655 PR rtl-optimization/83699
15656 * expmed.c (extract_bit_field_1): Restrict the vector usage of
15657 extract_bit_field_as_subreg to cases in which the extracted
15658 value is also a vector.
15660 * lra-constraints.c (process_alt_operands): Test for the equivalence
15661 substitutions when detecting a possible reload cycle.
15663 2018-01-06 Jakub Jelinek <jakub@redhat.com>
15666 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
15667 by default if flag_selective_schedling{,2}. Formatting fixes.
15669 PR rtl-optimization/83682
15670 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
15671 if it has non-VECTOR_MODE element mode.
15672 (vec_duplicate_p): Likewise.
15674 PR middle-end/83694
15675 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
15676 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
15678 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15681 * config/i386/i386-builtin.def
15682 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
15683 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
15684 Require also OPTION_MASK_ISA_AVX512F in addition to
15685 OPTION_MASK_ISA_GFNI.
15686 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
15687 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
15688 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
15689 to OPTION_MASK_ISA_GFNI.
15690 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
15691 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
15692 OPTION_MASK_ISA_AVX512BW.
15693 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
15694 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
15695 addition to OPTION_MASK_ISA_GFNI.
15696 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
15697 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
15698 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
15699 to OPTION_MASK_ISA_GFNI.
15700 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
15701 a requirement for all ISAs rather than any of them with a few
15703 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
15705 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
15706 bitmasks to be enabled with 3 exceptions, instead of requiring any
15707 enabled ISA with lots of exceptions.
15708 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
15709 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
15710 Change avx512bw in isa attribute to avx512f.
15711 * config/i386/sgxintrin.h: Add license boilerplate.
15712 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
15713 to __AVX512F__ and __AVX512VL to __AVX512VL__.
15714 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
15715 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
15717 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
15718 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
15719 temporarily sse2 rather than sse if not enabled already.
15722 * config/i386/sse.md (VI248_VLBW): Rename to ...
15723 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
15724 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
15725 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
15726 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
15727 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
15728 mode iterator instead of VI248_VLBW.
15730 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
15732 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
15733 (record_modified): Skip clobbers; add debug output.
15734 (param_change_prob): Use sreal frequencies.
15736 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15738 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
15739 punt for user-aligned variables.
15741 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15743 * tree-chrec.c (chrec_contains_symbols): Return true for
15746 2018-01-05 Sudakshina Das <sudi.das@arm.com>
15749 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
15750 of (x|y) == x for BICS pattern.
15752 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15754 PR tree-optimization/83605
15755 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
15756 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
15759 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
15761 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
15762 * config/epiphany/rtems.h: New file.
15764 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15765 Uros Bizjak <ubizjak@gmail.com>
15768 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
15769 QIreg_operand instead of register_operand predicate.
15770 * config/i386/i386.c (ix86_rop_should_change_byte_p,
15771 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
15772 comments instead of -fmitigate[-_]rop.
15774 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15777 * cgraphunit.c (symbol_table::compile): Switch to text_section
15778 before calling assembly_start debug hook.
15779 * run-rtl-passes.c (run_rtl_passes): Likewise.
15782 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15784 * tree-vrp.c (extract_range_from_binary_expr_1): Check
15785 range_int_cst_p rather than !symbolic_range_p before calling
15786 extract_range_from_multiplicative_op_1.
15788 2018-01-04 Jeff Law <law@redhat.com>
15790 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
15791 redundant test in assertion.
15793 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15795 * doc/rtl.texi: Document machine_mode wrapper classes.
15797 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15799 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
15800 using tree_to_uhwi.
15802 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15804 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
15805 the VEC_PERM_EXPR fold to fail.
15807 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15810 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
15811 to switched_sections.
15813 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15816 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
15817 test for d.testing.
15819 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
15822 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
15823 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
15825 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15828 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
15829 is BLKmode and bitpos not zero or mode change is needed.
15831 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15834 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
15837 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
15840 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
15841 instead of MULT rtx. Update all corresponding splitters.
15842 (*saddl_se): Ditto.
15843 (*ssub<modesuffix>): Ditto.
15844 (*ssubl_se): Ditto.
15845 (*cmp_sadd_di): Update split patterns.
15846 (*cmp_sadd_si): Ditto.
15847 (*cmp_sadd_sidi): Ditto.
15848 (*cmp_ssub_di): Ditto.
15849 (*cmp_ssub_si): Ditto.
15850 (*cmp_ssub_sidi): Ditto.
15851 * config/alpha/predicates.md (const23_operand): New predicate.
15852 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
15853 Look for ASHIFT, not MULT inner operand.
15854 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
15856 2018-01-04 Martin Liska <mliska@suse.cz>
15858 PR gcov-profile/83669
15859 * gcov.c (output_intermediate_file): Add version to intermediate
15861 * doc/gcov.texi: Document new field 'version' in intermediate
15862 file format. Fix location of '-k' option of gcov command.
15864 2018-01-04 Martin Liska <mliska@suse.cz>
15867 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
15869 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15871 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
15873 2018-01-03 Martin Sebor <msebor@redhat.com>
15875 PR tree-optimization/83655
15876 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
15877 checking calls with invalid arguments.
15879 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15881 * tree-vect-stmts.c (vect_get_store_rhs): New function.
15882 (vectorizable_mask_load_store): Delete.
15883 (vectorizable_call): Return false for masked loads and stores.
15884 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
15885 instead of gimple_assign_rhs1.
15886 (vectorizable_load): Handle IFN_MASK_LOAD.
15887 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
15889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15891 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
15893 (vectorizable_mask_load_store): ...here.
15894 (vectorizable_load): ...and here.
15896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15898 * tree-vect-stmts.c (vect_build_all_ones_mask)
15899 (vect_build_zero_merge_argument): New functions, split out from...
15900 (vectorizable_load): ...here.
15902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15904 * tree-vect-stmts.c (vect_check_store_rhs): New function,
15906 (vectorizable_mask_load_store): ...here.
15907 (vectorizable_store): ...and here.
15909 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15911 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
15913 (vectorizable_mask_load_store): ...here.
15915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15917 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
15918 (vect_model_store_cost): Take a vec_load_store_type instead of a
15920 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
15921 (vect_model_store_cost): Take a vec_load_store_type instead of a
15923 (vectorizable_mask_load_store): Update accordingly.
15924 (vectorizable_store): Likewise.
15925 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
15927 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15929 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
15930 IFN_MASK_LOAD calls here rather than...
15931 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
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 * expmed.c (extract_bit_field_1): For vector extracts,
15938 fall back to extract_bit_field_as_subreg if vec_extract
15941 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15942 Alan Hayward <alan.hayward@arm.com>
15943 David Sherwood <david.sherwood@arm.com>
15945 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
15946 they are variable or constant sized.
15947 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
15948 slots for constant-sized data.
15950 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15951 Alan Hayward <alan.hayward@arm.com>
15952 David Sherwood <david.sherwood@arm.com>
15954 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
15955 handling COND_EXPRs with boolean comparisons, try to find a better
15956 basis for the mask type than the boolean itself.
15958 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15960 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
15961 is calculated and how it can be overridden.
15962 * genmodes.c (max_bitsize_mode_any_mode): New variable.
15963 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
15965 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
15968 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15969 Alan Hayward <alan.hayward@arm.com>
15970 David Sherwood <david.sherwood@arm.com>
15972 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
15973 Remove the mode argument.
15974 (aarch64_simd_valid_immediate): Remove the mode and inverse
15976 * config/aarch64/iterators.md (bitsize): New iterator.
