1 2018-09-21 Olivier Hainque <hainque@adacore.com>
3 * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
4 (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
6 2018-09-21 Olivier Hainque <hainque@adacore.com>
8 * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
9 (PTRDIFF_TYPE): Likewise.
11 2018-09-21 Olivier Hainque <hainque@adacore.com>
13 * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
14 triplet, similar to support for VxWorks7.
15 * config/vxworks-dummy.h: Provide a default definition
16 of TARGET_VXWORKS64 to 0.
18 2018-09-21 Olivier Hainque <hainque@adacore.com>
20 * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
21 * config/vxworks-dummy.h: here.
23 2018-09-21 Olivier Hainque <hainque@adacore.com>
25 * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
27 2018-09-21 Shaokun Zhang <zhangshaokun@hisilicon.com>
28 Bo Zhou <zbo.zhou@hisilicon.com>
30 * config/aarch64/aarch64-cores.def (tsv110): New CPU.
31 * config/aarch64/aarch64-tune.md: Regenerated.
32 * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
33 * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
34 * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
36 2018-09-21 Andrew Stubbs <ams@codesourcery.com>
37 Julian Brown <julian@codesourcery.com>
39 * builtins.c (get_builtin_sync_mem): Handle address spaces.
41 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
43 * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
44 if the call takes a static chain.
46 2018-09-21 Martin Liska <mliska@suse.cz>
48 * auto-profile.c (autofdo_source_profile::read): Do not
50 (read_profile): Do not add working sets.
51 (read_autofdo_file): Remove sum_all.
52 (afdo_callsite_hot_enough_for_early_inline): Remove const
54 * coverage.c (struct counts_entry): Remove gcov_summary.
55 (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
56 do not support GCOV_TAG_PROGRAM_SUMMARY.
57 (get_coverage_counts): Remove summary and expected
59 * coverage.h (get_coverage_counts): Likewise.
60 * doc/gcov-dump.texi: Remove -w option.
61 * gcov-dump.c (dump_working_sets): Remove.
62 (main): Do not support '-w' option.
63 (print_usage): Likewise.
64 (tag_summary): Likewise.
65 * gcov-io.c (gcov_write_summary): Do not dump
67 (gcov_read_summary): Likewise.
68 (gcov_histo_index): Remove.
69 (gcov_histogram_merge): Likewise.
70 (compute_working_sets): Likewise.
71 * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
73 (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
74 (GCOV_TAG_SUMMARY_LENGTH): Adjust.
75 (GCOV_HISTOGRAM_SIZE): Remove.
76 (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
77 (struct gcov_summary): Simplify rapidly just
78 to runs and sum_max fields.
79 (gcov_histo_index): Remove.
80 (NUM_GCOV_WORKING_SETS): Likewise.
81 (compute_working_sets): Likewise.
82 * gcov-tool.c (print_overlap_usage_message): Remove
84 * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
85 (output_lines): Remove program related line.
86 * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
87 * lto-cgraph.c (output_profile_summary): Do not stream GCOV
89 (input_profile_summary): Do not read it.
90 (merge_profile_summaries): And do not merge it.
91 (input_symtab): Do not call removed function.
92 * modulo-sched.c (sms_schedule): Do not print sum_max.
93 * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
94 removed when histogram method was invented.
95 (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
97 * postreload-gcse.c (eliminate_partially_redundant_load): Fix
99 * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
100 and dump selected value.
101 * profile.c (add_working_set): Remove.
102 (get_working_sets): Likewise.
103 (find_working_set): Likewise.
104 (get_exec_counts): Do not work with working sets.
105 (read_profile_edge_counts): Do not inform as sum_max is removed.
106 (compute_branch_probabilities): Likewise.
107 (compute_value_histograms): Remove argument for call of
109 * profile.h: Do not make gcov_summary const.
111 2018-09-21 Monk Chiang <sh.chiang04@gmail.com>
113 * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
115 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
117 PR tree-optimization/86990
118 * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
119 Check that the entire merged store group is made of constants only for
122 2018-09-20 Allan Sandfeld Jensen <allan.jensen@qt.io>
124 * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
125 (VTABLE_VERIFICATION_SPEC): Likewise.
126 (SANITIZER_EARLY_SPEC): Likewise.
127 (SANITIZER_SPEC): Likewise.
128 * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
129 * doc/invoke.texi (Link Options): Document -r.
131 2018-09-20 Richard Biener <rguenther@suse.de>
134 * gimplify.c (gimplify_expr): Retain alignment of
135 addressable lvalue in dereference.
137 2018-09-20 Alexandre Oliva <aoliva@redhat.com>
140 * configure.ac: Check for .loc is_stmt support.
141 * configure, config.in: Rebuilt.
142 * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
145 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
147 * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
148 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
151 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
153 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
155 * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
158 2018-09-20 Marek Polacek <polacek@redhat.com>
160 * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
162 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
164 PR tree-optimization/87288
165 * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
166 into account when determining PEELING_FOR_NITERS.
168 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
170 PR tree-optimization/86877
171 * tree-vect-loop.c (vect_analyze_loop_2): Call
172 vect_verify_datarefs_alignment.
174 2018-09-19 Marek Polacek <polacek@redhat.com>
176 * doc/invoke.texi: Document -Wclass-conversion.
178 2018-09-19 John David Anglin <danglin@gcc.gnu.org>
180 * config/pa/pa.c (pa_adjust_priority): Delete.
181 (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
183 * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
184 (atomic_storehi): Likewise.
185 (atomic_storesi): Likewise.
186 (atomic_loaddi): Restore compare and swap exchange loop code.
188 2018-09-19 Segher Boessenkool <segher@kernel.crashing.org>
190 PR rtl-optimization/86902
191 * combine.c (try_combine): When changing the CC mode used, don't change
192 an unrelated mode in other_insn to that new CC mode.
194 2018-09-19 David Malcolm <dmalcolm@redhat.com>
196 * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
197 with %T in place of calls to dump_generic_expr.
198 (prune_runtime_alias_test_list): Likewise.
199 (create_runtime_alias_checks): Likewise.
200 * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
201 (vect_analyze_data_ref_dependence): Likewise.
202 (vect_slp_analyze_data_ref_dependence): Likewise.
203 (vect_record_base_alignment): Likewise. Use %G in place of call
205 (vect_compute_data_ref_alignment): Likewise.
206 (verify_data_ref_alignment): Likewise.
207 (vect_find_same_alignment_drs): Likewise.
208 (vect_analyze_group_access_1): Likewise.
209 (vect_analyze_data_ref_accesses): Likewise.
210 (dependence_distance_ge_vf): Likewise.
211 (dump_lower_bound): Likewise.
212 (vect_prune_runtime_alias_test_list): Likewise.
213 (vect_find_stmt_data_reference): Likewise.
214 (vect_analyze_data_refs): Likewise.
215 (vect_create_addr_base_for_vector_ref): Likewise.
216 (vect_create_data_ref_ptr): Likewise.
217 * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
218 (vect_can_advance_ivs_p): Likewise.
219 (vect_update_ivs_after_vectorizer): Likewise.
220 (vect_gen_prolog_loop_niters): Likewise.
221 (vect_prepare_for_masked_peels): Likewise.
222 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
223 (vect_determine_vectorization_factor): Likewise.
224 (vect_is_simple_iv_evolution): Likewise.
225 (vect_analyze_scalar_cycles_1): Likewise.
226 (vect_analyze_loop_operations): Likewise.
227 (report_vect_op): Likewise.
228 (vect_is_slp_reduction): Likewise.
229 (check_reduction_path): Likewise.
230 (vect_is_simple_reduction): Likewise.
231 (vect_create_epilog_for_reduction): Likewise.
232 (vect_finalize_reduction:): Likewise.
233 (vectorizable_induction): Likewise.
234 (vect_transform_loop_stmt): Likewise.
235 (vect_transform_loop): Likewise.
236 (optimize_mask_stores): Likewise.
237 * tree-vect-patterns.c (vect_pattern_detected): Likewise.
238 (vect_split_statement): Likewise.
239 (vect_recog_over_widening_pattern): Likewise.
240 (vect_recog_average_pattern): Likewise.
241 (vect_determine_min_output_precision_1): Likewise.
242 (vect_determine_precisions_from_range): Likewise.
243 (vect_determine_precisions_from_users): Likewise.
244 (vect_mark_pattern_stmts): Likewise.
245 (vect_pattern_recog_1): Likewise.
246 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
247 (vect_record_max_nunits): Likewise.
248 (vect_build_slp_tree_1): Likewise.
249 (vect_build_slp_tree_2): Likewise.
250 (vect_print_slp_tree): Likewise.
251 (vect_analyze_slp_instance): Likewise.
252 (vect_detect_hybrid_slp_stmts): Likewise.
253 (vect_detect_hybrid_slp_1): Likewise.
254 (vect_slp_analyze_operations): Likewise.
255 (vect_slp_analyze_bb_1): Likewise.
256 (vect_transform_slp_perm_load): Likewise.
257 (vect_schedule_slp_instance): Likewise.
258 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
259 (vect_mark_stmts_to_be_vectorized): Likewise.
260 (vect_init_vector_1): Likewise.
261 (vect_get_vec_def_for_operand): Likewise.
262 (vect_finish_stmt_generation_1): Likewise.
263 (vect_check_load_store_mask): Likewise.
264 (vectorizable_call): Likewise.
265 (vectorizable_conversion): Likewise.
266 (vectorizable_operation): Likewise.
267 (vectorizable_load): Likewise.
268 (vect_analyze_stmt): Likewise.
269 (vect_is_simple_use): Likewise.
270 (vect_get_vector_types_for_stmt): Likewise.
271 (vect_get_mask_type_for_stmt): Likewise.
272 * tree-vectorizer.c (increase_alignment): Likewise.
274 2018-09-19 Andrew Stubbs <ams@codesourcery.com>
276 * doc/rtl.texi: Adjust vec_select description.
277 * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
278 non-constant selectors.
280 2018-09-19 Matthew Malcomson <matthew.malcomson@arm.com>
282 * config/aarch64/aarch64-protos.h
283 (aarch64_offset_9bit_signed_unscaled_p): New declaration.
284 * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
285 (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
286 * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
287 (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
288 (AARCH64_FL_PROFILE): Move index so flags are ordered.
289 (AARCH64_ISA_RCPC8_4): New flag.
290 * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
291 to aarch64_offset_9bit_signed_unscaled_p.
292 * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
294 * config/aarch64/constraints.md (Ust): New constraint.
295 * config/aarch64/predicates.md.
296 (aarch64_9bit_offset_memory_operand): New predicate.
297 (aarch64_rcpc_memory_operand): New predicate.
299 2018-09-19 Eric Botcazou <ebotcazou@adacore.com>
301 PR rtl-optimization/87361
302 * rtlanal.c (nonzero_bits1): Revert accidental change.
304 2018-09-19 Richard Biener <rguenther@suse.de>
306 PR tree-optimization/87349
307 PR tree-optimization/87342
308 * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
310 2018-09-18 Marek Polacek <polacek@redhat.com>
312 P1064R0 - Allowing Virtual Function Calls in Constant Expressions
313 * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
315 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
317 * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
319 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
321 PR rtl-optimization/86882
322 * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
324 2018-09-18 Uros Bizjak <ubizjak@gmail.com>
326 * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
327 *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
329 2018-09-18 Jonathan Wakely <jwakely@redhat.com>
332 * doc/invoke.texi (Link Options): Fix formatting and grammar.
334 2018-09-18 Richard Biener <rguenther@suse.de>
337 * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
338 (compute_samebase_partition_bases): Likewise.
339 (coalesce_ssa_name): Always use compute_optimized_partition_bases.
340 (gimple_can_coalesce_p): Simplify.
342 2018-09-18 Hans-Peter Nilsson <hp@bitrange.com>
344 Handle a library implementation of ffs calling __builtin_ffs.
345 * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
346 (mmix_init_libfuncs): New function: make __builtin_ffs expand
349 2018-09-17 David Malcolm <dmalcolm@redhat.com>
351 * diagnostic-show-locus.c (class layout_range): Add field
353 (layout_range::layout_range): Add "original_idx" param and use it
354 to initialize new field.
355 (make_range): Use 0 for original_idx.
356 (layout::layout): Pass in index to calls to
357 maybe_add_location_range.
358 (layout::maybe_add_location_range): Add param "original_idx" and
359 pass it on to layout_range.
360 (layout::print_any_labels): Pass on range->m_original_idx to
362 (gcc_rich_location::add_location_if_nearby): Use 0 for
364 * gcc-rich-location.h (text_range_label::get_text): Update for new
366 (range_label_for_type_mismatch::get_text): Likewise.
368 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
370 * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
372 2018-09-17 David Malcolm <dmalcolm@redhat.com>
374 * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
375 format_string_diagnostic_t.
376 (fmtwarn_n): Likewise.
377 * substring-locations.c
378 (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
379 (format_warning_n_va): Convert to...
380 (format_string_diagnostic_t::emit_warning_n_va): ...this.
381 (format_warning_va): Convert to...
382 (format_string_diagnostic_t::emit_warning_va): ...this.
383 (format_warning_at_substring): Convert to...
384 (format_string_diagnostic_t::emit_warning): ...this.
385 (format_warning_at_substring_n): Convert to...
386 (format_string_diagnostic_t::emit_warning_n): ...this.
387 * substring-locations.h (class format_string_diagnostic_t): New
389 (format_warning_va): Convert to
390 format_string_diagnostic_t::emit_warning_va.
391 (format_warning_n_va): Convert to
392 format_string_diagnostic_t::emit_warning_n_va.
393 (format_warning_at_substring): Convert to
394 format_string_diagnostic_t::emit_warning.
395 (format_warning_at_substring_n): Convert to
396 format_string_diagnostic_t::emit_warning_n.
398 2018-09-17 Cesar Philippidis <cesar@codesourcery.com>
399 Bernd Schmidt <bernds_cb1@t-online.de>
401 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
404 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
406 * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
407 operand 0 predicate to nonimmediate operand.
408 (rint<mode>2_frndint): Remove insn pattern.
409 (rint<mode>2): Change operand 1 predicate to general_operand.
410 Extend operand 1 to XFmode and generate rintxf2 insn.
411 (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
412 Do not use X87MODEF mode macro.
413 (frndintxf2_<rounding>_i387): Rename from
414 frndint<mode>2_<rounding>_i387. Do not use X87MODEF mode macro.
415 (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
416 to XFmode and generate significandxf3 insn.
418 2018-09-17 Richard Biener <rguenther@suse.de>
420 PR tree-optimization/87328
421 * tree-ssa-sccvn.c (process_bb): Remove assertion about not
422 visiting unexecutable backedges when not iterating.
423 (do_rpo_vn): Mark all edges not executable even when not
426 2018-09-17 Martin Jambor <mjambor@suse.cz>
429 * doc/invoke.texi (Warning Options): Likewise.
431 2018-09-17 Richard Biener <rguenther@suse.de>
433 PR tree-optimization/87301
434 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
435 clean EH info from leftover copy assignments.
437 2018-09-17 Martin Liska <mliska@suse.cz>
439 PR gcov-profile/85871
440 * gcov.c (output_intermediate_file): Fix out of bounds
443 2018-09-17 Vineet Gupta <vgupta@synopsys.com>
445 * config/arc/arc.c: Object attributes for core4 not reflected
447 * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
450 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
452 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
454 2018-09-17 Martin Liska <mliska@suse.cz>
456 * doc/gcov.texi: Document new option --use-hotness-colors.
457 * gcov.c (struct source_info): Declare new field.
458 (source_info::source_info): Set default for maximum_count.
459 (print_usage): Add new -q option.
460 (process_args): Process it.
461 (accumulate_line_info): Save src->maximum_count.
462 (output_line_beginning): Make color line number if
463 flag_use_hotness_colors is set.
464 (output_line_details): Pass default argument value.
465 (output_lines): Pass src->maximum_count.
467 2018-09-17 Martin Liska <mliska@suse.cz>
469 * common/config/i386/i386-common.c (ix86_get_valid_option_values):
470 Use processor_names table.
471 * config/i386/i386.c (ix86_default_align): Use
472 processor_cost_table for alignment values.
473 (ix86_option_override_internal): Use processor_names.
474 (ix86_function_specific_print): Likewise.
475 * config/i386/i386.h (struct processor_costs):
476 Add alignment values.
477 (struct ptt): Remove and replace with const char *.
478 * config/i386/x86-tune-costs.h (struct processor_costs):
479 Declare default alignments for all costs.
481 2018-09-17 Aldy Hernandez <aldyh@redhat.com>
483 * tree-vrp.c (extract_range_from_unary_expr): Do not special case
484 symbolics or VR_VARYING ranges for ABS_EXPR.
485 * wide-int-range.cc (wide_int_range_abs): Return positive numbers
486 when range will wrap.
488 2018-09-15 Eric Botcazou <ebotcazou@adacore.com>
491 * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
492 before and after a JUMP_TABLE_DATA.
494 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
497 * dojump.c (do_compare_and_jump): Canonicalize function pointers
498 when one operand is a function pointer. Use POINTER_TYPE_P and
499 FUNC_OR_METHOD_TYPE_P.
500 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
501 * fold-const.c (build_range_check): Likewise.
502 * match.pd (simple_comparison): Likewise.
504 2018-09-14 David Malcolm <dmalcolm@redhat.com>
507 * spellcheck.c (get_edit_distance_cutoff): New function.
508 (selftest::test_edit_distance_unit_test_oneway): Rename to...
509 (selftest::test_get_edit_distance_one_way): ...this.
510 (selftest::test_get_edit_distance_unit): Rename to...
511 (selftest::test_get_edit_distance_both_ways): ...this.
512 (selftest::test_edit_distances): Move tests to this new function,
513 and test some more pairs of strings. Update for above renaming.
514 (selftest::get_old_cutoff): New function.
515 (selftest::test_get_edit_distance_cutoff): New function.
516 (selftest::assert_suggested_for): New function.
517 (ASSERT_SUGGESTED_FOR): New macro.
518 (selftest::assert_not_suggested_for): New function.
519 (ASSERT_NOT_SUGGESTED_FOR): New macro.
520 (selftest::test_suggestions): New function.
521 (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
522 tests to selftest::test_edit_distances and call it. Add calls to
523 selftest::test_get_edit_distance_cutoff and
524 selftest::test_suggestions.
525 * spellcheck.h (get_edit_distance_cutoff): New function declaration.
526 (best_match::consider): Replace hard-coded cutoff calculation with
528 (best_match::get_cutoff): New declaration.
529 (best_match::get_best_meaningful_candidate): Likewise.
531 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
533 * builtins.c (fold_builtin_strlen): Remove TODO comment.
535 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
538 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
540 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
541 terminated string literal.
543 2018-09-14 Martin Sebor <msebor@redhat.com>
545 * builtins.c (unterminated_array): Handle ARRAY_REF.
546 (expand_builtin_stpcpy_1): Detect unterminated char arrays.
547 * builtins.h (unterminated_array): Declare extern.
548 * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
550 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
553 2018-09-14 Martin Sebor <msebor@redhat.com>
554 Jeff Law <law@redhat.com>
556 * builtins.c (unterminated_array): New.
557 (expand_builtin_strcpy): Adjust.
558 (expand_builtin_strcpy_args): Detect unterminated arrays.
559 * gimple-fold.c (get_maxval_strlen): Add argument. Detect
561 * gimple-fold.h (get_maxval_strlen): Add argument.
562 (gimple_fold_builtin_strcpy): Detec unterminated arrays.
564 * gimple-fold.c (get_range_strlen): Add argument.
565 (get_maxval_strlen): Adjust.
566 * gimple-fold.h (get_range_strlen): Add argument.
568 2018-09-14 Wei Xiao <wei3.xiao@intel.com>
570 * config/i386/movdirintrin.h: Fix copyright year.
572 2018-09-14 Uros Bizjak <ubizjak@gmail.com>
574 * reg-stack.c: Include regs.h.
575 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
576 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
577 FIRST_STACK_REG, not DFmode.
578 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
579 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX.
580 (change stack): Default register mode to the reg_raw_mode of
581 FIRST_STACK_REG, not DFmode.
582 * config/i386/i386.md (*swap<mode>): Remove insn pattern.
583 (*swapxf): Rename from swapxf.
585 2018-09-14 Carl Love <cel@us.ibm.com>
587 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
588 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
590 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
593 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
596 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
599 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
601 2018-09-14 Jason Merrill <jason@redhat.com>
603 Fix --enable-gather-detailed-mem-stats.
604 * hash-table.c (hash_table_usage): Change from variable to function.
605 * hash-table.h: Adjust.
606 * Makefile.in: Add missing dependencies on hash-table.h.
608 2018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
610 PR tree-optimization/87259
612 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
613 execute_cse_reciprocals_1 has tried transforming.
615 2018-09-14 Aldy Hernandez <aldyh@redhat.com>
617 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
618 VR_VARYING for PLUS/MINUS_EXPR.
620 2018-09-14 Ilya Leoshkevich <iii@linux.ibm.com>
622 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
625 2018-09-14 Richard Biener <rguenther@suse.de>
628 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
629 bits for the merged partition.
631 2018-09-13 Martin Sebor <msebor@redhat.com>
632 Bernd Edlinger <bernd.edlinger@hotmail.de>
634 * builtins.h (c_srlen): Add argument.
635 * builtins.c (warn_string_no_nul): New function.
636 (c_strlen): Add argument and use it. Update recursive calls.
637 Pass DECL argument to string_constant to get info on non
638 terminated strings. Update *NONSTR as needed.
639 (fold_builtin_strlen): Add argument to calls to c_strlen.
640 Warn for unterminated arrays.
641 (warn_string_no_null): Add prototype.
642 * expr.c (string_constant): Update arguments. Update recursive
643 calls appropriately. Detect missing NUL terminator and outermost
644 declaration its missing in.
645 Improve checks for arrays with nonzero lower bound or elements
646 that are not a single byte. Simplify offset computation.
647 Simplify checks for non-NUL terminated strings.
648 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
649 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
651 2018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
653 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
655 * fold-const.c (c_getstr): Fix function comment. Remove unused third
656 argument. Fix range checks.
657 * fold-const.h (c_getstr): Adjust protoype.
658 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
659 string is constant but contains no NUL byte.
661 * expr.c (string_constant): Adjust function comment.
662 Remove bogus check for zero termination.
664 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
666 * varasm.c (compare_constant): Compare type size of STRING_CSTs.
667 (get_constant_size): Don't make STRING_CSTs larger than they are.
668 (check_string_literal): New check function for STRING_CSTs.
669 (output_constant): Use it.
671 2018-09-13 Eric Botcazou <ebotcazou@adacore.com>
674 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
676 2018-09-13 Richard Biener <rguenther@suse.de>
678 PR tree-optimization/87263
679 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
680 (struct unwind_state): Add max_rpo field.
681 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
682 Compute max_rpo, the max RPO number a block can be backwards reached
683 from. Re-write non-iterating mode to a RPO ordered worklist approach,
684 separating it from the iterating mode.
686 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
688 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
689 (rfs_decision): New scheduling decision.
691 2018-09-13 Richard Biener <rguenther@suse.de>
694 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
695 (vn_nary_op_insert_pieces_predicated): Do not write useless
696 valid_dominated_by_p entry outside of the allocated storage.
698 2018-09-13 Omar Sandoval <osandov@osandov.com>
699 Tom de Vries <tdevries@suse.de>
702 * dwarf2out.c (is_c): New function.
703 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
705 2018-09-13 Sam Tebbs <sam.tebbs@arm.com>
708 * config/aarch64/aarch64.md (*aarch64_bfxil):
710 * config/aarch64/constraints.md (Ulc): Define.
711 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
713 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
716 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
718 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
719 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
720 aarch64_layout_frame call.
721 (aarch64_expand_epilogue): Likewise.
722 (aarch64_initial_elimination_offset): Likewise.
723 (aarch64_get_separate_components): Likewise.
724 (aarch64_use_return_insn_p): Likewise.
725 (aarch64_layout_frame): Remove unneeded check.
727 2018-09-13 Jakub Jelinek <jakub@redhat.com>
729 * configure.ac: Only append
730 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
731 gcc_config_arguments if it was never reconfigured or last reconfigure
732 was with different arguments.
733 * configure: Regenerated.
735 2018-09-13 Jakub Jelinek <jakub@redhat.com>
736 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
739 * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
740 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
742 2018-09-13 Jakub Jelinek <jakub@redhat.com>
744 PR tree-optimization/87287
745 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
746 X % C == 0 to X % (unsigned) C == 0 optimization to ...
747 * match.pd (X % C == 0): ... here. New optimization.
749 2018-09-12 Jakub Jelinek <jakub@redhat.com>
752 * expr.h (maybe_optimize_mod_cmp): Declare.
753 * expr.c (mod_inv): New function.
754 (maybe_optimize_mod_cmp): New function.
755 (do_store_flag): Use it.
756 * cfgexpand.c (expand_gimple_cond): Likewise.
758 2018-09-09 Cesar Philippidis <cesar@codesourcery.com>
759 Julian Brown <julian@codesourcery.com>
762 * gimplify.c (gimplify_scan_omp_clauses): Set
763 target_firstprivatize_array_bases in OpenACC parallel and kernels
764 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
765 OpenACC data regions.
767 2018-09-12 Uros Bizjak <ubizjak@gmail.com>
769 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
770 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
772 2018-09-12 Richard Biener <rguenther@suse.de>
774 PR tree-optimization/87280
775 * tree-ssa-sccvn.c (process_bb): Handle the case of executable
776 edge but unreachable target.
777 (do_rpo_vn): For conservatively handling a PHI only mark
778 the backedge executable but not the block reachable.
780 2018-09-12 Richard Biener <rguenther@suse.de>
782 PR tree-optimization/87266
783 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
786 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
788 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
790 ("trunc<BFP:mode><DFP_ALL:mode>2")
791 ("trunc<DFP_ALL:mode><BFP:mode>2")
792 ("extend<BFP:mode><DFP_ALL:mode>2")
793 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
794 according to the target operand type.
796 2018-09-12 Jakub Jelinek <jakub@redhat.com>
797 Andreas Krebbel <krebbel@linux.ibm.com>
799 PR tree-optimization/86844
800 * gimple-ssa-store-merging.c
801 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
802 there are any overlapping stores in between them, make sure they are
803 also coalesced or we give up completely.
805 2018-09-12 Jakub Jelinek <jakub@redhat.com>
808 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
809 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
811 2018-09-12 Tom de Vries <tdevries@suse.de>
813 * common.opt (gdescribe-dies): Add option.
814 * dwarf2out.c (add_name_and_src_coords_attributes): Add description
815 attribute for artifical and nameless decls.
816 (dwarf2out_register_external_die): Add description attribute to
817 external reference die.
818 (add_desc_attribute): New functions.
819 (gen_subprogram_die): Add description attribute to
820 DW_TAG_call_site_parameter.
821 * tree-pretty-print.c (print_generic_expr_to_str): New function.
822 * tree-pretty-print.h (print_generic_expr_to_str): Declare.
823 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
825 (@item -gdescribe-dies): Add.
827 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
829 * tree-vrp.c (vrp_shift_undefined_p): Remove.
830 (extract_range_from_binary_expr_1: Call
831 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
832 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
835 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
837 * gimple-ssa-warn-alloca.c
838 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
839 field for ALLOCA_BOUND_*_LARGE.
841 2018-09-11 Nathan Sidwell <nathan@acm.org>
843 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
845 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
847 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
848 for clobbers. Remove obsolete comment.
850 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
852 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
853 mpxchk, mpxld and mpxst types.
854 (define_attr length_immediate): Remove all processing of mpx types.
855 (define_attr prefix_0f): Ditto.
856 (define_attr memory): Ditto.
858 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
860 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
861 (log<mode>2): Change operand 1 predicate to general_operand.
862 Extend operand 1 to XFmode and generate logxf3 insn.
863 (log10<mode>2): Change operand 1 predicate to general_operand.
864 Extend operand 1 to XFmode and generate log10xf3 insn.
865 (log2<mode>2): Change operand 1 predicate to general_operand.
866 Extend operand 1 to XFmode and generate log2xf3 insn.
867 (fyl2xp1_extend<mode>xf3_i387): Remove.
868 (log1p<mode>2): Change operand 1 predicate to general_operand.
869 Extend operand 1 to XFmode and generate log1pxf3 insn.
870 (fxtract_extend<mode>xf3_i387): Remove.
871 (logb<mode>2): Change operand 1 predicate to general_operand.
872 Extend operand 1 to XFmode and generate logbxf3 insn.
873 (ilogb<mode>2): Change operand 1 predicate to general_operand.
874 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
875 (significand<mode>2): Change operand 1 predicate to general_operand.
876 Extend operand 1 to XFmode and generate significandxf3 insn.
878 2018-09-11 Nathan Sidwell <nathan@acm.org>
880 * gcc.c (perror_with_name, pfatal_with_name): Delete.