15977 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
15978 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
15979 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
15980 aarch64_simd_valid_immediate.
15981 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
15982 (aarch64_reg_or_bic_imm): Likewise.
15983 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
15984 with an insn_type enum and msl with a modifier_type enum.
15985 Replace element_width with a scalar_mode. Change the shift
15986 to unsigned int. Add constructors for scalar_float_mode and
15987 scalar_int_mode elements.
15988 (aarch64_vect_float_const_representable_p): Delete.
15989 (aarch64_can_const_movi_rtx_p)
15990 (aarch64_simd_scalar_immediate_valid_for_move)
15991 (aarch64_simd_make_constant): Update call to
15992 aarch64_simd_valid_immediate.
15993 (aarch64_advsimd_valid_immediate_hs): New function.
15994 (aarch64_advsimd_valid_immediate): Likewise.
15995 (aarch64_simd_valid_immediate): Remove mode and inverse
15996 arguments. Rewrite to use the above. Use const_vec_duplicate_p
15997 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
15998 and aarch64_float_const_representable_p on the result.
15999 (aarch64_output_simd_mov_immediate): Remove mode argument.
16000 Update call to aarch64_simd_valid_immediate and use of
16001 simd_immediate_info.
16002 (aarch64_output_scalar_simd_mov_immediate): Update call
16005 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16006 Alan Hayward <alan.hayward@arm.com>
16007 David Sherwood <david.sherwood@arm.com>
16009 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
16010 (mode_nunits): Likewise CONST_MODE_NUNITS.
16011 * machmode.def (ADJUST_NUNITS): Document.
16012 * genmodes.c (mode_data::need_nunits_adj): New field.
16013 (blank_mode): Update accordingly.
16014 (adj_nunits): New variable.
16015 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
16017 (emit_mode_size_inline): Set need_bytesize_adj for all modes
16018 listed in adj_nunits.
16019 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
16020 listed in adj_nunits. Don't emit case statements for such modes.
16021 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
16022 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
16023 nothing if adj_nunits is nonnull.
16024 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
16025 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
16026 (emit_mode_fbit): Update use of print_maybe_const_decl.
16027 (emit_move_size): Likewise. Treat the array as non-const
16029 (emit_mode_adjustments): Handle adj_nunits.
16031 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16033 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
16034 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
16035 (VECTOR_MODES): Use it.
16036 (make_vector_modes): Take the prefix as an argument.
16038 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16039 Alan Hayward <alan.hayward@arm.com>
16040 David Sherwood <david.sherwood@arm.com>
16042 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
16043 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
16044 for MODE_VECTOR_BOOL.
16045 * machmode.def (VECTOR_BOOL_MODE): Document.
16046 * genmodes.c (VECTOR_BOOL_MODE): New macro.
16047 (make_vector_bool_mode): New function.
16048 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
16050 * lto-streamer-in.c (lto_input_mode_table): Likewise.
16051 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
16053 * stor-layout.c (int_mode_for_mode): Likewise.
16054 * tree.c (build_vector_type_for_mode): Likewise.
16055 * varasm.c (output_constant_pool_2): Likewise.
16056 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
16057 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
16058 for MODE_VECTOR_BOOL.
16059 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
16060 of mode class checks.
16061 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
16062 instead of a list of mode class checks.
16063 (expand_vector_scalar_condition): Likewise.
16064 (type_for_widest_vector_mode): Handle BImode as an inner mode.
16066 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16067 Alan Hayward <alan.hayward@arm.com>
16068 David Sherwood <david.sherwood@arm.com>
16070 * machmode.h (mode_size): Change from unsigned short to
16072 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
16073 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16074 or if measurement_type is not polynomial.
16075 (fixed_size_mode::includes_p): Check for constant-sized modes.
16076 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
16077 return a poly_uint16 rather than an unsigned short.
16078 (emit_mode_size): Change the type of mode_size from unsigned short
16079 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
16080 (emit_mode_adjustments): Cope with polynomial vector sizes.
16081 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16083 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16085 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
16086 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
16087 * caller-save.c (setup_save_areas): Likewise.
16088 (replace_reg_with_saved_mem): Likewise.
16089 * calls.c (emit_library_call_value_1): Likewise.
16090 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
16091 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
16092 (gen_lowpart_for_combine): Likewise.
16093 * convert.c (convert_to_integer_1): Likewise.
16094 * cse.c (equiv_constant, cse_insn): Likewise.
16095 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
16096 (cselib_subst_to_values): Likewise.
16097 * dce.c (word_dce_process_block): Likewise.
16098 * df-problems.c (df_word_lr_mark_ref): Likewise.
16099 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
16100 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
16101 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
16102 (rtl_for_decl_location): Likewise.
16103 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
16104 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
16105 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
16106 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
16107 (expand_expr_real_1): Likewise.
16108 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
16109 (pad_below): Likewise.
16110 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16111 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
16112 * ira.c (get_subreg_tracking_sizes): Likewise.
16113 * ira-build.c (ira_create_allocno_objects): Likewise.
16114 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
16115 (ira_sort_regnos_for_alter_reg): Likewise.
16116 * ira-costs.c (record_operand_costs): Likewise.
16117 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
16118 (resolve_simple_move): Likewise.
16119 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
16120 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
16121 (lra_constraints): Likewise.
16122 (CONST_POOL_OK_P): Reject variable-sized modes.
16123 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
16124 (add_pseudo_to_slot, lra_spill): Likewise.
16125 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16126 * optabs-query.c (get_best_extraction_insn): Likewise.
16127 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16128 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
16129 (expand_mult_highpart, valid_multiword_target_p): Likewise.
16130 * recog.c (offsettable_address_addr_space_p): Likewise.
16131 * regcprop.c (maybe_mode_change): Likewise.
16132 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
16133 * regrename.c (build_def_use): Likewise.
16134 * regstat.c (dump_reg_info): Likewise.
16135 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
16136 (find_reloads, find_reloads_subreg_address): Likewise.
16137 * reload1.c (eliminate_regs_1): Likewise.
16138 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
16139 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
16140 (simplify_binary_operation_1, simplify_subreg): Likewise.
16141 * targhooks.c (default_function_arg_padding): Likewise.
16142 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
16143 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
16144 (verify_gimple_assign_ternary): Likewise.
16145 * tree-inline.c (estimate_move_cost): Likewise.
16146 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16147 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
16148 (get_address_cost_ainc): Likewise.
16149 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
16150 (vect_supportable_dr_alignment): Likewise.
16151 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16152 (vectorizable_reduction): Likewise.
16153 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
16154 (vectorizable_operation, vectorizable_load): Likewise.
16155 * tree.c (build_same_sized_truth_vector_type): Likewise.
16156 * valtrack.c (cleanup_auto_inc_dec): Likewise.
16157 * var-tracking.c (emit_note_insn_var_location): Likewise.
16158 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
16159 (ADDR_VEC_ALIGN): Likewise.
16161 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16162 Alan Hayward <alan.hayward@arm.com>
16163 David Sherwood <david.sherwood@arm.com>
16165 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
16167 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16168 or if measurement_type is polynomial.
16169 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
16170 * combine.c (make_extraction): Likewise.
16171 * dse.c (find_shift_sequence): Likewise.
16172 * dwarf2out.c (mem_loc_descriptor): Likewise.
16173 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
16174 (extract_bit_field, extract_low_bits): Likewise.