881 (load_specs): Use fatal_error.
882 (DELETE_IF_ORDINARY, process_command): Use error.
883 (execute, run_attempt): Use fatal_error.
885 2018-09-11 Andrew Stubbs <ams@codesourcery.com>
887 * diagnostic-core.h (sorry_at): New prototype.
888 * diagnostic.c (sorry_at): New function.
890 2018-09-11 Aldy Hernandez <aldyh@redhat.com>
892 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
893 by zero as VR_UNDEFINED.
895 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
897 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
898 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
899 (<sincos>mode2): New expander.
900 (sincos_extend<mode>xf3_i387): Remove insn pattern.
901 (sincos -> sin, cos splitters): Remove splitter patterns.
902 (sincos<mode>3): Change operand 2 predicate to general_operand.
903 Extend operand 2 to XFmode and generate sincosxf3 insn.
904 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
905 Change operand 3 predicate to const1_operand.
906 (fptan_extend<mode>xf4_i387): Remove insn pattern.
907 (tanxf2): Update operands in the call to fptanxf4_i387.
908 (tan<mode>2): Change operand 1 predicate to general_operand.
909 Extend operand 1 to XFmode and generate tanxf3 insn.
910 (atan2xf3): Rename from *fpatanxf3_i387.
911 (fpatan_extend<mode>xf3_i387): Remove insn pattern.
912 (atan2xf3): Remove expander.
913 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand.
914 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
915 (atan<mode>2): Change operand 1 predicate to general_operand.
916 Extend operand 1 to XFmode and generate atanxf3 insn.
918 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
920 * config/i386/i386.md (x87/SSE constant load splitter): Use
921 memory_operand instead of nonimmediate_operand for input operand
924 2018-09-09 Uros Bizjak <ubizjak@gmail.com>
926 * config/i386/i386.md (float partial SSE register stall splitter): Move
927 splitter near its instruction pattern.
928 (float_extend partial SSE register stall splitter): Ditto.
929 (float_truncate partial SSE register stall splitter): Ditto.
931 2018-09-09 Hans-Peter Nilsson <hp@bitrange.com>
934 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
935 to speculation_safe_value_not_needed.
938 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
939 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
940 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
941 leaf_function_p, instead use has_hard_reg_initial_val.
943 2018-09-09 Nathan Sidwell <nathan@acm.org>
945 * gcc.h (pfatal_with_name): Don't declare here.
946 * gcc.c (pfatal_with_name): Make static.
948 2018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com>
950 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
953 2018-09-08 John David Anglin <danglin@gcc.gnu.org>
955 PR rtl-optimization/85458
956 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
957 priority hook to reduce the priority of EXPR.
959 2018-09-07 Uros Bizjak <ubizjak@gmail.com>
961 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
962 DImode for x87 on 32bit targets. Conditionally disable x87 modes
963 with X87_ENABLE_FLOAT. Remove preparation code.
964 (*float<SWI48:mode><MODEF:mode>2): Rename from
965 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed
966 math using "enabled" attribute.
967 (*floatdi<MODEF:mode>2_i387): Rename from
968 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and
969 enable for 32bit targets only.
970 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
972 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
973 as operand 1 predicate. Rewrite as define_insn_and_split.
974 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
976 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
978 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
979 to fallthru to FLOAT case.
981 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
984 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
985 around folding of vec_sl to handle out of range shift values.
987 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
989 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
990 Update callers to gen_fix_trunc<mode>_i387_fisttp
991 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
992 nonimmediate_operand.
993 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
994 and corresponding splitters.
995 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
996 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
997 (fix_truncdi_i387_with_temp): Remove insn pattern
998 and corresponding splitters.
999 (fix_trunc<mode>_i387): Change operand 0 predicate to
1000 nonimmediate_operand.
1001 (fix_trunc<mode>_i387_with_temp): Remove insn pattern
1002 and corresponding splitters.
1003 (*fistdi2_1): Remove.
1005 (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
1006 (lrintxfdi2): Remove expander. Reimplement as define_insn.
1007 (*fist<mode>2_1): Remove.
1008 (fist<mode>2): Ditto.
1009 (fist<mode>2_with_temp): Remove insn pattern and corresponding
1011 (lrintxf<mode>2): Remove expander. Reimplement as define_insn.
1012 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
1013 (fistdi2_<rounding>): Change operand 0 predicate to
1014 nonimmediate_operand.
1015 (fistdi2_<rounding>_with_temp): Remove insn pattern
1016 and corresponding splitters.
1017 (fist<mode>2_<rounding>): Change operand 0 predicate to
1018 nonimmediate_operand.
1019 (fist<mode>2_<rounding>_with_temp): Remove insn pattern
1020 and corresponding splitters.
1022 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
1024 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
1026 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
1029 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
1031 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
1032 early gimple folding of vec_splat().
1033 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
1034 * gimple-fold.h: Add an extern define for tree_vec_extract().
1036 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
1038 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
1039 wrappers around TREE_TYPE comparisons.
1041 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
1044 * config/s390/predicates.md: Add nonsym_memory_operand.
1045 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
1046 contains a SYMBOL_REF, load it into an intermediate pseudo.
1047 (s390_emit_compare_and_swap): Legitimize operand.
1048 * config/s390/s390.md: Use the new nonsym_memory_operand
1049 with UNSPECV_CAS patterns.
1051 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
1054 * config/s390/s390-passes.def: New file.
1055 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
1057 (make_pass_s390_early_mach): Add declaration.
1058 * config/s390/s390.c (make_pass_s390_early_mach):
1059 (s390_option_override): Remove dynamic registration.
1060 * config/s390/t-s390: Add s390-passes.def.
1062 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
1064 * config/s390/s390.c (s390_decompose_constant_pool_ref):
1065 Remove UNSPEC_LTREL_BASE check.
1066 (annotate_constant_pool_refs): Likewise.
1067 (find_constant_pool_ref): Likewise.
1068 (find_ltrel_base): Removed.
1069 (replace_ltrel_base): Removed.
1070 (s390_mainpool_finish): Remove replace_ltrel_base call.
1071 (s390_chunkify_start): Remove pending LTREL_BASE logic.
1072 (s390_chunkify_finish): Remove replace_ltrel_base call.
1073 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
1075 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
1078 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
1079 to speculation_safe_value_not_needed.
1081 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
1082 Bernd Schmidt <bernds_cb1@t-online.de>
1084 * config/nvptx/nvptx-opts.h: New file.
1085 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
1086 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
1088 (TARGET_SM35): New macro.
1089 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
1091 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
1093 (misa=): New option.
1094 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
1096 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
1098 * config/i386/i386.md (truncdfsf2): Remove expander.
1099 (truncdfsf2_with_temp): Ditto.
1100 (truncxf<mode>2): Ditto.
1101 (*truncdfsf_fast_mixed): Remove insn pattern.
1102 (*truncdfsf_fast_i387): Ditto.
1103 (*truncdfsf_mixed): Ditto.
1104 (*truncdfsf_i387): Ditto.
1105 (*truncdfsf2_i387_1): Ditto.
1106 (*truncxfsf2_mixed): Ditto.
1107 (*truncxfdf2_mixed): Ditto.
1108 (*truncxf<mode>2_i387_noop): Ditto. Update callers
1109 to call gen_truncxf<mode>2 instead.
1110 (*truncxf<mode>2_i387): Remove.
1111 (reg->reg splitters): Remove splitter pattern.
1112 (reg->mem splitters): Ditto.
1114 (truncdfsf2): New insn pattern.
1115 (truncxf<mode>2): Ditto.
1117 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1119 * tree-ssa-math-opts.c (is_mult_by): New function.
1120 (is_square_of): Use the above.
1121 (optimize_recip_sqrt): New function.
1122 (pass_cse_reciprocals::execute): Use the above.
1124 2018-09-05 Richard Biener <rguenther@suse.de>
1127 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
1128 to zero-init the emplaced vec.
1130 2018-09-05 Martin Liska <mliska@suse.cz>
1132 PR tree-optimization/87205
1133 * tree-switch-conversion.c (pass_lower_switch::execute):
1134 Group cases for switch statements.
1136 2018-09-05 Richard Biener <rguenther@suse.de>
1138 PR tree-optimization/87217
1139 * tree-ssa-sccvn.c (vuse_valueize): New.
1140 (vn_reference_lookup_pieces): Use it.
1141 (vn_reference_lookup): Likewise.
1143 2018-09-05 Nathan Sidwell <nathan@acm.org>
1146 * stor-layout.c (place_field): Scan forwards to check last
1147 bitfield when ms_bitfield_placement is in effect.
1149 2018-09-05 Richard Biener <rguenther@suse.de>
1152 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
1155 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
1156 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
1158 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
1159 * config.gcc (extra_objs): Build it.
1160 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
1162 * config/aarch64/aarch64-passes.def
1163 (pass_tag_collision_avoidance): New pass.
1164 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
1165 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
1166 (aarch64_classify_address): Remove static qualifier.
1167 (aarch64_address_info, aarch64_address_type): Move to...
1168 * config/aarch64/aarch64-protos.h: ... here.
1169 (make_pass_tag_collision_avoidance): New function.
1170 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
1173 2018-09-05 Martin Liska <mliska@suse.cz>
1175 * doc/gcov.texi: Update documentation of humar
1177 * gcov.c (format_count): Print one decimal place, it provides
1178 more fine number of situations like '1G' vs. '1.4G'.
1180 2018-09-05 Martin Liska <mliska@suse.cz>
1183 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
1184 * optc-gen.awk: Allow 'Var' for Deprecated options in order
1185 to generate a MASK value.
1187 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
1190 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
1191 if frame pointer isn't used.
1192 (compute_frame_pointer_to_fb_displacement): Likewise.
1194 2018-09-04 Jakub Jelinek <jakub@redhat.com>
1197 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
1198 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
1199 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
1200 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
1201 and OPTION_MASK_ISA_XSAVEC_UNSET.
1203 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
1205 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
1206 XOR operations in NAND case.
1208 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
1210 * wide-int-range.cc (wide_int_range_convert): New.
1211 * wide-int-range.h (wide_int_range_convert): New.
1212 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
1213 code into wide_int_range_convert.
1214 (extract_range_into_wide_ints): Do not munge anti range constants
1215 into the entire domain. Just return the range back.
1217 2018-09-04 Martin Liska <mliska@suse.cz>
1219 * genmatch.c (output_line_directive): Add new argument
1221 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
1223 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
1225 * doc/invoke.texi (Option Summary): Add whitespace.
1227 * doc/invoke.texi (Option Summary): Add -Waligned-new.
1229 2018-09-04 Richard Biener <rguenther@suse.de>
1231 PR tree-optimization/87211
1232 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
1233 backedge value we're supposed to treat as VARYING also number
1234 the PHI to VARYING in case it got a different value-number already.
1236 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
1238 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
1239 (extract_range_from_binary_expr_1): Do not call
1240 vrp_can_optimize_bit_op.
1241 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
1243 (wide_int_range_get_mask_and_bounds): New.
1244 (wide_int_range_optimize_bit_op): New.
1245 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
1246 (wide_int_range_bit_and): Same.
1247 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
1248 (wide_int_range_optimize_bit_op): New.
1249 (wide_int_range_get_mask_and_bounds): New.
1251 2018-09-04 Richard Biener <rguenther@suse.de>
1253 PR tree-optimization/87176
1254 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
1255 variable. When value-numbering a virtual PHI node make sure
1256 to not value-number to the backedge value.
1258 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
1260 * doc/extend.texi (Long Long, Hex Floats): Document support for
1261 long long and hex floats in more recent versions of ISO C++.
1263 2018-09-03 Richard Biener <rguenther@suse.de>
1265 PR tree-optimization/87177
1266 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
1269 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
1271 * bb-reorder.c (edge_order): Convert to C-qsort-style
1272 tri-state comparator.
1273 (reorder_basic_blocks_simple): Change std::stable_sort to gcc_stablesort.
1275 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
1277 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
1278 tri-state comparator.
1279 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
1281 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
1283 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
1284 (mergesort): ... here as maximum count for using netsort.
1285 (gcc_qsort): Set nlim to 3 if stable sort is requested.
1286 (gcc_stablesort): New.
1287 * system.h (gcc_stablesort): Declare.
1289 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
1291 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
1292 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
1293 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
1295 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
1297 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
1298 lxsdx and stxsdx alternatives.
1299 (*mov<mode>_hardfloat64): Ditto.
1300 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
1302 2018-09-03 Richard Biener <rguenther@suse.de>
1304 PR tree-optimization/87200
1305 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
1308 2018-09-03 Martin Liska <mliska@suse.cz>
1310 PR tree-optimization/87201
1311 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
1312 Fix parenthesis in an expression.
1314 2018-09-03 Richard Biener <rguenther@suse.de>
1316 PR tree-optimization/87197
1317 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
1318 visited. CSE the VN_INFO hashtable lookup.
1320 PR tree-optimization/87169
1321 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
1322 iterating make sure there's no extra backedges from irreducible
1323 regions feeding the header. Mark the destination block
1326 2018-09-03 Martin Liska <mliska@suse.cz>
1329 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
1330 * common/common-targhooks.c (default_get_valid_option_values):
1332 * common/common-targhooks.h (default_get_valid_option_values):
1334 * common/config/i386/i386-common.c: Move processor_target_table
1336 (ix86_get_valid_option_values): New function.
1337 (TARGET_GET_VALID_OPTION_VALUES): New macro.
1338 * config/i386/i386.c (struct ptt): Move to i386-common.c.
1339 (PTA_*): Move all defined masks into i386-common.c.
1340 (ix86_function_specific_restore): Use new processor_cost_table.
1341 * config/i386/i386.h (struct ptt): Moved from i386.c.
1342 (struct pta): Likewise.
1343 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
1344 * doc/tm.texi.in: Likewise.
1345 * opt-suggestions.c (option_proposer::suggest_option):
1346 Pass prefix to build_option_suggestions.
1347 (option_proposer::get_completions): Likewise.
1348 (option_proposer::build_option_suggestions): Use the new target
1350 * opts.c (struct option_help_tuple): New struct.
1351 (print_filtered_help): Use the new target hook.
1353 2018-09-03 Martin Liska <mliska@suse.cz>
1356 * predict.c (set_even_probabilities): Add likely_edges
1357 argument and handle cases where we have precisely one
1359 (combine_predictions_for_bb): Catch also likely_edges.
1360 (tree_predict_by_opcode): Handle gswitch statements.
1361 * tree-cfg.h (find_case_label_for_value): New declaration.
1362 (find_taken_edge_switch_expr): Likewise.
1363 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
1364 Find pivot in decision tree based on probabily, not by number of
1367 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
1369 * doc/standards.texi (Standards): Update Objective-C reference.
1371 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
1373 * doc/install.texi (Prerequisites): Update link for MPC.
1375 2018-09-01 Michael Matz <matz@suse.de>
1377 PR tree-optimization/87074
1378 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
1379 PHIs for outer-loop uses.
1381 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
1383 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
1384 * doc/invoke.texi (C Dialect Options): Ditto.
1386 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
1388 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
1390 2018-08-31 Richard Biener <rguenther@suse.de>
1392 PR tree-optimization/87168
1393 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
1394 (rpo_elim::eliminate_avail): When OP was not visited it must
1397 2018-08-31 David Malcolm <dmalcolm@redhat.com>
1399 * tree-vrp.c (copy_value_range): Convert param "from" from
1400 "value_range *" to "const value_range *".
1401 (range_is_null): Likewise for param "vr".
1402 (range_int_cst_p): Likewise.
1403 (range_int_cst_singleton_p): Likewise.
1404 (symbolic_range_p): Likewise.
1405 (value_ranges_intersect_p): Likewise for both params.
1406 (value_range_nonnegative_p): Likewise for param "vr".
1407 (value_range_constant_singleton): Likewise.
1408 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
1409 (extract_range_into_wide_ints): Likewise for param "vr".
1410 (extract_range_from_multiplicative_op): Likewise for params "vr0"
1412 (vrp_can_optimize_bit_op): Likewise.
1413 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
1415 (extract_range_from_unary_expr): Likewise.
1416 (debug_value_range): Likewise for param "vr".
1417 (value_range::dump): Add "const" qualifier.
1418 (vrp_prop::check_array_ref): Convert local "vr" from
1419 "value_range *" to "const value_range *".
1420 (vrp_prop::check_mem_ref): Likewise.
1421 (vrp_prop::visit_stmt): Likewise for local "old_vr".
1422 (vrp_intersect_ranges_1): Likewise for param "vr_1".
1423 (vrp_intersect_ranges): Likewise.
1424 (simplify_stmt_for_jump_threading): Likewise for local "vr".
1425 (vrp_prop::vrp_finalize): Likewise.
1426 * tree-vrp.h (value_range::dump): Add "const" qualifier.
1427 (vrp_intersect_ranges): Add "const" qualifier to params as above.
1428 (extract_range_from_unary_expr): Likewise.
1429 (value_range_constant_singleton): Likewise.
1430 (symbolic_range_p): Likewise.
1431 (copy_value_range): Likewise.
1432 (extract_range_from_binary_expr_1): Likewise.
1433 (range_int_cst_p): Likewise.
1434 (vrp_set_zero_nonzero_bits): Likewise.
1435 (range_int_cst_singleton_p): Likewise.
1437 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
1439 * config/aarch64/arm_neon.h (vabsd_s64): New.
1440 (vnegd_s64): Likewise.
1442 2018-08-31 Martin Jambor <mjambor@suse.cz>
1444 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
1446 2018-08-31 Martin Liska <mliska@suse.cz>
1448 * ipa-icf.c (sem_item::add_type): Use
1449 sem_item::m_type_hash_cache.
1450 * ipa-icf.h: Move the cache from sem_item_optimizer
1453 2018-08-31 Nathan Sidwell <nathan@acm.org>
1455 * doc/extend.texi (Backwards Compatibility): Remove implicit
1456 extern C leeway of () being (...).
1458 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1460 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
1462 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
1466 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
1468 2018-08-31 Jakub Jelinek <jakub@redhat.com>
1471 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
1472 gen_int_mode. Formatting fixes.
1474 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
1476 * target.def (custom_function_descriptors): Improve documentation.
1477 * doc/tm.texi.in (Trampolines): Expand discussion of function
1478 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
1479 beginning of the section.
1480 * doc/tm.texi: Regenerated.
1482 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
1484 * cfg.h (class auto_edge_flag): Spell out the template-id of the
1485 base class in the initializer list. This is a workaround for
1486 building with older GCC.
1487 (class auto_bb_flag): Likewise.
1489 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
1491 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
1492 (altivec_vcmpequ<VI_char>_p): Remove star.
1493 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
1495 (expand_strncmp_vec_sequence): New function.
1496 (emit_final_str_compare_vec): New function.
1497 (expand_strn_compare): Add support for vector strncmp.
1498 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
1499 length specification to bytes.
1500 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
1501 (vcmpnezb_p): New pattern.
1502 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
1503 for option -mstring-compare-inline-limit.
1505 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
1507 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
1508 (PTA_SKYLAKE): Add PTA_AES.
1509 (PTA_GOLDMONT): Likewise.
1511 2018-08-29 Jan Hubicka <jh@suse.cz>
1514 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
1515 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
1517 2018-08-29 Jan Hubicka <jh@suse.cz>
1519 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
1521 (hash_tree): Do not visit TYPE_STUB_DECL.
1522 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
1523 stream TYPE_STUB_DECL.
1524 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
1525 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
1526 after free_lang_data.
1527 (type_in_anonymous_namespace_p): Likewise.
1529 2018-08-29 Jan Hubicka <jh@suse.cz>
1531 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
1532 comment that it has to be even number.
1533 (class sreal): Change m_sig type to int32_t.
1534 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
1535 int64_t for temporary calculations.
1536 (sreal_verify_basics): Drop one bit from minimum and maximum.
1538 2018-08-30 Richard Biener <rguenther@suse.de>
1540 PR tree-optimization/87147
1541 * tree-ssa-sccvn.c (SSA_VISITED): New function.
1542 (visit_phi): When the degenerate result is from the backedge and
1543 we didn't visit its definition yet drop to VARYING.
1544 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
1546 2018-08-29 Jan Hubicka <jh@suse.cz>
1548 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
1550 (hash_tree): Likewise.
1552 2018-08-29 Jan Hubicka <jh@suse.cz>
1554 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
1555 and TYPE_NEXT_REF_TO.
1557 2018-08-29 Jan Hubicka <jh@suse.cz>
1559 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
1560 comment that it has to be even number.
1561 (class sreal): Change m_sig type to int32_t.
1562 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
1563 int64_t for temporary calculations.
1564 (sreal_verify_basics): Drop one bit from minimum and maximum.
1566 2018-08-30 Tamar Christina <tamar.christina@arm.com>
1568 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
1570 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
1573 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
1574 if to_add is negative.
1576 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
1579 * builtins.c (c_strlen): Improve range checks.
1581 2018-08-29 Martin Sebor <msebor@redhat.com>
1582 Jeff Law <law@redhat.com>
1584 PR tree-optimization/86714
1585 PR tree-optimization/86711
1586 * builtins.c (c_strlen): Add arguments to call to string_constant.
1587 * expr.c (string_constant): Add argument. Detect missing nul
1588 terminator and outermost declaration it's missing in.
1589 * expr.h (string_constant): Add argument.
1590 * fold-const.c (read_from_constant_string): Add arguments to call to
1592 (c_getstr): Likewise.
1593 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
1595 * tree-ssa-strlen.c (get_stridx): Likewise.
1597 2018-08-29 Jan Hubicka <jh@suse.cz>
1599 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
1600 Do not stream DECL_VINDEX.
1601 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
1602 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
1603 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
1605 2018-08-29 Richard Biener <rguenther@suse.de>
1607 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
1608 virtual operands that are not default defs to honor region
1610 (rpo_vn_valueize): Remove ineffective code here.
1612 2018-08-29 Richard Biener <rguenther@suse.de>
1614 PR tree-optimization/87132
1615 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
1616 when skipping defs reachable over backedges.
1618 2018-08-29 Richard Biener <rguenther@suse.de>
1620 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
1621 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
1622 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
1623 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
1624 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
1625 (vn_reference_lookup_3): Remove use of const_parms.
1626 (free_rpo_vn): Do not free const_parms.
1627 (do_rpo_vn): Do not call init_const_parms.
1628 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
1629 SSA_NAME_POINTS_TO_READONLY_MEMORY.
1630 (call_may_clobber_ref_p_1): Likewise.
1632 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
1635 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
1637 (-ftree-scev-cprop): Document.
1639 2018-08-29 Jan Hubicka <jh@suse.cz>
1641 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
1643 (sreal constructor): Update.
1644 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
1645 sreal:operator/): Update.
1647 2018-08-29 Martin Liska <mliska@suse.cz>
1649 * tree-switch-conversion.c (switch_conversion::expand):
1650 Strenghten assumption about gswitch statements.
1652 2018-08-29 Richard Biener <rguenther@suse.de>
1654 PR tree-optimization/87117
1655 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
1656 re-value-number released SSA VDEFs.
1658 2018-08-29 Richard Biener <rguenther@suse.de>
1660 PR tree-optimization/87126
1661 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
1663 2018-08-28 Jim Wilson <jimw@sifive.com>
1665 * config/riscv/pic.md: Rewrite.
1666 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
1668 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
1669 (SOFTF, default_load, softload, softstore): New.
1671 2018-08-28 Jeff Law <law@redhat.com>
1673 * fold-const.c (fold_binary_loc): Remove recently added assert.
1675 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
1677 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
1678 to OP parmeter of generated function.
1680 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
1682 PR tree-optimization/87009
1683 * match.pd: Add boolean optimizations.
1685 2018-08-28 Martin Sebor <msebor@redhat.com>
1688 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
1689 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
1690 (pass_walloca::gate): Use it.
1691 (alloca_call_type): Same.
1692 (pass_walloca::execute): Same.
1693 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
1695 2018-08-28 David Malcolm <dmalcolm@redhat.com>
1697 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
1698 GCC_VERSION for usage of "__gcc_dump_printf__" format from
1701 2018-08-28 Richard Biener <rguenther@suse.de>
1703 PR tree-optimization/87124
1704 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
1705 constants before looking up avail.
1707 2018-08-28 Jakub Jelinek <jakub@redhat.com>
1710 * calls.c (maybe_warn_nonstring_arg): Punt early if
1711 warn_stringop_overflow is zero. Don't call get_range_strlen
1712 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
1713 Swap comparison operands to have constants on rhs. Only use
1714 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
1715 increment lenrng[0].
1717 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
1719 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
1720 use of tree_to_shwi. Remove duplicated test for the size being
1721 a whole number of bytes.
1723 2018-08-28 Richard Biener <rguenther@suse.de>
1725 PR tree-optimization/87117
1726 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
1727 Handle removed stmt without LHS (GIMPLE_NOP).
1729 2018-08-28 Richard Biener <rguenther@suse.de>
1731 PR tree-optimization/87117
1732 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
1733 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
1735 2018-08-28 Richard Biener <rguenther@suse.de>
1737 PR tree-optimization/87117
1738 * tree-ssa-pre.c (compute_avail): Do not make expressions
1739 with predicated values available.
1740 (get_expr_value_id): Assert we do not run into predicated value
1743 2018-08-28 Richard Biener <rguenther@suse.de>
1745 PR tree-optimization/87117
1746 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
1747 get virtual operands.
1748 (get_expr_operands): Handle STRING_CST like other decls.
1750 2018-08-28 Martin Liska <mliska@suse.cz>
1752 * tree.h: Update documentation of fndecl_built_in_p
1756 2018-08-27 Jeff Law <law@redhat.com>
1757 PR tree-optimization/87110
1758 * tree-ssa-dse.c (compute_trims): Handle non-constant
1761 2018-08-27 Martin Sebor <msebor@redhat.com>
1763 PR tree-optimization/86914
1764 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
1766 2018-08-27 Martin Sebor <msebor@redhat.com>
1768 PR tree-optimization/87112
1769 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
1770 the type of the bound argument.
1772 2018-08-27 Jeff Law <law@redhat.com>
1774 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
1775 type does not have a TYPE_SIZE_UNIT.
1777 2018-08-27 Steve Ellcey <sellcey@cavium.com>
1779 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
1780 with include of backend.h.
1782 2018-08-27 Richard Biener <rguenther@suse.de>
1784 PR tree-optimization/86927
1785 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
1786 use const cond reduction code.
1788 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
1790 PR tree-optimization/85758
1791 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
1793 2018-08-27 David Malcolm <dmalcolm@redhat.com>
1796 * diagnostic-show-locus.c (class layout_range): Update for
1797 conversion of show_caret_p to a tri-state.
1798 (layout_range::layout_range): Likewise.
1799 (make_range): Likewise.
1800 (layout::maybe_add_location_range): Likewise.
1801 (layout::should_print_annotation_line_p): Don't show annotation
1802 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
1803 (layout::get_state_at_point): Update for conversion of
1804 show_caret_p to a tri-state. Bail out early for
1805 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
1806 underlining or source colorization.
1807 (gcc_rich_location::add_location_if_nearby): Update for conversion
1808 of show_caret_p to a tri-state.
1809 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
1810 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
1812 (selftest::test_one_liner_labels): Likewise.
1813 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
1814 conversion of show_caret_p to a tri-state.
1815 * pretty-print.c (text_info::set_location): Likewise.
1816 * pretty-print.h (text_info::set_location): Likewise.
1817 * substring-locations.c (format_warning_n_va): Likewise.
1818 * tree-diagnostic.c (default_tree_printer): Likewise.
1819 * tree-pretty-print.c (newline_and_indent): Likewise.
1821 2018-08-27 David Malcolm <dmalcolm@redhat.com>
1824 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
1825 line above for line-insertion fix-it hints.
1826 (selftest::test_fixit_insert_containing_newline): Update the
1827 expected results, and add a test with line-numbering enabled.
1829 2018-08-27 Martin Liska <mliska@suse.cz>
1832 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
1833 params with DECL_HAS_VALUE_EXPR_P.
1835 2018-08-27 Martin Liska <mliska@suse.cz>
1837 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
1838 selected expansion strategy.
1840 2018-08-27 Martin Liska <mliska@suse.cz>
1842 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
1843 * builtins.c (is_builtin_fn): Likewise.
1844 * attribs.c (diag_attr_exclusions): Use new function
1845 fndecl_built_in_p and remove check for FUNCTION_DECL if
1847 (builtin_mathfn_code): Likewise.
1848 (fold_builtin_expect): Likewise.
1849 (fold_call_expr): Likewise.
1850 (fold_builtin_call_array): Likewise.
1851 (fold_call_stmt): Likewise.
1852 (set_builtin_user_assembler_name): Likewise.
1853 (is_simple_builtin): Likewise.
1854 * calls.c (gimple_alloca_call_p): Likewise.
1855 (maybe_warn_nonstring_arg): Likewise.