16175 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
16176 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
16177 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
16178 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
16179 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16180 * reload.c (find_reloads): Likewise.
16181 * reload1.c (alter_reg): Likewise.
16182 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
16183 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
16184 * tree-if-conv.c (predicate_mem_writes): Likewise.
16185 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
16186 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
16187 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
16188 * valtrack.c (dead_debug_insert_temp): Likewise.
16189 * varasm.c (mergeable_constant_section): Likewise.
16190 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
16192 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16193 Alan Hayward <alan.hayward@arm.com>
16194 David Sherwood <david.sherwood@arm.com>
16196 * expr.c (expand_assignment): Cope with polynomial mode sizes
16197 when assigning to a CONCAT.
16199 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16200 Alan Hayward <alan.hayward@arm.com>
16201 David Sherwood <david.sherwood@arm.com>
16203 * machmode.h (mode_precision): Change from unsigned short to
16205 (mode_to_precision): Return a poly_uint16 rather than an unsigned
16207 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
16208 or if measurement_type is not polynomial.
16209 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
16210 in which the mode is already known to be a scalar_int_mode.
16211 * genmodes.c (emit_mode_precision): Change the type of mode_precision
16212 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
16214 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16215 for GET_MODE_PRECISION.
16216 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16217 for GET_MODE_PRECISION.
16218 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
16220 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
16221 (expand_field_assignment, make_extraction): Likewise.
16222 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
16223 (get_last_value): Likewise.
16224 * convert.c (convert_to_integer_1): Likewise.
16225 * cse.c (cse_insn): Likewise.
16226 * expr.c (expand_expr_real_1): Likewise.
16227 * lra-constraints.c (simplify_operand_subreg): Likewise.
16228 * optabs-query.c (can_atomic_load_p): Likewise.
16229 * optabs.c (expand_atomic_load): Likewise.
16230 (expand_atomic_store): Likewise.
16231 * ree.c (combine_reaching_defs): Likewise.
16232 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
16233 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
16234 * tree.h (type_has_mode_precision_p): Likewise.
16235 * ubsan.c (instrument_si_overflow): Likewise.
16237 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16238 Alan Hayward <alan.hayward@arm.com>
16239 David Sherwood <david.sherwood@arm.com>
16241 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
16242 polynomial numbers of units.
16243 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
16244 (valid_vector_subparts_p): New function.
16245 (build_vector_type): Remove temporary shim and take the number
16246 of units as a poly_uint64 rather than an int.
16247 (build_opaque_vector_type): Take the number of units as a
16248 poly_uint64 rather than an int.
16249 * tree.c (build_vector_from_ctor): Handle polynomial
16250 TYPE_VECTOR_SUBPARTS.
16251 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
16252 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
16253 (build_vector_from_val): If the number of units is variable,
16254 use build_vec_duplicate_cst for constant operands and
16255 VEC_DUPLICATE_EXPR otherwise.
16256 (make_vector_type): Remove temporary is_constant ().
16257 (build_vector_type, build_opaque_vector_type): Take the number of
16258 units as a poly_uint64 rather than an int.
16259 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
16261 * cfgexpand.c (expand_debug_expr): Likewise.
16262 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
16263 (store_constructor, expand_expr_real_1): Likewise.
16264 (const_scalar_mask_from_tree): Likewise.
16265 * fold-const-call.c (fold_const_reduction): Likewise.
16266 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
16267 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
16268 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
16269 (fold_relational_const): Likewise.
16270 (native_interpret_vector): Likewise. Change the size from an
16271 int to an unsigned int.
16272 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
16273 TYPE_VECTOR_SUBPARTS.
16274 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
16275 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
16276 duplicating a non-constant operand into a variable-length vector.
16277 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
16278 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
16279 * ipa-icf.c (sem_variable::equals): Likewise.
16280 * match.pd: Likewise.
16281 * omp-simd-clone.c (simd_clone_subparts): Likewise.
16282 * print-tree.c (print_node): Likewise.
16283 * stor-layout.c (layout_type): Likewise.
16284 * targhooks.c (default_builtin_vectorization_cost): Likewise.
16285 * tree-cfg.c (verify_gimple_comparison): Likewise.
16286 (verify_gimple_assign_binary): Likewise.
16287 (verify_gimple_assign_ternary): Likewise.
16288 (verify_gimple_assign_single): Likewise.
16289 * tree-pretty-print.c (dump_generic_node): Likewise.
16290 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16291 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
16292 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
16293 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
16294 (vect_shift_permute_load_chain): Likewise.
16295 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
16296 (expand_vector_condition, optimize_vector_constructor): Likewise.
16297 (lower_vec_perm, get_compute_type): Likewise.
16298 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16299 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
16300 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
16301 (vect_recog_mask_conversion_pattern): Likewise.
16302 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
16303 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
16304 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16305 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
16306 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
16307 (vectorizable_shift, vectorizable_operation, vectorizable_store)
16308 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
16309 (supportable_widening_operation): Likewise.
16310 (supportable_narrowing_operation): Likewise.
16311 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
16313 * varasm.c (output_constant): Likewise.
16315 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16316 Alan Hayward <alan.hayward@arm.com>
16317 David Sherwood <david.sherwood@arm.com>
16319 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
16320 so that both the length == 3 and length != 3 cases set up their
16321 own permute vectors. Add comments explaining why we know the
16322 number of elements is constant.
16323 (vect_permute_load_chain): Likewise.
16325 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16326 Alan Hayward <alan.hayward@arm.com>
16327 David Sherwood <david.sherwood@arm.com>
16329 * machmode.h (mode_nunits): Change from unsigned char to
16331 (ONLY_FIXED_SIZE_MODES): New macro.
16332 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
16333 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
16334 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
16336 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
16337 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
16338 or if measurement_type is not polynomial.
16339 * genmodes.c (ZERO_COEFFS): New macro.
16340 (emit_mode_nunits_inline): Make mode_nunits_inline return a
16342 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
16343 Use ZERO_COEFFS when emitting initializers.
16344 * data-streamer.h (bp_pack_poly_value): New function.
16345 (bp_unpack_poly_value): Likewise.
16346 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16347 for GET_MODE_NUNITS.
16348 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16349 for GET_MODE_NUNITS.
16350 * tree.c (make_vector_type): Remove temporary shim and make
16351 the real function take the number of units as a poly_uint64
16352 rather than an int.
16353 (build_vector_type_for_mode): Handle polynomial nunits.
16354 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
16355 * emit-rtl.c (const_vec_series_p_1): Likewise.
16356 (gen_rtx_CONST_VECTOR): Likewise.
16357 * fold-const.c (test_vec_duplicate_folding): Likewise.
16358 * genrecog.c (validate_pattern): Likewise.
16359 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
16360 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16361 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
16362 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
16363 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
16364 * rtlanal.c (subreg_get_info): Likewise.
16365 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16366 (vect_grouped_load_supported): Likewise.
16367 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
16368 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
16369 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
16370 (simplify_const_unary_operation, simplify_binary_operation_1)
16371 (simplify_const_binary_operation, simplify_ternary_operation)
16372 (test_vector_ops_duplicate, test_vector_ops): Likewise.
16373 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
16374 instead of CONST_VECTOR_NUNITS.
16375 * varasm.c (output_constant_pool_2): Likewise.
16376 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
16377 explicit-encoded elements in the XVEC for variable-length vectors.
16379 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16381 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
16383 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16384 Alan Hayward <alan.hayward@arm.com>
16385 David Sherwood <david.sherwood@arm.com>
16387 * coretypes.h (fixed_size_mode): Declare.
16388 (fixed_size_mode_pod): New typedef.
16389 * builtins.h (target_builtins::x_apply_args_mode)
16390 (target_builtins::x_apply_result_mode): Change type to
16391 fixed_size_mode_pod.
16392 * builtins.c (apply_args_size, apply_result_size, result_vector)
16393 (expand_builtin_apply_args_1, expand_builtin_apply)
16394 (expand_builtin_return): Update accordingly.
16396 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16398 * cse.c (hash_rtx_cb): Hash only the encoded elements.
16399 * cselib.c (cselib_hash_rtx): Likewise.
16400 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
16401 CONST_VECTOR encoding.
16403 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16404 Jeff Law <law@redhat.com>
16407 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
16408 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
16409 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
16410 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
16413 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
16414 explicitly probe *sp in a noreturn function if there were any callee
16415 register saves or frame pointer is needed.
16417 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16420 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
16421 BLKmode for ternary, binary or unary expressions.
16424 * var-tracking.c (delete_vta_debug_insn): New inline function.
16425 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
16426 insns from get_insns () to NULL instead of each bb separately.
16427 Use delete_vta_debug_insn. No longer static.
16428 (vt_debug_insns_local, variable_tracking_main_1): Adjust
16429 delete_vta_debug_insns callers.
16430 * rtl.h (delete_vta_debug_insns): Declare.
16431 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
16432 instead of variable_tracking_main.
16434 2018-01-03 Martin Sebor <msebor@redhat.com>
16436 PR tree-optimization/83603
16437 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
16438 arguments past the endof the argument list in functions declared
16439 without a prototype.
16440 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
16441 Avoid checking when arguments are null.
16443 2018-01-03 Martin Sebor <msebor@redhat.com>
16446 * doc/extend.texi (attribute const): Fix a typo.
16447 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
16448 issuing -Wsuggest-attribute for void functions.
16450 2018-01-03 Martin Sebor <msebor@redhat.com>
16452 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
16453 offset_int::from instead of wide_int::to_shwi.
16454 (maybe_diag_overlap): Remove assertion.
16455 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
16456 * gimple-ssa-sprintf.c (format_directive): Same.
16457 (parse_directive): Same.
16458 (sprintf_dom_walker::compute_format_length): Same.
16459 (try_substitute_return_value): Same.
16461 2018-01-03 Jeff Law <law@redhat.com>
16463 PR middle-end/83654
16464 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
16465 non-constant residual for zero at runtime and avoid probing in
16466 that case. Reorganize code for trailing problem to mirror handling
16469 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16471 PR tree-optimization/83501
16472 * tree-ssa-strlen.c (get_string_cst): New.
16473 (handle_char_store): Call get_string_cst.
16475 2018-01-03 Martin Liska <mliska@suse.cz>
16477 PR tree-optimization/83593
16478 * tree-ssa-strlen.c: Include tree-cfg.h.
16479 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
16480 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
16481 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
16483 (strlen_dom_walker::before_dom_children): Call
16484 gimple_purge_dead_eh_edges. Dump tranformation with details
16486 (strlen_dom_walker::before_dom_children): Update call by adding
16487 new argument cleanup_eh.
16488 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
16490 2018-01-03 Martin Liska <mliska@suse.cz>
16493 * cif-code.def (VARIADIC_THUNK): New enum value.
16494 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
16497 2018-01-03 Jan Beulich <jbeulich@suse.com>
16499 * sse.md (mov<mode>_internal): Tighten condition for when to use
16500 vmovdqu<ssescalarsize> for TI and OI modes.
16502 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16504 Update copyright years.
16506 2018-01-03 Martin Liska <mliska@suse.cz>
16509 * ipa-visibility.c (function_and_variable_visibility): Skip
16510 functions with noipa attribure.
16512 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16514 * gcc.c (process_command): Update copyright notice dates.
16515 * gcov-dump.c (print_version): Ditto.
16516 * gcov.c (print_version): Ditto.
16517 * gcov-tool.c (print_version): Ditto.
16518 * gengtype.c (create_file): Ditto.
16519 * doc/cpp.texi: Bump @copying's copyright year.
16520 * doc/cppinternals.texi: Ditto.
16521 * doc/gcc.texi: Ditto.
16522 * doc/gccint.texi: Ditto.
16523 * doc/gcov.texi: Ditto.
16524 * doc/install.texi: Ditto.
16525 * doc/invoke.texi: Ditto.
16527 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16529 * vector-builder.h (vector_builder::m_full_nelts): Change from
16530 unsigned int to poly_uint64.
16531 (vector_builder::full_nelts): Update prototype accordingly.
16532 (vector_builder::new_vector): Likewise.
16533 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
16534 (vector_builder::operator ==): Likewise.
16535 (vector_builder::finalize): Likewise.
16536 * int-vector-builder.h (int_vector_builder::int_vector_builder):
16537 Take the number of elements as a poly_uint64 rather than an
16539 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
16540 from unsigned int to poly_uint64.
16541 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
16542 (vec_perm_indices::new_vector): Likewise.
16543 (vec_perm_indices::length): Likewise.
16544 (vec_perm_indices::nelts_per_input): Likewise.
16545 (vec_perm_indices::input_nelts): Likewise.
16546 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
16547 number of elements per input as a poly_uint64 rather than an
16548 unsigned int. Use the original encoding for variable-length
16549 vectors, rather than clamping each individual element.
16550 For the second and subsequent elements in each pattern,
16551 clamp the step and base before clamping their sum.
16552 (vec_perm_indices::series_p): Handle polynomial element counts.
16553 (vec_perm_indices::all_in_range_p): Likewise.
16554 (vec_perm_indices_to_tree): Likewise.
16555 (vec_perm_indices_to_rtx): Likewise.
16556 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
16557 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
16558 (tree_vector_builder::new_binary_operation): Handle polynomial
16559 element counts. Return false if we need to know the number
16560 of elements at compile time.
16561 * fold-const.c (fold_vec_perm): Punt if the number of elements
16562 isn't known at compile time.
16564 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16566 * vec-perm-indices.h (vec_perm_builder): Change element type
16567 from HOST_WIDE_INT to poly_int64.
16568 (vec_perm_indices::element_type): Update accordingly.
16569 (vec_perm_indices::clamp): Handle polynomial element_types.
16570 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16571 (vec_perm_indices::all_in_range_p): Likewise.
16572 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
16574 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
16575 polynomial vec_perm_indices element types.
16576 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
16577 * fold-const.c (fold_vec_perm): Likewise.
16578 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
16579 * tree-vect-generic.c (lower_vec_perm): Likewise.
16580 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16581 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
16582 element type to HOST_WIDE_INT.
16584 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16585 Alan Hayward <alan.hayward@arm.com>
16586 David Sherwood <david.sherwood@arm.com>
16588 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
16589 rather than an int. Use plus_constant.
16590 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
16591 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
16593 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16594 Alan Hayward <alan.hayward@arm.com>
16595 David Sherwood <david.sherwood@arm.com>
16597 * calls.c (emit_call_1, expand_call): Change struct_value_size from
16598 a HOST_WIDE_INT to a poly_int64.
16600 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16601 Alan Hayward <alan.hayward@arm.com>
16602 David Sherwood <david.sherwood@arm.com>
16604 * calls.c (load_register_parameters): Cope with polynomial
16605 mode sizes. Require a constant size for BLKmode parameters
16606 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
16607 forces a parameter to be padded at the lsb end in order to
16608 fill a complete number of words, require the parameter size
16609 to be ordered wrt UNITS_PER_WORD.