1856 * cfgexpand.c (expand_call_stmt): Likewise.
1857 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
1858 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
1859 (cgraph_node::verify_node): Likewise.
1860 * cgraphclones.c (build_function_decl_skip_args): Likewise.
1861 (cgraph_node::create_clone): Likewise.
1862 * config/arm/arm.c (arm_insert_attributes): Likewise.
1863 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
1864 * dse.c (scan_insn): Likewise.
1865 * expr.c (expand_expr_real_1): Likewise.
1866 * fold-const.c (operand_equal_p): Likewise.
1867 (fold_binary_loc): Likewise.
1868 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
1869 * gimple-low.c (lower_stmt): Likewise.
1870 * gimple-pretty-print.c (dump_gimple_call): Likewise.
1871 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Likewise.
1872 * gimple.c (gimple_build_call_from_tree): Likewise.
1873 (gimple_call_builtin_p): Likewise.
1874 (gimple_call_combined_fn): Likewise.
1875 * gimplify.c (gimplify_call_expr): Likewise.
1876 (gimple_boolify): Likewise.
1877 (gimplify_modify_expr): Likewise.
1878 (gimplify_addr_expr): Likewise.
1879 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
1880 * ipa-cp.c (determine_versionability): Likewise.
1881 * ipa-fnsummary.c (compute_fn_summary): Likewise.
1882 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
1883 * ipa-split.c (visit_bb): Likewise.
1884 (split_function): Likewise.
1885 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1886 * lto-cgraph.c (input_node): Likewise.
1887 * lto-streamer-out.c (write_symbol): Likewise.
1888 * omp-low.c (setjmp_or_longjmp_p): Likewise.
1889 (lower_omp_1): Likewise.
1890 * predict.c (strip_predict_hints): Likewise.
1891 * print-tree.c (print_node): Likewise.
1892 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
1893 * trans-mem.c (is_tm_irrevocable): Likewise.
1894 (is_tm_load): Likewise.
1895 (is_tm_simple_load): Likewise.
1896 (is_tm_store): Likewise.
1897 (is_tm_simple_store): Likewise.
1898 (is_tm_abort): Likewise.
1899 (tm_region_init_1): Likewise.
1900 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
1901 * tree-cfg.c (verify_gimple_call): Likewise.
1902 (move_stmt_r): Likewise.
1903 (stmt_can_terminate_bb_p): Likewise.
1904 * tree-eh.c (lower_eh_constructs_2): Likewise.
1905 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
1906 * tree-inline.c (remap_gimple_stmt): Likewise.
1907 (copy_bb): Likewise.
1908 (estimate_num_insns): Likewise.
1909 (fold_marked_statements): Likewise.
1910 * tree-sra.c (scan_function): Likewise.
1911 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
1912 (optimize_stack_restore): Likewise.
1913 (pass_fold_builtins::execute): Likewise.
1914 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
1915 (mark_all_reaching_defs_necessary_1): Likewise.
1916 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
1917 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
1918 (pass_forwprop::execute): Likewise.
1919 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1920 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
1921 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
1922 * tree-ssa-strlen.c (get_string_length): Likewise.
1923 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
1924 (find_func_aliases_for_call): Likewise.
1925 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
1926 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
1927 * tree-tailcall.c (find_tail_calls): Likewise.
1928 * tree.c (need_assembler_name_p): Likewise.
1929 (free_lang_data_in_decl): Likewise.
1930 (get_call_combined_fn): Likewise.
1931 * ubsan.c (is_ubsan_builtin_p): Likewise.
1932 * varasm.c (incorporeal_function_p): Likewise.
1933 * tree.h (DECL_BUILT_IN): Remove and replace with
1935 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
1936 (fndecl_built_in_p): New.
1938 2018-08-27 Martin Liska <mliska@suse.cz>
1940 PR tree-optimization/86847
1941 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
1942 Dump also subtree probability.
1943 (switch_decision_tree::do_jump_if_equal): New function.
1944 (switch_decision_tree::emit_case_nodes): Handle special
1945 situations in balanced tree that can be emitted much simpler.
1946 Fix calculation of probabilities that happen in tree expansion.
1947 * tree-switch-conversion.h (struct cluster): Add
1949 (struct simple_cluster): Likewise.
1950 (struct case_tree_node): Add new function has_child.
1951 (do_jump_if_equal): New.
1953 2018-08-27 Martin Liska <mliska@suse.cz>
1955 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
1956 Add new argument to bit_test_cluster constructor.
1957 (bit_test_cluster::emit): Set bits really number of values
1959 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
1960 probability argument.
1961 * tree-switch-conversion.h (struct bit_test_cluster):
1962 Add m_handles_entire_switch.
1964 2018-08-27 Martin Liska <mliska@suse.cz>
1966 PR tree-optimization/86702
1967 * tree-switch-conversion.c (jump_table_cluster::emit):
1968 Make probabilities even for values in jump table
1969 according to number of cases handled.
1970 (switch_decision_tree::compute_cases_per_edge): Pass
1971 argument to reset_out_edges_aux function.
1972 (switch_decision_tree::analyze_switch_statement): Likewise.
1973 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
1976 2018-08-27 Martin Liska <mliska@suse.cz>
1978 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
1979 cfun argument explicitly.
1980 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
1981 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
1982 function gimple_switch_default_bb.
1983 (convert_switch_statements):
1985 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
1986 * stmt.c (label_to_block_fn): Use label_to_block and pass
1987 cfun argument explicitly and use gimple_switch_label_bb.
1988 (expand_case): Likewise.
1989 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
1990 cfun argument explicitly. Likewise.
1991 (make_edges_bb): Likewise.
1992 (make_cond_expr_edges): Likewise.
1993 (get_cases_for_edge): Likewise.
1994 (make_gimple_switch_edges): Likewise.
1995 (label_to_block_fn): Likewise.
1996 (label_to_block): Likewise.
1997 (make_goto_expr_edges): Likewise.
1998 (make_gimple_asm_edges): Likewise.
1999 (main_block_label): Likewise.
2000 (group_case_labels_stmt): Likewise.
2001 (find_taken_edge_computed_goto): Likewise.
2002 (find_taken_edge_switch_expr): Likewise.
2003 (gimple_verify_flow_info): Likewise.
2004 (gimple_redirect_edge_and_branch): Likewise.
2005 (gimple_switch_label_bb): New function.
2006 (gimple_switch_default_bb): Likewise.
2007 (gimple_switch_edge): Likewise.
2008 (gimple_switch_default_edge): Likewise.
2009 * tree-cfg.h (label_to_block_fn): Remove and replace ...
2010 (label_to_block): ... with this.
2011 (gimple_switch_label_bb): New.
2012 (gimple_switch_default_bb): Likewise.
2013 (gimple_switch_edge): Likewise.
2014 (gimple_switch_default_edge): Likewise.
2015 * tree-cfgcleanup.c (convert_single_case_switch): Use
2016 new gimple functions and pass new argument to label_to_block.
2017 (cleanup_control_flow_bb):
2018 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
2019 cfun argument explicitly.
2020 (make_eh_edges): Likewise.
2021 (redirect_eh_dispatch_edge): Likewise.
2022 (lower_resx): Likewise.
2023 (lower_eh_dispatch): Likewise.
2024 (maybe_remove_unreachable_handlers): Likewise.
2025 (unsplit_eh): Likewise.
2026 (cleanup_empty_eh): Likewise.
2027 (verify_eh_edges): Likewise.
2028 (verify_eh_dispatch_edge): Likewise.
2029 * tree-ssa-dom.c (record_edge_info): Likewise.
2030 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
2031 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
2032 (thread_through_normal_block): Likewise.
2033 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
2034 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
2035 * tree-switch-conversion.c (switch_conversion::collect): Use new
2037 (switch_conversion::check_final_bb): Likewise.
2038 (switch_conversion::gather_default_values): Pass new argument
2040 (switch_conversion::build_constructors): Likewise.
2041 (switch_decision_tree::compute_cases_per_edge): Use new
2042 gimple_switch_edge function.
2043 (switch_decision_tree::analyze_switch_statement): Pass new argument
2045 (switch_decision_tree::try_switch_expansion): Use
2046 gimple_switch_default_edge.
2047 * tree-vrp.c (find_switch_asserts): Pass new argument
2049 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
2050 (vr_values::simplify_switch_using_ranges): Likewise.
2052 2018-08-27 Richard Biener <rguenther@suse.de>
2054 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
2055 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
2057 * tree-ssa-sccvn.h (struct vn_pval): New structure.
2058 (struct vn_nary_op_s): Add unwind_to member. Add
2059 predicated_values flag and put result into a union together
2060 with a linked list of vn_pval.
2061 (struct vn_ssa_aux): Add name member to make maintaining
2062 a map of SSA name to vn_ssa_aux possible. Remove no longer
2063 needed info, dfsnum, low, visited, on_sccstack, use_processed
2064 and range_info_anti_range_p members.
2065 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
2066 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
2068 (vn_valueize): New global.
2069 (vn_context_bb): Likewise.
2070 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
2071 VN_INFO_PTR_INFO): Remove.
2072 * tree-ssa-sccvn.c: ... (rewrite)
2073 (pass_fre::execute): For -O2+ initialize loops and run
2074 RPO VN in optimistic mode (iterating). For -O1 and -Og
2075 run RPO VN in non-optimistic mode.
2076 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
2077 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
2078 * doc/invoke.texi (sccvn-max-scc-size): Remove.
2079 (rpo-vn-max-loop-depth): Document.
2080 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
2081 when valuezing the VUSE signals we walked out of the region.
2082 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
2083 (phi_translate): Set VN context block to use for availability
2085 (compute_avail): Likewise.
2086 (pre_valueize): New function.
2087 (pass_pre::execute): Adjust to the RPO VN API.
2089 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
2090 (propagate_constants_for_unrolling): Remove.
2091 (tree_unroll_loops_completely): Perform value-numbering
2092 on the unrolled bodies loop parent.
2094 2018-08-27 Richard Biener <rguenther@suse.de>
2096 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
2097 for partial antic compute.
2099 2018-08-27 Jakub Jelinek <jakub@redhat.com>
2101 PR rtl-optimization/87065
2102 * combine.c (simplify_if_then_else): Formatting fix.
2103 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
2105 (known_cond): Don't return const_true_rtx for vector modes. Use
2106 CONST0_RTX instead of const0_rtx. Formatting fixes.
2108 2018-08-27 Martin Liska <mliska@suse.cz>
2110 PR gcov-profile/87069
2111 * gcov.c (process_file): Record files already processed
2112 and warn about a file being processed multiple times.
2114 2018-08-27 Martin Liska <mliska@suse.cz>
2117 * config/aarch64/aarch64.c (aarch64_override_options_internal):
2118 Set default values for x_aarch64_*_string strings.
2119 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
2120 prefix. For -mabi do not print '=ABI' in help and use
2121 <option_value> format for -msve-vector-bits and -moverride
2124 2018-08-26 Jeff Law <law@redhat.com>
2126 * config/mips/frame-header-opt.c: Include "backend.h" rather than
2129 2018-08-26 Marek Polacek <polacek@redhat.com>
2131 PR c++/87029, Implement -Wredundant-move.
2132 * doc/invoke.texi: Document -Wredundant-move.
2134 2018-08-25 Martin Sebor <msebor@redhat.com>
2136 PR tree-optimization/87059
2137 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
2138 to the same type as the other.
2139 * fold-const.c (fold_binary_loc): Assert expectation.
2141 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
2143 * config/darwin.c (machopic_legitimize_pic_address): Clean up
2144 extraneous parentheses, dead code section and formatting.
2146 2018-08-24 David Malcolm <dmalcolm@redhat.com>
2149 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
2150 wide enough for jumps in the line-numbering to be visible.
2151 (layout::print_gap_in_line_numbering): New member function.
2152 (layout::calculate_line_spans): When using line numbering, merge
2153 line spans that are only 1 line apart.
2154 (diagnostic_show_locus): When printing line numbers, show gaps in
2155 line numbering directly, rather than printing headers.
2156 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
2157 line-numbering with multiple line spans.
2158 (selftest::test_fixit_insert_containing_newline_2): Add test of
2159 line-numbering, in which the spans are close enough to be merged.
2161 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
2163 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
2164 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
2165 * tree-vrp.c (range_is_nonnull): Remove.
2166 (range_includes_zero_p): Accept value_range instead of min/max.
2167 (extract_range_from_binary_expr_1): Do not early bail on
2169 Use range_includes_zero_p instead of range_is_nonnull.
2170 (extract_range_from_unary_expr): Use range_includes_zero_p instead
2171 of range_is_nonnull.
2172 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
2173 special case VR_ANTI_RANGE.
2174 (vrp_finalize): Same.
2175 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
2177 (range_is_nonnull): Remove.
2178 * vr-values.c (vrp_stmt_computes_nonzero): Use
2179 range_includes_zero_p instead of range_is_nonnull.
2180 (extract_range_basic): Pass value_range to range_includes_zero_p
2181 instead of range_is_nonnull.
2183 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
2185 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
2186 * emit-rtl.h (rtl_data): Remove return_bnd.
2187 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
2188 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
2189 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
2190 (POINTER_BOUNDS_MODE): Remove definition.
2191 (make_pointer_bounds_mode): Remove.
2192 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
2193 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
2194 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
2195 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
2196 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
2197 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
2198 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
2200 * config/i386/i386-modes.def (BND32, BND64): Remove.
2201 * config/i386/i386.c (dbx_register_map): Remove bound registers.
2202 (dbx64_register_map): Ditto.
2203 (svr4_dbx_register_map): Ditto.
2204 (indirect_thunk_bnd_needed): Remove.
2205 (indirect_thunks_bnd_used): Ditto.
2206 (indirect_return_bnd_needed): Ditto.
2207 (indirect_return_via_cx_bnd): Ditto.
2208 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
2209 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
2210 (output_indirect_thunk): Ditto. Remove need_prefix argument.
2211 (output_indirect_thunk_function): Remove handling of
2212 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
2213 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
2214 (ix86_save_reg): Remove handling of crtl->return_bnd.
2215 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
2216 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
2217 and UNSPEC_BNDLX_ADDR.
2218 (ix86_output_indirect_branch_via_reg): Remove handling of
2219 indirect_thunk_prefix_bnd.
2220 (ix86_output_indirect_branch_via_push): Ditto.
2221 (ix86_output_function_return): Ditto.
2222 (ix86_output_indirect_function_return): Ditto.
2223 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
2224 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
2225 (CALL_USED_REGISTERS): Ditto.
2226 (REG_ALLOC_ORDER): Update for removal of bound registers.