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 * reload1.c (spill_stack_slot_width): Change element type
16616 from unsigned int to poly_uint64_pod.
16617 (alter_reg): Treat mode sizes as polynomial.
16619 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16620 Alan Hayward <alan.hayward@arm.com>
16621 David Sherwood <david.sherwood@arm.com>
16623 * reload.c (complex_word_subreg_p): New function.
16624 (reload_inner_reg_of_subreg, push_reload): Use it.
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 * lra-constraints.c (process_alt_operands): Reject matched
16631 operands whose sizes aren't ordered.
16632 (match_reload): Refer to this check here.
16634 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16635 Alan Hayward <alan.hayward@arm.com>
16636 David Sherwood <david.sherwood@arm.com>
16638 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
16639 that the mode size is in the set {1, 2, 4, 8, 16}.
16641 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16642 Alan Hayward <alan.hayward@arm.com>
16643 David Sherwood <david.sherwood@arm.com>
16645 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
16646 Use plus_constant instead of gen_rtx_PLUS.
16648 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16649 Alan Hayward <alan.hayward@arm.com>
16650 David Sherwood <david.sherwood@arm.com>
16652 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
16653 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
16654 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
16655 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
16656 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
16657 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
16658 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
16659 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
16660 * config/i386/i386.c (ix86_push_rounding): ...this new function.
16661 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
16663 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
16664 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
16665 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
16666 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
16667 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
16668 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
16669 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
16670 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
16671 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
16672 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
16674 * expr.c (emit_move_resolve_push): Treat the input and result
16675 of PUSH_ROUNDING as a poly_int64.
16676 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
16677 (emit_push_insn): Likewise.
16678 * lra-eliminations.c (mark_not_eliminable): Likewise.
16679 * recog.c (push_operand): Likewise.
16680 * reload1.c (elimination_effects): Likewise.
16681 * rtlanal.c (nonzero_bits1): Likewise.
16682 * calls.c (store_one_arg): Likewise. Require the padding to be
16683 known at compile time.
16685 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16686 Alan Hayward <alan.hayward@arm.com>
16687 David Sherwood <david.sherwood@arm.com>
16689 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
16690 Use plus_constant instead of gen_rtx_PLUS.
16692 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16693 Alan Hayward <alan.hayward@arm.com>
16694 David Sherwood <david.sherwood@arm.com>
16696 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
16697 rather than an int.
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_1): Use tree_to_poly_uint64
16704 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
16705 via stack temporaries. Treat the mode size as polynomial too.
16707 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16708 Alan Hayward <alan.hayward@arm.com>
16709 David Sherwood <david.sherwood@arm.com>
16711 * expr.c (expand_expr_real_2): When handling conversions involving
16712 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
16713 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
16714 as a poly_uint64 too.
16716 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16717 Alan Hayward <alan.hayward@arm.com>
16718 David Sherwood <david.sherwood@arm.com>
16720 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
16722 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16723 Alan Hayward <alan.hayward@arm.com>
16724 David Sherwood <david.sherwood@arm.com>
16726 * combine.c (can_change_dest_mode): Handle polynomial
16727 REGMODE_NATURAL_SIZE.
16728 * expmed.c (store_bit_field_1): Likewise.
16729 * expr.c (store_constructor): Likewise.
16730 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
16731 and polynomial REGMODE_NATURAL_SIZE.
16732 (gen_lowpart_common): Likewise.
16733 * reginfo.c (record_subregs_of_mode): Likewise.
16734 * rtlanal.c (read_modify_subreg_p): Likewise.
16736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16737 Alan Hayward <alan.hayward@arm.com>
16738 David Sherwood <david.sherwood@arm.com>
16740 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
16741 numbers of elements.
16743 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16744 Alan Hayward <alan.hayward@arm.com>
16745 David Sherwood <david.sherwood@arm.com>
16747 * match.pd: Cope with polynomial numbers of vector elements.
16749 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16750 Alan Hayward <alan.hayward@arm.com>
16751 David Sherwood <david.sherwood@arm.com>
16753 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
16754 in a POINTER_PLUS_EXPR.
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 * omp-simd-clone.c (simd_clone_subparts): New function.
16761 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
16762 (ipa_simd_modify_function_body): Likewise.
16764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16765 Alan Hayward <alan.hayward@arm.com>
16766 David Sherwood <david.sherwood@arm.com>
16768 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
16769 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
16770 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
16771 (expand_vector_condition, vector_element): Likewise.
16772 (subparts_gt): New function.
16773 (get_compute_type): Use subparts_gt.
16774 (count_type_subparts): Delete.
16775 (expand_vector_operations_1): Use subparts_gt instead of
16776 count_type_subparts.
16778 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16779 Alan Hayward <alan.hayward@arm.com>
16780 David Sherwood <david.sherwood@arm.com>
16782 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
16783 (vect_compile_time_alias): ...this new function. Do the calculation
16784 on poly_ints rather than trees.
16785 (vect_prune_runtime_alias_test_list): Update call accordingly.
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_build_slp_tree_1): Handle polynomial
16793 (vect_schedule_slp_instance): Likewise.
16795 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16796 Alan Hayward <alan.hayward@arm.com>
16797 David Sherwood <david.sherwood@arm.com>
16799 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
16800 constant and extern definitions for variable-length vectors.
16801 (vect_get_constant_vectors): Note that the number of units
16802 is known to be constant.
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 (vectorizable_conversion): Treat the number
16809 of units as polynomial. Choose between WIDE and NARROW based
16812 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16813 Alan Hayward <alan.hayward@arm.com>
16814 David Sherwood <david.sherwood@arm.com>
16816 * tree-vect-stmts.c (simd_clone_subparts): New function.
16817 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
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 (vectorizable_call): Treat the number of
16824 vectors as polynomial. Use build_index_vector for
16825 IFN_GOMP_SIMD_LANE.
16827 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16828 Alan Hayward <alan.hayward@arm.com>
16829 David Sherwood <david.sherwood@arm.com>
16831 * tree-vect-stmts.c (get_load_store_type): Treat the number of
16832 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
16833 for variable-length vectors.
16834 (vectorizable_mask_load_store): Treat the number of units as
16835 polynomial, asserting that it is constant if the condition has
16836 already been enforced.
16837 (vectorizable_store, vectorizable_load): Likewise.
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_live_operation): Treat the number
16844 of units as polynomial. Punt if we can't tell at compile time
16845 which vector contains the final result.
16847 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16848 Alan Hayward <alan.hayward@arm.com>
16849 David Sherwood <david.sherwood@arm.com>
16851 * tree-vect-loop.c (vectorizable_induction): Treat the number
16852 of units as polynomial. Punt on SLP inductions. Use an integer
16853 VEC_SERIES_EXPR for variable-length integer reductions. Use a
16854 cast of such a series for variable-length floating-point
16857 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16858 Alan Hayward <alan.hayward@arm.com>
16859 David Sherwood <david.sherwood@arm.com>
16861 * tree.h (build_index_vector): Declare.
16862 * tree.c (build_index_vector): New function.
16863 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
16864 of units as polynomial, forcibly converting it to a constant if
16865 vectorizable_reduction has already enforced the condition.
16866 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
16867 to create a {1,2,3,...} vector.
16868 (vectorizable_reduction): Treat the number of units as polynomial.