2227 (HI_REGISTER_NAMES): Ditto.
2228 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
2229 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
2230 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
2231 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
2232 (FIRST_PSEUDO_REG): Update.
2233 (BND): Remove mode iterator.
2234 * config/i386/predicates.md (bnd_mem_operator): Remove.
2236 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
2238 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
2241 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
2243 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
2244 the case in which the permute needs only a single element and
2245 repeats for every vector of the result. Extend that case to
2246 handle variable-length vectors.
2247 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
2249 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
2252 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
2253 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
2255 2018-08-24 Richard Biener <rguenther@suse.de>
2257 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
2258 bb_flags_allocated members.
2259 (auto_flag): New RAII class for allocating flags.
2260 (auto_edge_flag): New RAII class for allocating edge flags.
2261 (auto_bb_flag): New RAII class for allocating bb flags.
2262 * cfgloop.c (verify_loop_structure): Allocate temporary edge
2264 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
2265 in favor of temporarily allocated BB flag.
2266 * hsa-brig.c: Re-order includes.
2267 * hsa-dump.c: Likewise.
2268 * hsa-regalloc.c: Likewise.
2269 * print-rtl.c: Likewise.
2270 * profile-count.c: Likewise.
2272 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
2275 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
2278 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
2281 * wide-int-range.cc (wide_int_range_div): Do not ignore result
2282 from wide_int_range_multiplicative_op.
2284 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2286 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
2289 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
2291 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
2294 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
2296 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
2297 full GENERIC expression used for replacement.
2299 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
2301 * tree-vrp.c (abs_extent_range): Remove.
2302 (extract_range_into_wide_ints): Pass wide ints by reference.
2303 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
2304 Pass wide ints by reference in all calls to
2305 extract_range_into_wide_ints.
2306 * wide-int-range.cc (wide_int_range_div): New.
2307 * wide-int-range.h (wide_int_range_div): New.
2308 (wide_int_range_includes_zero_p): New.
2309 (wide_int_range_zero_p): New.
2311 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
2313 * config/aarch64/aarch64.md (arches): New enum.
2314 (arch): New enum attr.
2315 (arch_enabled): New attr.
2316 (enabled): Now uses arch_enabled only.
2317 (simd, sve, fp16): Removed attribute.
2318 (fp): Attr now defined in terms of 'arch'.
2319 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
2320 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
2321 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
2322 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
2323 attributes into 'arch'.
2324 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
2325 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
2326 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
2327 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
2328 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
2329 'simd' attribute into 'arch'.
2330 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
2331 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
2332 Convert use of 'fp' attribute to 'arch'.
2333 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
2334 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
2336 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
2337 (different modes) Merge 'fp' and 'simd' into 'arch'.
2338 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
2341 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
2343 PR rtl-optimization/87026
2344 * expmed.c (canonicalize_comparison): If we can no longer create
2345 pseudoregisters, don't.
2347 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
2350 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
2352 * config/arm/arm.c (speculation_barrier_libfunc): New static
2354 (arm_init_libfuncs): Initialize it.
2355 (arm_emit_speculation_barrier): New function.
2356 * config/arm/arm.md (speculation_barrier): Call
2357 arm_emit_speculation_barrier for architectures that do not have
2359 (speculation_barrier_insn): Only match on Armv7 or later.
2361 2018-08-23 Richard Biener <rguenther@suse.de>
2364 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
2367 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
2369 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
2370 of single-vector TBLs.
2371 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
2374 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
2377 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
2378 aarch64_evpc_tbl guard.
2380 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
2382 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
2385 2018-08-22 Martin Sebor <msebor@redhat.com>
2388 * tree-pretty-print.c (pretty_print_string): Add argument.
2389 (dump_generic_node): Call to pretty_print_string with string size.
2391 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
2393 PR rtl-optimization/86771
2394 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
2395 of two SETs into those two SETs, one to be placed at i2, if that SETs
2396 destination is modified between i2 and i3.
2398 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2400 PR tree-optimization/86725
2401 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
2403 (vect_analyze_scalar_cycles_1): Check it.
2405 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2407 PR tree-optimization/86725
2408 * tree-vect-loop.c (vect_is_simple_reduction): When treating
2409 an outer loop phi as a double reduction, make sure that the
2410 single user of the phi result is an inner loop phi.
2412 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2414 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
2415 grouped stores with gaps to a strided group.
2417 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2419 * tree-vect-stmts.c (get_group_load_store_type)
2420 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
2421 first statement in a group.
2423 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2425 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
2426 the sequence used in gcc/gcc.c.
2428 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2431 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
2434 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2436 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
2437 Darwin10-specific unwinder-shim.
2438 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
2439 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
2440 New to cater for Darwin10 Rosetta.
2442 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2444 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
2445 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
2447 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2452 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
2453 (dwarf2out_switch_text_section): Generate a local label for the
2454 second function sub-section and apply it as the second FDE start
2456 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
2457 second sub-section start.
2459 2018-08-22 Richard Biener <rguenther@suse.de>
2461 PR tree-optimization/86988
2462 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
2464 2018-08-22 Richard Biener <rguenther@suse.de>
2466 PR tree-optimization/86945
2467 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
2469 2018-08-22 Alexandre Oliva <oliva@adacore.com>
2471 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
2472 a comment about how uses of r2 for .sdata2 come about.
2474 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
2476 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
2478 2018-08-21 Marek Polacek <polacek@redhat.com>
2480 PR c++/86981, Implement -Wpessimizing-move.
2481 * doc/invoke.texi: Document -Wpessimizing-move.
2483 2018-08-21 Jan Hubicka <jh@suse.cz>
2485 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
2486 * tree.h (is_redundant_typedef): Remove.
2487 * dwarf2out.c (is_redundant_typedef): Turn into static function.
2489 2018-08-21 Jan Hubicka <jh@suse.cz>
2491 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
2494 2018-08-21 Tamar Christina <tamar.christina@arm.com>
2496 * expmed.c (extract_low_bits): Reject invalid subregs early.
2498 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
2501 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
2504 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2506 * config/vxworks.h: Guard vxworks_asm_out_constructor and
2507 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
2508 * config/vxworks.c: Likewise.
2510 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2512 * config/vxworks.c: Set targetm.have_ctors_dtors
2513 if HAVE_INITFINI_ARRAY_SUPPORT.
2514 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
2515 if HAVE_INITFINI_ARRAY_SUPPORT.
2517 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2519 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
2520 default search path for VxWorks < 7.
2522 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
2524 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
2525 (get_format_string): Refer to c_getstr.
2527 2018-08-21 Tom de Vries <tdevries@suse.de>
2529 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
2530 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
2531 (debuginfo_early_stop): Declare.
2532 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
2533 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
2534 (debuginfo_early_stop): New function.
2535 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
2536 and debuginfo_early_stop.
2537 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
2538 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
2539 (general_init): Call debuginfo_early_init.
2540 (finalize): Call debuginfo_fini.
2541 (do_compile): Call debuginfo_init.
2542 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
2544 (@item -fdump-debug, @item -fdump-earlydebug): Add.
2546 2018-08-21 Tom de Vries <tdevries@suse.de>
2548 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
2549 flag_dump_noaddr and flag_dump_unnumbered.
2551 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
2553 * wide-int-range.cc (wide_int_range_abs): New.
2554 (wide_int_range_order_set): Rename from wide_int_range_min_max.
2555 * wide-int-range.h (wide_int_range_abs): New.
2556 (wide_int_range_min_max): New.
2557 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
2558 case to call wide_int_range_abs.
2559 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
2560 (extract_range_from_abs_expr): Delete.
2562 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
2565 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
2566 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
2569 2018-08-20 Nathan Sidwell <nathan@acm.org>
2570 Jeff Law <law@redhat.com>
2572 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
2573 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
2575 2018-08-20 David Malcolm <dmalcolm@redhat.com>
2578 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
2579 (decl_attributes): Likewise.
2580 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
2582 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
2583 * diagnostic-core.h (class auto_diagnostic_group): New class.
2584 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
2585 (diagnostic_report_diagnostic): Handle the first diagnostics within
2587 (emit_diagnostic): Add auto_diagnostic_group instance.
2589 (inform_n): Likewise.
2590 (warning): Likewise.
2591 (warning_at): Likewise.
2592 (warning_n): Likewise.
2593 (pedwarn): Likewise.
2594 (permerror): Likewise.
2596 (error_n): Likewise.
2597 (error_at): Likewise.
2599 (fatal_error): Likewise.
2600 (internal_error): Likewise.
2601 (internal_error_no_backtrace): Likewise.
2602 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
2603 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
2604 * diagnostic.h (struct diagnostic_context): Add fields
2605 "diagnostic_group_nesting_depth",
2606 "diagnostic_group_emission_count", "begin_group_cb",
2608 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
2609 Add auto_diagnostic_group instance(s).
2610 (find_explicit_erroneous_behavior): Likewise.
2611 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
2612 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
2613 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
2614 (gimplify_va_arg_expr): Likewise.
2615 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
2616 (HSA_SORRY_AT): Likewise.
2617 * ipa-devirt.c (compare_virtual_tables): Likewise.
2618 (warn_odr): Likewise.
2619 * multiple_target.c (expand_target_clones): Likewise.
2620 * opts-common.c (cmdline_handle_error): Likewise.
2621 * reginfo.c (globalize_reg): Likewise.
2622 * substring-locations.c (format_warning_n_va): Likewise.
2623 * tree-inline.c (expand_call_inline): Likewise.
2624 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
2625 * tree-ssa-loop-niter.c
2626 (do_warn_aggressive_loop_optimizations): Likewise.
2627 * tree-ssa-uninit.c (warn_uninit): Likewise.
2628 * tree.c (warn_deprecated_use): Likewise.
2630 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
2633 * config/i386/i386.md (eh_return): Always update EH return
2634 address in word_mode.
2636 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
2638 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
2639 TARGET_SPLIT_COMPLEX_ARG is defined.
2641 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
2643 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
2645 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
2648 * expr.c (expand_assignment): Assert that bitpos is positive.
2649 (store_field): Likewise
2650 (expand_expr_real_1): Make sure that bitpos is positive.
2651 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
2654 2018-08-20 Nathan Sidwell <nathan@acm.org>
2656 * Makefile.in (CPP_ID_DATA_H): Delete.
2657 (CPP_INTERNAL_H): Don't add it.
2658 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
2659 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
2661 2018-08-20 Richard Biener <rguenther@suse.de>
2663 PR tree-optimization/78655
2664 * tree-vrp.c (extract_range_from_binary_expr_1): Make
2665 pointer + offset nonnull if either operand is nonnull work.
2667 2018-08-20 Tom de Vries <tdevries@suse.de>
2669 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
2670 unless the referenced die describes the added property using
2671 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
2672 Otherwise, add a DW_AT_location to the referenced die.
2674 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
2677 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
2678 register_operand when calling ix86_set_reg_reg_cost.
2679 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
2680 Set *total to 0 for operands that satisfy x86_64_immediate_operand
2681 predicate and to 1 otherwise.
2683 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
2685 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
2686 emit a diagnostic that it is not supported and reset the option.
2687 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
2688 supported and consume the option. (ASM_FINAL_SPEC): New.
2690 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
2692 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
2695 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
2697 C-SKY port: Documentation
2699 * doc/extend.texi (C-SKY Function Attributes): New section.
2700 * doc/invoke.texi (Option Summary): Add C-SKY options.
2701 (C-SKY Options): New section.
2702 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
2704 2018-08-17 Jojo <jijie_rong@c-sky.com>
2705 Huibin Wang <huibin_wang@c-sky.com>
2706 Sandra Loosemore <sandra@codesourcery.com>
2707 Chung-Lin Tang <cltang@codesourcery.com>
2709 C-SKY port: Backend implementation
2711 * config/csky/*: New.
2712 * common/config/csky/*: New.
2714 2018-08-17 Jojo <jijie_rong@c-sky.com>
2715 Huibin Wang <huibin_wang@c-sky.com>
2716 Sandra Loosemore <sandra@codesourcery.com>
2717 Chung-Lin Tang <cltang@codesourcery.com>
2718 Andrew Jenner <andrew@codesourcery.com>
2720 C-SKY port: Configury
2722 * config.gcc (csky-*-*): New.
2723 * configure.ac: Add csky to targets for dwarf2 debug_line support.
2724 * configure: Regenerated.
2726 2018-08-17 David Malcolm <dmalcolm@redhat.com>
2728 * dump-context.h: Include "dumpfile.h".
2729 (dump_context::dump_printf_va): Convert final param from va_list
2730 to va_list *. Convert from ATTRIBUTE_PRINTF to
2731 ATTRIBUTE_GCC_DUMP_PRINTF.
2732 (dump_context::dump_printf_loc_va): Likewise.
2733 * dumpfile.c: Include "stringpool.h".
2734 (make_item_for_dump_printf_va): Delete.
2735 (make_item_for_dump_printf): Delete.
2736 (class dump_pretty_printer): New class.
2737 (dump_pretty_printer::dump_pretty_printer): New ctor.
2738 (dump_pretty_printer::emit_items): New member function.
2739 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
2741 (dump_pretty_printer::emit_item): New member function.
2742 (dump_pretty_printer::stash_item): New member function.
2743 (dump_pretty_printer::format_decoder_cb): New member function.
2744 (dump_pretty_printer::decode_format): New member function.
2745 (dump_context::dump_printf_va): Reimplement in terms of
2746 dump_pretty_printer.
2747 (dump_context::dump_printf_loc_va): Convert final param from va_list
2749 (dump_context::begin_scope): Reimplement call to
2750 make_item_for_dump_printf.
2751 (dump_printf): Update for change to dump_printf_va.
2752 (dump_printf_loc): Likewise.
2753 (selftest::test_capture_of_dump_calls): Convert "stmt" from
2754 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
2755 with %T, %E, and %G.
2756 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
2757 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
2758 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
2759 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
2760 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
2761 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
2762 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
2763 within a dump_printf_loc call to "%wu".
2764 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
2765 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
2766 missing space after "=".
2767 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
2768 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
2769 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
2770 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
2771 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
2772 duplicate "vectorized" from message.
2774 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
2776 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
2777 polyNxK_t element's TYPE_STRING_FLAG.
2779 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
2781 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
2782 (they were unnamed before). Fix comments.
2784 2018-08-17 Nathan Sidwell <nathan@acm.org>
2786 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
2788 2018-08-17 Richard Biener <rguenther@suse.de>
2790 PR tree-optimization/86841
2791 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
2793 2018-08-17 Martin Liska <mliska@suse.cz>
2795 * common.opt: Remove Warn, Init and Report for options with
2796 Ignore/Deprecated flag. Warning is done automatically for
2798 * config/i386/i386.opt: Likewise.
2799 * config/ia64/ia64.opt: Likewise.
2800 * config/rs6000/rs6000.opt: Likewise.
2801 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
2802 Remove usage of flag_check_pointer_bounds.
2803 * lto-wrapper.c (merge_and_complain): Do not handle
2804 OPT_fcheck_pointer_bounds.
2805 (append_compiler_options): Likewise.
2806 * opt-functions.awk: Do not handle Deprecated.
2807 * optc-gen.awk: Check that Var, Report and Init are not
2808 used for an option with Ignore/Deprecated flag.
2809 * opts-common.c (decode_cmdline_option): Do not report
2811 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
2813 * opts.h (struct cl_option): Remove cl_deprecated flag.
2814 (CL_ERR_DEPRECATED): Remove error enum value.
2816 2018-08-17 Richard Biener <rguenther@suse.de>
2819 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
2820 across a va-arg-pack using call adjust its return value accordingly.
2822 2018-08-16 Martin Sebor <msebor@redhat.com>
2824 PR tree-optimization/86853
2825 * gimple-ssa-sprintf.c (struct format_result): Rename member.
2826 (struct fmtresult): Add member and initialize it in ctors.
2827 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
2828 (format_string): Handle %S the same as %ls. Set MAYFAIL.
2829 (format_directive): Set POSUNDER4K when MAYFAIL is set.
2830 (parse_directive): Handle %C same as %c.
2831 (sprintf_dom_walker::compute_format_length): Adjust.
2832 (is_call_safe): Adjust.
2834 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
2836 * builtins.c (c_strlen): Add new parameter eltsize. Use it
2837 for determining how to count the elements.
2838 * builtins.h (c_strlen): Adjust prototype.
2839 * expr.c (string_constant): Add new parameter mem_size.
2840 Set *mem_size appropriately.
2841 * expr.h (string_constant): Adjust protoype.
2842 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
2843 * gimple-fold.h (get_range_strlen): Adjust prototype.
2844 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
2845 (format_string): Call get_string_length with eltsize.
2847 2018-08-16 David Malcolm <dmalcolm@redhat.com>
2849 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
2850 to emit the span, rather than setting it as the prefix.
2852 2018-08-16 David Malcolm <dmalcolm@redhat.com>
2854 * diagnostic-show-locus.c (layout::start_annotation_line): Add
2855 "margin_char" parameter, defaulting to space. Use it in place
2856 of pp_space for the initial part of the margin.
2857 (layout::print_leading_fixits): Use '+' when filling the margin
2858 of line-insertion fix-it hints.
2860 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2862 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
2865 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2867 * config/rs6000/altivec.md: Don't set length attribute to the default
2869 * config/rs6000/darwin.md: Ditto.
2870 * config/rs6000/dfp.md: Ditto.
2871 * config/rs6000/htm.md: Ditto.
2872 * config/rs6000/rs6000.md: Ditto.
2873 * config/rs6000/sync.md: Ditto.
2874 * config/rs6000/vsx.md: Ditto.
2876 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2878 * config/rs6000/altivec.md: Don't set length attribute to the default
2879 value, for branch instructions.
2880 * config/rs6000/darwin.md: Ditto.
2881 * config/rs6000/rs6000.md: Ditto.
2883 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2885 * config/rs6000/rs6000.md (length): Always define as const_int 4.
2886 (unnamed conditional branch define_insn): Set length to 4 or 8
2887 depending on offset.
2888 (<bd>_<mode>): Similar, for alternative 0.
2889 (<bd>tf_<mode>): Ditto.
2891 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2893 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
2895 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
2897 * doc/rtl.texi: Replace old RTX class names with new names.
2900 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
2902 * expmed.h (canonicalize_comparison): New declaration.
2903 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
2904 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
2905 * optabs.c (prepare_cmp_insn): Likewise.
2906 * rtl.h (unsigned_condition_p): New function which checks if a
2907 comparison operator is unsigned.
2909 2018-08-16 Nathan Sidwell <nathan@acm.org>
2911 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
2912 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
2914 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2917 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
2918 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
2919 (mov<mov>): ..this and enable unconditionally.
2921 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2923 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
2925 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
2927 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
2928 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
2929 "Common" with "Target".
2931 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
2933 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
2934 * doc/invoke.texi (mmitigate-rop): Remove.
2935 * config/i386/i386.c: Do not include "regrename.h".
2936 (ix86_rop_should_change_byte_p, reg_encoded_number)
2937 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
2939 (ix86_reorg): Remove call to ix86_mitigate_rop.
2940 * config/i386/i386.md (attr "modrm_class"): Remove.
2941 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
2942 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
2943 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
2945 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
2947 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
2948 allow folding of mergeh() and mergel() for the float and double types.
2949 (fold_mergehl_helper): Rework to handle building a permute tree
2952 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
2954 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
2957 2018-08-15 David Malcolm <dmalcolm@redhat.com>
2959 * common.opt (fdiagnostics-show-labels): New option.
2960 * diagnostic-show-locus.c (class layout_range): Add field
2962 (class layout): Add field "m_show_labels_p".
2963 (layout_range::layout_range): Add param "label" and use it to
2965 (make_range): Pass in NULL for new "label" param of layout_range's
2967 (layout::layout): Initialize m_show_labels_p.
2968 (layout::maybe_add_location_range): Pass in loc_range->m_label
2969 when constructing layout_range instances.
2970 (struct line_label): New struct.
2971 (layout::print_any_labels): New member function.
2972 (layout::print_line): Call it if label-printing is enabled.
2973 (selftest::test_one_liner_labels): New test.
2974 (selftest::test_diagnostic_show_locus_one_liner): Call it.
2975 * diagnostic.c (diagnostic_initialize): Initialize
2976 context->show_labels_p.
2977 * diagnostic.h (struct diagnostic_context): Add field
2979 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
2980 -fno-diagnostics-show-labels.
2981 * dwarf2out.c (gen_producer_string): Add
2982 OPT_fdiagnostics_show_labels to the ignored options.
2983 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
2985 (gcc_rich_location::maybe_add_expr): Likewise.
2986 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
2987 label" param, defaulting to NULL.
2988 (gcc_rich_location::add_expr): Add "label" param.
2989 (gcc_rich_location::maybe_add_expr): Likewise.
2990 (class text_range_label): New class.
2991 (class range_label_for_type_mismatch): New class.
2992 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
2993 of format_warning_va.
2994 (fmtwarn_n): Likewise for new params of format_warning_n_va.
2995 * lto-wrapper.c (merge_and_complain): Add
2996 OPT_fdiagnostics_show_labels to the "pick one setting" options.
2997 (append_compiler_options): Likewise to the dropped options.
2998 (append_diag_options): Likewise to the passed-on options.
2999 * opts.c (common_handle_option): Handle the new option.
3000 * selftest-diagnostic.c
3001 (test_diagnostic_context::test_diagnostic_context): Enable
3003 * substring-locations.c: Include "gcc-rich-location.h".
3004 (format_warning_n_va): Add "fmt_label" and "param_label" params
3005 and use them as appropriate.
3006 (format_warning_va): Add "fmt_label" and "param_label" params,
3007 passing them on to format_warning_n_va.
3008 (format_warning_at_substring): Likewise.
3009 (format_warning_at_substring_n): Likewise.
3010 * substring-locations.h (format_warning_va): Add "fmt_label" and
3011 "param_label" params.
3012 (format_warning_n_va): Likewise.
3013 (format_warning_at_substring): Likewise.
3014 (format_warning_at_substring_n): Likewise.
3015 * toplev.c (general_init): Initialize global_dc->show_labels_p.
3017 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
3020 * builtins.c (expand_builtin_memcmp): Do not expand the call
3021 when overflow is detected.
3023 2018-08-15 Martin Sebor <msebor@redhat.com>
3025 PR tree-optimization/71625
3026 * config/aarch64/aarch64-builtins.c
3027 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
3029 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
3031 * config/s390/s390.c (s390_reorg): Remove loop.
3033 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
3036 (darwin_function_switched_text_sections): Delete.
3037 * gcc/config/darwin.h
3038 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
3040 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
3043 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
3044 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
3045 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
3047 2018-08-15 Martin Liska <mliska@suse.cz>
3049 PR tree-optimization/86925
3050 * predict.c (expr_expected_value_1): When taking
3051 later predictor, assign also probability.
3052 Use fold_build2_initializer_loc in order to fold
3053 the expression in -frounding-math.
3055 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
3057 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
3059 (expand_vec_perm_1): Try the new method.
3061 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
3064 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
3065 Check whether lra_live_max_point is 0 before dividing.
3067 2018-08-14 Martin Sebor <msebor@redhat.com>
3069 PR tree-optimization/86650
3070 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
3071 (vrp_prop::check_mem_ref): Same.
3073 2018-08-13 Liu Hao <lh_mouse@126.com>
3075 * pretty-print.c (eat_esc_sequence): Swap the foreground and
3076 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
3077 and clear it thereafter, as it only works for DBCS.
3079 2018-08-13 Liu Hao <lh_mouse@126.com>
3081 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
3082 handle returned by _get_osf_handle().
3084 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
3086 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
3087 for folding vec_perm.
3089 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
3091 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
3092 Add support for gimple-folding of vec_pack() and vec_unpack()
3095 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
3097 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
3098 vec_xst variants to the list.
3099 (rs6000_gimple_fold_builtin): Add support for folding unaligned
3100 vector loads and stores.
3102 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
3104 * config.gcc (rs6000-ibm-aix4.x): Delete.
3105 (rs6000-ibm-aix5.1): Delete.
3106 (rs6000-ibm-aix5.2): Delete.
3107 (rs6000-ibm-aix5.3): Delete.
3108 * config/rs6000/aix43.h: Delete.
3109 * config/rs6000/aix51.h: Delete.
3110 * config/rs6000/aix52.h: Delete.
3111 * config/rs6000/t-aix43: Delete.
3113 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
3115 * config/s390/s390.c (s390_decompose_constant_pool_ref):
3117 (s390_decompose_address): Factor out constant pool ref
3120 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
3122 * config/nds32/nds32-predicates.c
3123 (nds32_can_use_bclr_p): Change return type as bool.
3124 (nds32_can_use_bset_p): Ditto.
3125 (nds32_can_use_btgl_p): Ditto.
3126 (nds32_can_use_bitci_p): Ditto.
3127 * config/nds32/nds32-protos.h
3128 (nds32_can_use_bclr_p): Change declaration.
3129 (nds32_can_use_bset_p): Ditto.
3130 (nds32_can_use_btgl_p): Ditto.
3131 (nds32_can_use_bitci_p): Ditto.
3133 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
3135 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
3136 Support -msched-prolog-epilog option.
3137 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
3139 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
3141 * common/config/nds32/nds32-common.c
3142 (nds32_option_optimization_table): Enalbe -malways-align.
3144 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
3146 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
3148 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
3149 OPT_misr_secure_ case.
3150 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
3151 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
3152 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
3154 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
3155 (nds32_isr_info): New field security_level.
3156 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
3157 * config/nds32/nds32.md (return_internal): Consider critical attribute.
3158 * config/nds32/nds32.opt (misr-secure): New option.
3159 * config/nds32/nds32_init.inc: New file.
3160 * config/nds32/nds32_isr.h: New file.
3162 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
3164 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
3165 Update comment for atomic instructions.
3166 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
3167 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
3169 (atomic_loaddi): Revise fence expansion to only emit fence prior to
3170 load for __ATOMIC_SEQ_CST model.
3171 (atomic_loaddi_1): Remove float register target.
3172 (atomic_storedi): Handle CONST_INT values.
3173 (atomic_storedi_1): Remove float register source. Add special case
3175 (memory_barrier): New expander and insn.
3177 2018-08-11 Jakub Jelinek <jakub@redhat.com>
3179 PR tree-optimization/86835
3180 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
3181 new_stmt after def_gsi, make sure to insert new_square_stmt after
3182 that stmt, not 2 stmts before it.
3184 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
3187 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
3190 2018-08-10 Martin Liska <mliska@suse.cz>
3193 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
3195 * builtins.c (expand_builtin_expect_with_probability):
3197 (expand_builtin_expect_with_probability): New function.
3198 (build_builtin_expect_predicate): Add new argumnet probability
3199 for BUILT_IN_EXPECT_WITH_PROBABILITY.
3200 (fold_builtin_expect):
3203 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
3204 * builtins.h (fold_builtin_expect): Set new argument.
3205 * doc/extend.texi: Document __builtin_expect_with_probability.
3206 * doc/invoke.texi: Likewise.
3207 * gimple-fold.c (gimple_fold_call): Pass new argument.
3208 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
3209 also BUILT_IN_EXPECT_WITH_PROBABILITY.
3210 * predict.c (get_predictor_value): New function.
3211 (expr_expected_value): Add new argument probability. Assume
3212 that predictor and probability are always non-null.
3213 (expr_expected_value_1): Likewise. For __builtin_expect and
3214 __builtin_expect_with_probability set probability. Handle
3215 combination in binary expressions.
3216 (tree_predict_by_opcode): Simplify code by simply calling
3217 get_predictor_value.
3218 (pass_strip_predict_hints::execute): Add handling of
3219 BUILT_IN_EXPECT_WITH_PROBABILITY.
3220 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
3222 * tree.h (DECL_BUILT_IN_P): New function.
3224 2018-08-10 Martin Liska <mliska@suse.cz>
3226 PR tree-optimization/85799
3227 * passes.def: Add argument for pass_strip_predict_hints.
3228 * predict.c (class pass_strip_predict_hints): Add new argument
3230 (strip_predictor_early): New function.
3231 (pass_strip_predict_hints::execute): Call the function to
3233 (strip_predict_hints): New function.
3234 * predict.def: Fix comment.
3236 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
3238 * Makefile.in: Clarify which tm.texi to copy over to assert the
3239 right to grant a GFDL license for all.
3241 2018-08-09 Jeff Law <law@redhat.com>
3243 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
3246 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
3248 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
3251 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
3253 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
3254 reductions for variable-length vectors.
3256 2018-08-09 David Malcolm <dmalcolm@redhat.com>
3259 * common.opt (fdiagnostics-show-line-numbers): New option.
3260 * diagnostic-show-locus.c (class layout): Add fields
3261 "m_show_line_numbers_p" and "m_linenum_width";
3262 (num_digits): New function.
3263 (test_num_digits): New function.
3264 (layout::layout): Initialize new fields. Update m_x_offset
3265 logic to handle any left margin.