16869 Choose vectype_in based on the largest scalar element size rather
16870 than the smallest number of units. Enforce the restrictions
16873 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16874 Alan Hayward <alan.hayward@arm.com>
16875 David Sherwood <david.sherwood@arm.com>
16877 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
16878 number of units as polynomial.
16880 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16881 Alan Hayward <alan.hayward@arm.com>
16882 David Sherwood <david.sherwood@arm.com>
16884 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
16885 * target.def (autovectorize_vector_sizes): Return the vector sizes
16886 by pointer, using vector_sizes rather than a bitmask.
16887 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
16888 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
16889 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
16891 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
16892 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
16893 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
16894 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
16895 * omp-general.c (omp_max_vf): Likewise.
16896 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16897 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
16898 * tree-vect-loop.c (vect_analyze_loop): Likewise.
16899 * tree-vect-slp.c (vect_slp_bb): Likewise.
16900 * doc/tm.texi: Regenerate.
16901 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
16903 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
16904 the vector size as a poly_uint64 rather than an unsigned int.
16905 (current_vector_size): Change from an unsigned int to a poly_uint64.
16906 (get_vectype_for_scalar_type): Update accordingly.
16907 * tree.h (build_truth_vector_type): Take the size and number of
16908 units as a poly_uint64 rather than an unsigned int.
16909 (build_vector_type): Add a temporary overload that takes
16910 the number of units as a poly_uint64 rather than an unsigned int.
16911 * tree.c (make_vector_type): Likewise.
16912 (build_truth_vector_type): Take the number of units as a poly_uint64
16913 rather than an unsigned int.
16915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16916 Alan Hayward <alan.hayward@arm.com>
16917 David Sherwood <david.sherwood@arm.com>
16919 * target.def (get_mask_mode): Take the number of units and length
16920 as poly_uint64s rather than unsigned ints.
16921 * targhooks.h (default_get_mask_mode): Update accordingly.
16922 * targhooks.c (default_get_mask_mode): Likewise.
16923 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
16924 * doc/tm.texi: Regenerate.
16926 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16927 Alan Hayward <alan.hayward@arm.com>
16928 David Sherwood <david.sherwood@arm.com>
16930 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
16931 * omp-general.c (omp_max_vf): Likewise.
16932 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
16933 (expand_omp_simd): Handle polynomial safelen.
16934 * omp-low.c (omplow_simd_context): Add a default constructor.
16935 (omplow_simd_context::max_vf): Change from int to poly_uint64.
16936 (lower_rec_simd_input_clauses): Update accordingly.
16937 (lower_rec_input_clauses): Likewise.
16939 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16940 Alan Hayward <alan.hayward@arm.com>
16941 David Sherwood <david.sherwood@arm.com>
16943 * tree-vectorizer.h (vect_nunits_for_cost): New function.
16944 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
16945 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
16946 (vect_analyze_slp_cost): Likewise.
16947 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
16948 (vect_model_load_cost): Likewise.
16950 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16951 Alan Hayward <alan.hayward@arm.com>
16952 David Sherwood <david.sherwood@arm.com>
16954 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
16955 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
16956 from an unsigned int * to a poly_uint64_pod *.
16957 (calculate_unrolling_factor): New function.
16958 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
16960 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16961 Alan Hayward <alan.hayward@arm.com>
16962 David Sherwood <david.sherwood@arm.com>
16964 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
16965 from an unsigned int to a poly_uint64.
16966 (_loop_vec_info::slp_unrolling_factor): Likewise.
16967 (_loop_vec_info::vectorization_factor): Change from an int
16969 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
16970 (vect_get_num_vectors): New function.
16971 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
16972 (vect_get_num_copies): Use vect_get_num_vectors.
16973 (vect_analyze_data_ref_dependences): Change max_vf from an int *
16974 to an unsigned int *.
16975 (vect_analyze_data_refs): Change min_vf from an int * to a
16977 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16978 than an unsigned HOST_WIDE_INT.
16979 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
16980 (vect_analyze_data_ref_dependence): Change max_vf from an int *
16981 to an unsigned int *.
16982 (vect_analyze_data_ref_dependences): Likewise.
16983 (vect_compute_data_ref_alignment): Handle polynomial vf.
16984 (vect_enhance_data_refs_alignment): Likewise.
16985 (vect_prune_runtime_alias_test_list): Likewise.
16986 (vect_shift_permute_load_chain): Likewise.
16987 (vect_supportable_dr_alignment): Likewise.
16988 (dependence_distance_ge_vf): Take the vectorization factor as a
16989 poly_uint64 rather than an unsigned HOST_WIDE_INT.
16990 (vect_analyze_data_refs): Change min_vf from an int * to a
16992 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
16993 vfm1 as a poly_uint64 rather than an int. Make the same change
16994 for the returned bound_scalar.
16995 (vect_gen_vector_loop_niters): Handle polynomial vf.
16996 (vect_do_peeling): Likewise. Update call to
16997 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
16998 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
17000 * tree-vect-loop.c (vect_determine_vectorization_factor)
17001 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
17002 (vect_get_known_peeling_cost): Likewise.
17003 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
17004 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
17005 (vect_transform_loop): Likewise. Use the lowest possible VF when
17006 updating the upper bounds of the loop.
17007 (vect_min_worthwhile_factor): Make static. Return an unsigned int
17008 rather than an int.
17009 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
17010 polynomial unroll factors.
17011 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
17012 (vect_make_slp_decision): Likewise.
17013 (vect_supported_load_permutation_p): Likewise, and polynomial
17015 (vect_analyze_slp_cost): Handle polynomial vf.
17016 (vect_slp_analyze_node_operations): Likewise.
17017 (vect_slp_analyze_bb_1): Likewise.
17018 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17019 than an unsigned HOST_WIDE_INT.
17020 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
17021 (vectorizable_load): Handle polynomial vf.
17022 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
17024 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
17026 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17027 Alan Hayward <alan.hayward@arm.com>
17028 David Sherwood <david.sherwood@arm.com>
17030 * match.pd: Handle bit operations involving three constants
17031 and try to fold one pair.
17033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17035 * tree-vect-loop-manip.c: Include gimple-fold.h.
17036 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
17037 niters_maybe_zero parameters. Handle other cases besides a step of 1.
17038 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
17039 Add a path that uses a step of VF instead of 1, but disable it
17041 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
17042 and niters_no_overflow parameters. Update calls to
17043 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
17044 Create a new SSA name if the latter choses to use a ste other
17045 than zero, and return it via niters_vector_mult_vf_var.
17046 * tree-vect-loop.c (vect_transform_loop): Update calls to
17047 vect_do_peeling, vect_gen_vector_loop_niters and
17048 slpeel_make_loop_iterate_ntimes.
17049 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
17050 (vect_gen_vector_loop_niters): Update declarations after above changes.
17052 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
17054 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
17055 128-bit round to integer instructions.
17056 (ceil<mode>2): Likewise.
17057 (btrunc<mode>2): Likewise.
17058 (round<mode>2): Likewise.
17060 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17062 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
17063 unaligned VSX load/store on P8/P9.
17064 (expand_block_clear): Allow the use of unaligned VSX
17065 load/store on P8/P9.
17067 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17069 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
17071 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
17072 swap associated with both a load and a store.
17074 2018-01-02 Andrew Waterman <andrew@sifive.com>
17076 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
17077 * config/riscv/riscv.md (clear_cache): Use it.
17079 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
17081 * web.c: Remove out-of-date comment.
17083 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17085 * expr.c (fixup_args_size_notes): Check that any existing
17086 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
17087 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
17088 (emit_single_push_insn): ...here.
17090 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17092 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
17093 (const_vector_encoded_nelts): New function.
17094 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
17095 (const_vector_int_elt, const_vector_elt): Declare.
17096 * emit-rtl.c (const_vector_int_elt_1): New function.
17097 (const_vector_elt): Likewise.
17098 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
17099 of CONST_VECTOR_ELT.
17101 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17103 * expr.c: Include rtx-vector-builder.h.
17104 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
17105 directly on the tree encoding.
17106 (const_vector_from_tree): Likewise.
17107 * optabs.c: Include rtx-vector-builder.h.
17108 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
17109 sequence of "u" values.
17110 * vec-perm-indices.c: Include rtx-vector-builder.h.
17111 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
17112 directly on the vec_perm_indices encoding.
17114 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17116 * doc/rtl.texi (const_vector): Describe new encoding scheme.
17117 * Makefile.in (OBJS): Add rtx-vector-builder.o.
17118 * rtx-vector-builder.h: New file.
17119 * rtx-vector-builder.c: Likewise.
17120 * rtl.h (rtx_def::u2): Add a const_vector field.
17121 (CONST_VECTOR_NPATTERNS): New macro.
17122 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
17123 (CONST_VECTOR_DUPLICATE_P): Likewise.
17124 (CONST_VECTOR_STEPPED_P): Likewise.
17125 (CONST_VECTOR_ENCODED_ELT): Likewise.
17126 (const_vec_duplicate_p): Check for a duplicated vector encoding.
17127 (unwrap_const_vec_duplicate): Likewise.
17128 (const_vec_series_p): Check for a non-duplicated vector encoding.
17129 Say that the function only returns true for integer vectors.
17130 * emit-rtl.c: Include rtx-vector-builder.h.
17131 (gen_const_vec_duplicate_1): Delete.
17132 (gen_const_vector): Call gen_const_vec_duplicate instead of
17133 gen_const_vec_duplicate_1.
17134 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
17135 (gen_const_vec_duplicate): Use rtx_vector_builder.
17136 (gen_const_vec_series): Likewise.
17137 (gen_rtx_CONST_VECTOR): Likewise.
17138 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
17139 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17140 Build a new vector rather than modifying a CONST_VECTOR in-place.
17141 (handle_special_swappables): Update call accordingly.
17142 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
17143 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17144 Build a new vector rather than modifying a CONST_VECTOR in-place.
17145 (handle_special_swappables): Update call accordingly.
17147 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17149 * simplify-rtx.c (simplify_const_binary_operation): Use
17150 CONST_VECTOR_ELT instead of XVECEXP.
17152 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17154 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
17155 the selector elements to be different from the data elements
17156 if the selector is a VECTOR_CST.
17157 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
17158 ssizetype for the selector.
17160 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17162 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
17163 before testing each element individually.
17164 * tree-vect-generic.c (lower_vec_perm): Likewise.
17166 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17168 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
17169 * selftest-run-tests.c (selftest::run_tests): Call it.
17170 * vector-builder.h (vector_builder::operator ==): New function.
17171 (vector_builder::operator !=): Likewise.
17172 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
17173 (vec_perm_indices::all_from_input_p): New function.
17174 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17175 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
17176 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
17177 instead of reading the VECTOR_CST directly. Detect whether both
17178 vector inputs are the same before constructing the vec_perm_indices,
17179 and update the number of inputs argument accordingly. Use the
17180 utility functions added above. Only construct sel2 if we need to.
17182 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17184 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
17185 the broadcast of the low byte.
17186 (expand_mult_highpart): Use an explicit encoding for the permutes.
17187 * optabs-query.c (can_mult_highpart_p): Likewise.
17188 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
17189 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17190 (vectorizable_bswap): Likewise.
17191 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
17192 explicit encoding for the power-of-2 permutes.
17193 (vect_permute_store_chain): Likewise.
17194 (vect_grouped_load_supported): Likewise.
17195 (vect_permute_load_chain): Likewise.
17197 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17199 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
17200 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
17201 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
17202 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17203 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
17204 (vect_gen_perm_mask_any): Likewise.
17206 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17208 * int-vector-builder.h: New file.
17209 * vec-perm-indices.h: Include int-vector-builder.h.
17210 (vec_perm_indices): Redefine as an int_vector_builder.
17211 (auto_vec_perm_indices): Delete.
17212 (vec_perm_builder): Redefine as a stand-alone class.
17213 (vec_perm_indices::vec_perm_indices): New function.
17214 (vec_perm_indices::clamp): Likewise.
17215 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
17216 (vec_perm_indices::new_vector): New function.
17217 (vec_perm_indices::new_expanded_vector): Update for new
17218 vec_perm_indices class.
17219 (vec_perm_indices::rotate_inputs): New function.
17220 (vec_perm_indices::all_in_range_p): Operate directly on the
17221 encoded form, without computing elided elements.
17222 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
17223 encoding. Update for new vec_perm_indices class.
17224 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
17225 the given vec_perm_builder.
17226 (expand_vec_perm_var): Update vec_perm_builder constructor.
17227 (expand_mult_highpart): Use vec_perm_builder instead of
17228 auto_vec_perm_indices.
17229 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
17230 vec_perm_indices instead of auto_vec_perm_indices. Use a single
17231 or double series encoding as appropriate.
17232 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
17233 vec_perm_indices instead of auto_vec_perm_indices.
17234 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17235 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17236 (vect_permute_store_chain): Likewise.
17237 (vect_grouped_load_supported): Likewise.
17238 (vect_permute_load_chain): Likewise.
17239 (vect_shift_permute_load_chain): Likewise.
17240 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17241 (vect_transform_slp_perm_load): Likewise.
17242 (vect_schedule_slp_instance): Likewise.
17243 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17244 (vectorizable_mask_load_store): Likewise.
17245 (vectorizable_bswap): Likewise.
17246 (vectorizable_store): Likewise.
17247 (vectorizable_load): Likewise.
17248 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
17249 vec_perm_indices instead of auto_vec_perm_indices. Use
17250 tree_to_vec_perm_builder to read the vector from a tree.
17251 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
17252 vec_perm_builder instead of a vec_perm_indices.
17253 (have_whole_vector_shift): Use vec_perm_builder and
17254 vec_perm_indices instead of auto_vec_perm_indices. Leave the
17255 truncation to calc_vec_perm_mask_for_shift.
17256 (vect_create_epilog_for_reduction): Likewise.
17257 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
17258 from auto_vec_perm_indices to vec_perm_indices.
17259 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17260 instead of changing individual elements.
17261 (aarch64_vectorize_vec_perm_const): Use new_vector to install
17262 the vector in d.perm.
17263 * config/arm/arm.c (expand_vec_perm_d::perm): Change
17264 from auto_vec_perm_indices to vec_perm_indices.
17265 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17266 instead of changing individual elements.
17267 (arm_vectorize_vec_perm_const): Use new_vector to install
17268 the vector in d.perm.
17269 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
17270 Update vec_perm_builder constructor.
17271 (rs6000_expand_interleave): Likewise.
17272 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
17273 (rs6000_expand_interleave): Likewise.
17275 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17277 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
17278 to qimode could truncate the indices.
17279 * optabs.c (expand_vec_perm_var): Likewise.
17281 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17283 * Makefile.in (OBJS): Add vec-perm-indices.o.
17284 * vec-perm-indices.h: New file.
17285 * vec-perm-indices.c: Likewise.
17286 * target.h (vec_perm_indices): Replace with a forward class
17288 (auto_vec_perm_indices): Move to vec-perm-indices.h.
17289 * optabs.h: Include vec-perm-indices.h.
17290 (expand_vec_perm): Delete.
17291 (selector_fits_mode_p, expand_vec_perm_var): Declare.
17292 (expand_vec_perm_const): Declare.
17293 * target.def (vec_perm_const_ok): Replace with...
17294 (vec_perm_const): ...this new hook.
17295 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
17296 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
17297 * doc/tm.texi: Regenerate.
17298 * optabs.def (vec_perm_const): Delete.
17299 * doc/md.texi (vec_perm_const): Likewise.
17300 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
17301 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
17302 expand_vec_perm for constant permutation vectors. Assert that
17303 the mode of variable permutation vectors is the integer equivalent
17304 of the mode that is being permuted.
17305 * optabs-query.h (selector_fits_mode_p): Declare.
17306 * optabs-query.c: Include vec-perm-indices.h.
17307 (selector_fits_mode_p): New function.
17308 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
17309 is defined, instead of checking whether the vec_perm_const_optab
17310 exists. Use targetm.vectorize.vec_perm_const instead of
17311 targetm.vectorize.vec_perm_const_ok. Check whether the indices
17312 fit in the vector mode before using a variable permute.
17313 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
17314 vec_perm_indices instead of an rtx.
17315 (expand_vec_perm): Replace with...
17316 (expand_vec_perm_const): ...this new function. Take the selector
17317 as a vec_perm_indices rather than an rtx. Also take the mode of
17318 the selector. Update call to shift_amt_for_vec_perm_mask.
17319 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
17320 Use vec_perm_indices::new_expanded_vector to expand the original
17321 selector into bytes. Check whether the indices fit in the vector
17322 mode before using a variable permute.
17323 (expand_vec_perm_var): Make global.
17324 (expand_mult_highpart): Use expand_vec_perm_const.
17325 * fold-const.c: Includes vec-perm-indices.h.
17326 * tree-ssa-forwprop.c: Likewise.
17327 * tree-vect-data-refs.c: Likewise.
17328 * tree-vect-generic.c: Likewise.
17329 * tree-vect-loop.c: Likewise.
17330 * tree-vect-slp.c: Likewise.
17331 * tree-vect-stmts.c: Likewise.
17332 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
17334 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
17335 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
17336 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
17337 (aarch64_vectorize_vec_perm_const): ...this new function.
17338 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17339 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17340 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
17341 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
17342 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17343 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17344 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
17346 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
17347 check for NEON modes.
17348 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
17349 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
17350 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
17351 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
17353 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
17354 the old VEC_PERM_CONST conditions.
17355 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
17356 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
17357 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
17358 (ia64_vectorize_vec_perm_const_ok): Merge into...
17359 (ia64_vectorize_vec_perm_const): ...this new function.
17360 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
17361 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
17362 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
17363 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
17364 * config/mips/mips.c (mips_expand_vec_perm_const)
17365 (mips_vectorize_vec_perm_const_ok): Merge into...
17366 (mips_vectorize_vec_perm_const): ...this new function.
17367 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
17368 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
17369 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
17370 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
17371 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
17372 (rs6000_expand_vec_perm_const): Delete.
17373 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
17375 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17376 (altivec_expand_vec_perm_const_le): Take each operand individually.
17377 Operate on constant selectors rather than rtxes.
17378 (altivec_expand_vec_perm_const): Likewise. Update call to
17379 altivec_expand_vec_perm_const_le.
17380 (rs6000_expand_vec_perm_const): Delete.
17381 (rs6000_vectorize_vec_perm_const_ok): Delete.
17382 (rs6000_vectorize_vec_perm_const): New function.
17383 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17384 an element count and rtx array.
17385 (rs6000_expand_extract_even): Update call accordingly.
17386 (rs6000_expand_interleave): Likewise.
17387 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
17388 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
17389 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
17390 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
17391 (rs6000_expand_vec_perm_const): Delete.
17392 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17393 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17394 (altivec_expand_vec_perm_const_le): Take each operand individually.
17395 Operate on constant selectors rather than rtxes.
17396 (altivec_expand_vec_perm_const): Likewise. Update call to
17397 altivec_expand_vec_perm_const_le.
17398 (rs6000_expand_vec_perm_const): Delete.
17399 (rs6000_vectorize_vec_perm_const_ok): Delete.
17400 (rs6000_vectorize_vec_perm_const): New function. Remove stray
17401 reference to the SPE evmerge intructions.
17402 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17403 an element count and rtx array.
17404 (rs6000_expand_extract_even): Update call accordingly.
17405 (rs6000_expand_interleave): Likewise.
17406 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
17407 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
17409 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17411 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17413 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
17414 vector mode and that that mode matches the mode of the data
17416 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
17417 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
17418 directly using expand_vec_perm_1 when forcing selectors into
17420 (expand_vec_perm_var): New function, split out from expand_vec_perm.
17422 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17424 * optabs-query.h (can_vec_perm_p): Delete.
17425 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
17426 * optabs-query.c (can_vec_perm_p): Split into...
17427 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
17428 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
17429 particular selector is valid.
17430 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17431 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17432 (vect_grouped_load_supported): Likewise.
17433 (vect_shift_permute_load_chain): Likewise.
17434 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17435 (vect_transform_slp_perm_load): Likewise.
17436 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17437 (vectorizable_bswap): Likewise.
17438 (vect_gen_perm_mask_checked): Likewise.
17439 * fold-const.c (fold_ternary_loc): Likewise. Don't take
17440 implementations of variable permutation vectors into account
17441 when deciding which selector to use.
17442 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
17443 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
17444 with a false third argument.
17445 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
17446 to test whether the constant selector is valid and can_vec_perm_var_p
17447 to test whether a variable selector is valid.
17449 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17451 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
17452 * optabs-query.c (can_vec_perm_p): Likewise.
17453 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
17454 instead of vec_perm_indices.
17455 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
17456 (vect_gen_perm_mask_checked): Likewise,
17457 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
17458 (vect_gen_perm_mask_checked): Likewise,
17460 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17462 * optabs-query.h (qimode_for_vec_perm): Declare.
17463 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
17464 (qimode_for_vec_perm): ...this new function.
17465 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
17467 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17469 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
17470 does not have a conditional at the top.
17472 2018-01-02 Richard Biener <rguenther@suse.de>
17474 * ipa-inline.c (big_speedup_p): Fix expression.
17476 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17479 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
17482 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17486 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
17487 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
17488 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
17489 cond_taken_branch_cost 3->4.
17491 2018-01-01 Jakub Jelinek <jakub@redhat.com>
17493 PR tree-optimization/83581
17494 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
17495 TODO_cleanup_cfg if any changes have been made.
17497 PR middle-end/83608
17498 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
17499 convert_modes if target mode has the right side, but different mode
17502 PR middle-end/83609
17503 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
17504 last argument when extracting from CONCAT. If either from_real or
17505 from_imag is NULL, use expansion through memory. If result is not
17506 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
17507 the parts directly to inner mode, if even that fails, use expansion
17510 PR middle-end/83623
17511 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
17512 check for bswap in mode rather than HImode and use that in expand_unop
17515 Copyright (C) 2018 Free Software Foundation, Inc.
17517 Copying and distribution of this file, with or without modification,
17518 are permitted in any medium without royalty provided the copyright
17519 notice and this notice are preserved.