3266 (layout::print_source_line): Print line number when requested.
3267 (layout::start_annotation_line): New member function.
3268 (layout::print_annotation_line): Call it.
3269 (layout::print_leading_fixits): Likewise.
3270 (layout::print_trailing_fixits): Likewise. Update calls to
3271 move_to_column for new parameter.
3272 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
3273 it to potentially call start_annotation_line.
3274 (layout::show_ruler): Call start_annotation_line.
3275 (selftest::test_line_numbers_multiline_range): New selftest.
3276 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
3277 and selftest::test_line_numbers_multiline_range.
3278 * diagnostic.c (diagnostic_initialize): Initialize
3279 show_line_numbers_p.
3280 * diagnostic.h (struct diagnostic_context): Add field
3281 "show_line_numbers_p".
3282 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
3283 -fno-diagnostics-show-line-numbers.
3284 * dwarf2out.c (gen_producer_string): Add
3285 OPT_fdiagnostics_show_line_numbers to the ignored options.
3286 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
3287 one setting" options.
3288 (append_compiler_options): Likewise to the dropped options.
3289 (append_diag_options): Likewise to the passed-on options.
3290 * opts.c (common_handle_option): Handle the new option.
3291 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
3293 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
3295 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
3296 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
3297 third argument of type "const signed char" to descriptions of
3298 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
3299 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
3300 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
3301 __builtin_bcdsub_ov functions.
3303 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
3305 PR tree-optimization/86858
3306 * tree-vect-loop.c (vect_is_simple_reduction): Restore
3307 flow_bb_inside_loop_p calls.
3309 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
3311 PR tree-optimization/86871
3312 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
3313 instead of gimple_assign_lhs.
3315 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
3318 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
3319 register constraint to operand 0.
3320 (add<mode>3_carryinC): Likewise.
3321 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
3323 2018-08-09 Martin Liska <mliska@suse.cz>
3326 * common.opt: Remove extra line.
3328 2018-08-09 Martin Liska <mliska@suse.cz>
3330 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
3331 at the end of a line, make first letter capital and end up
3332 a sentence with a dot.
3333 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
3334 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
3335 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
3336 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
3337 (PARAM_MAX_ISL_OPERATIONS): Likewise.
3338 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
3339 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
3340 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
3341 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
3342 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
3343 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
3344 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
3345 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
3346 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
3347 (PARAM_TREE_REASSOC_WIDTH): Likewise.
3348 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
3349 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
3350 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
3352 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
3355 * config/s390/s390.c (s390_option_override_internal): Reduce the
3356 stack-clash-protection-probe-interval param if it would be too big
3359 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
3362 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
3363 * config/m68k/m68k.c (handle_move_double): Don't call
3364 m68k_final_prescan_insn.
3365 (m68k_adjust_decorated_operand): Renamed from
3366 m68k_final_prescan_insn, remove first and third operand and
3368 (print_operand): Call it.
3369 (print_operand_address): Call it.
3371 2018-08-08 Nathan Sidwell <nathan@acm.org>
3373 * diagnostic.c (diagnostic_report_current_module): Use
3374 linemap_included_from & linemap_included_from_linemap.
3376 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
3378 * config/aarch64/aarch64-cores.def: Add phecda core.
3379 * config/aarch64/aarch64-tune.md: Regenerate.
3380 * doc/invoke.texi: Add phecda core.
3382 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
3385 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
3387 * config/s390/s390.md ("movti"): Add more alternatives for
3388 constant to GPR copies.
3390 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
3392 * config/s390/s390.c: Fix whitespace damage throughout the file.
3393 * config/s390/s390.h: Likewise.
3394 * config/s390/tpf.h: Likewise.
3396 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
3398 * config/s390/s390.c (s390_loadrelative_operand_p):
3399 Remove TARGET_CPU_ZARCH usages.
3400 (s390_rtx_costs): Likewise.
3401 (s390_legitimate_constant_p): Likewise.
3402 (s390_cannot_force_const_mem): Likewise.
3403 (legitimate_reload_constant_p): Likewise.
3404 (s390_preferred_reload_class): Likewise.
3405 (legitimize_pic_address): Likewise.
3406 (legitimize_tls_address): Likewise.
3407 (s390_split_branches): Removed.
3408 (s390_add_execute): Removed.
3409 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
3410 (s390_mainpool_start): Likewise.
3411 (s390_mainpool_finish): Likewise.
3412 (s390_mainpool_cancel): Removed.
3413 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
3414 (s390_chunkify_cancel): Likewise.
3415 (s390_return_addr_rtx): Likewise.
3416 (s390_register_info): Remove split_branches_pending_p uages.
3417 (s390_optimize_register_info): Likewise.
3418 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
3419 split_branches_pending_p usages.
3420 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
3421 (s390_load_got): Likewise.
3422 (s390_expand_split_stack_prologue): Likewise.
3423 (output_asm_nops): Likewise.
3424 (s390_function_profiler): Likewise.
3425 (s390_emit_call): Likewise.
3426 (s390_conditional_register_usage): Likewise.
3427 (s390_optimize_prologue): Likewise.
3428 (s390_reorg): Remove TARGET_CPU_ZARCH and
3429 split_branches_pending_p usages.
3430 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
3432 (s390_output_indirect_thunk_function): Likewise.
3433 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
3434 (TARGET_CPU_ZARCH_P): Removed.
3435 (struct machine_function): Remove split_branches_pending_p.
3436 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
3438 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
3440 * common/config/s390/s390-common.c (processor_flags_table):
3442 * config.gcc: Remove with_arch/with_tune support.
3443 * config/s390/2064.md: Remove cpu attribute comparisons.
3444 * config/s390/driver-native.c (s390_host_detect_local_cpu):
3446 * config/s390/linux.h (ASM_SPEC):
3447 Remove -march support.
3448 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
3449 Use a table to get an arch level.
3450 * config/s390/s390-opts.h (enum processor_type):
3452 * config/s390/s390.c
3453 (processor_table): Remove entries, add arch_level values.
3454 (s390_issue_rate): Remove cases.
3455 (s390_option_override): Adjust
3456 s390_option_override_internal() call.
3457 (s390_option_override_internal): Remove deprecation warning.
3458 (s390_valid_target_attribute_tree): Adjust
3459 s390_option_override_internal() call.
3460 * config/s390/s390.h (struct s390_processor):
3461 Share with s390-c.c, add arch_level field.
3462 * config/s390/s390.md:
3463 Remove occurrences in cpu attribute.
3464 * config/s390/s390.opt: Remove -march/-mtune support.
3465 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
3466 * doc/invoke.texi: Remove deprecation warning.
3468 2018-08-08 Luis Machado <luis.machado@linaro.org>
3470 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
3472 (qdf24xx_tunings): Set vector cost structure to
3473 qdf24xx_vector_cost.
3475 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
3476 <register_sextend>: Set to 3.
3478 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
3481 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
3482 * config/aarch64/aarch64-simd.md
3483 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
3484 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
3485 (aarch64_frecpx<mode>): ...this new pattern.
3486 * config/aarch64/aarch64-simd-builtins.def: Remove comment
3487 about aarch64_frecp<FRECP:frecp_suffix><mode>.
3489 2018-08-07 Martin Liska <mliska@suse.cz>
3492 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
3493 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
3494 * predict.def (PRED_MALLOC_NONNULL): New predictor.
3495 * doc/extend.texi: Document that malloc attribute adds
3498 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
3501 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3502 Define to speculation_safe_value_not_needed.
3504 2018-08-06 Jeff Law <law@redhat.com>
3506 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
3507 the vr_values instance to cprop_into_stmt.
3508 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
3509 (cprop_operand): Also query EVRP to determine if OP is a constant.
3511 2018-08-06 Nathan Sidwell <nathan@acm.org>
3513 * diagnostic.c (diagnostic_report_current_module): Reroll
3514 included-at loop. Translate text.
3516 2018-08-06 David Malcolm <dmalcolm@redhat.com>
3518 * function-tests.c (selftest::test_expansion_to_rtl): Call
3519 free_after_compilation.
3521 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3523 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
3525 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
3527 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
3528 loops with memory block operations from getting unrolled.
3530 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
3533 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3534 Define to speculation_safe_value_not_needed.
3536 2018-08-06 Jeff Law <law@redhat.com>
3538 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
3541 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3544 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
3545 with all enabled __intN types.
3547 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
3549 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3551 * alias.c (record_set): Check for clobber high.
3552 * cfgexpand.c (expand_gimple_stmt): Likewise.
3553 * combine-stack-adj.c (single_set_for_csa): Likewise.
3554 * combine.c (find_single_use_1): Likewise.
3555 (set_nonzero_bits_and_sign_copies): Likewise.
3556 (get_combine_src_dest): Likewise.
3557 (is_parallel_of_n_reg_sets): Likewise.
3558 (try_combine): Likewise.
3559 (record_dead_and_set_regs_1): Likewise.
3560 (reg_dead_at_p_1): Likewise.
3561 (reg_dead_at_p): Likewise.
3562 * dce.c (deletable_insn_p): Likewise.
3563 (mark_nonreg_stores_1): Likewise.
3564 (mark_nonreg_stores_2): Likewise.
3565 * df-scan.c (df_find_hard_reg_defs): Likewise.
3566 (df_uses_record): Likewise.
3567 (df_get_call_refs): Likewise.
3568 * dwarf2out.c (mem_loc_descriptor): Likewise.
3569 * haifa-sched.c (haifa_classify_rtx): Likewise.
3570 * ira-build.c (create_insn_allocnos): Likewise.
3571 * ira-costs.c (scan_one_insn): Likewise.
3572 * ira.c (equiv_init_movable_p): Likewise.
3573 (rtx_moveable_p): Likewise.
3574 (interesting_dest_for_shprep): Likewise.
3575 * jump.c (mark_jump_label_1): Likewise.
3576 * postreload-gcse.c (record_opr_changes): Likewise.
3577 * postreload.c (reload_cse_simplify): Likewise.
3578 (struct reg_use): Add source expr.
3579 (reload_combine): Check for clobber high.
3580 (reload_combine_note_use): Likewise.
3581 (reload_cse_move2add): Likewise.
3582 (move2add_note_store): Likewise.
3583 * print-rtl.c (print_pattern): Likewise.
3584 * recog.c (decode_asm_operands): Likewise.
3585 (store_data_bypass_p): Likewise.
3586 (if_test_bypass_p): Likewise.
3587 * regcprop.c (kill_clobbered_value): Likewise.
3588 (kill_set_value): Likewise.
3589 * reginfo.c (reg_scan_mark_refs): Likewise.
3590 * reload1.c (maybe_fix_stack_asms): Likewise.
3591 (eliminate_regs_1): Likewise.
3592 (elimination_effects): Likewise.
3593 (mark_not_eliminable): Likewise.
3594 (scan_paradoxical_subregs): Likewise.
3595 (forget_old_reloads_1): Likewise.
3596 * reorg.c (find_end_label): Likewise.
3597 (try_merge_delay_insns): Likewise.
3598 (redundant_insn): Likewise.
3599 (own_thread_p): Likewise.
3600 (fill_simple_delay_slots): Likewise.
3601 (fill_slots_from_thread): Likewise.
3602 (dbr_schedule): Likewise.
3603 * resource.c (update_live_status): Likewise.
3604 (mark_referenced_resources): Likewise.
3605 (mark_set_resources): Likewise.
3606 * rtl.c (copy_rtx): Likewise.
3607 * rtlanal.c (reg_referenced_p): Likewise.
3608 (single_set_2): Likewise.
3609 (noop_move_p): Likewise.
3610 (note_stores): Likewise.
3611 * sched-deps.c (sched_analyze_reg): Likewise.
3612 (sched_analyze_insn): Likewise.
3614 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3616 * cse.c (invalidate_reg): New function extracted from...
3617 (invalidate): ...here.
3618 (canonicalize_insn): Check for clobber high.
3619 (invalidate_from_clobbers): invalidate clobber highs.
3620 (invalidate_from_sets_and_clobbers): Likewise.
3621 (count_reg_usage): Check for clobber high.
3622 (insn_live_p): Likewise.
3623 * cselib.c (cselib_expand_value_rtx_1):Likewise.
3624 (cselib_invalidate_regno): Check for clobber in setter.
3625 (cselib_invalidate_rtx): Pass through setter.
3626 (cselib_invalidate_rtx_note_stores):
3627 (cselib_process_insn): Check for clobber high.
3628 * cselib.h (cselib_invalidate_rtx): Add operand.
3630 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3632 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
3633 (mark_not_eliminable): Likewise.
3634 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
3635 * lra-lives.c (process_bb_lives): Check for clobber high.
3636 * lra.c (new_insn_reg): Remember clobber highs.
3637 (collect_non_operand_hard_regs): Check for clobber high.
3638 (lra_set_insn_recog_data): Likewise.
3639 (add_regs_to_insn_regno_info): Likewise.
3640 (lra_update_insn_regno_info): Likewise.
3642 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3644 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
3645 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
3647 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3649 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
3650 (copy_insn_1): Likewise.
3651 (gen_hard_reg_clobber_high): New gen function.
3652 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
3653 * genemit.c (gen_exp): Likewise.
3654 (gen_emit_seq): Pass through info.
3655 (gen_insn): Check for CLOBBER_HIGH.
3656 (gen_expand): Pass through info.
3657 (gen_split): Likewise.
3658 (output_add_clobbers): Likewise.
3659 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
3660 (remove_clobbers): Likewise.
3661 * rtl.h (gen_hard_reg_clobber_high): New declaration.
3663 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3665 * doc/rtl.texi (clobber_high): Add.
3666 (parallel): Add in clobber high
3667 * rtl.c (rtl_check_failed_code3): Add function.
3668 * rtl.def (CLOBBER_HIGH): Add expression.
3669 * rtl.h (RTL_CHECKC3): Add macro.
3670 (rtl_check_failed_code3): Add declaration.
3671 (XC3EXP): Add macro.
3673 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
3676 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
3677 cfun->machine->max_used_stack_alignment if needed.
3679 2018-08-04 Martin Sebor <msebor@redhat.com>
3681 PR tree-optimization/86571
3682 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
3685 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
3688 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3691 2018-08-03 Jeff Law <law@redhat.com>
3694 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3695 Define to speculation_safe_value_not_needed.
3697 2018-08-03 David Malcolm <dmalcolm@redhat.com>
3699 * doc/gcov.texi (-x): Remove duplicate "to".
3700 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
3701 (-Wif-not-aligned): Remove duplicate "is".
3702 (-flto): Remove duplicate "the".
3703 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
3704 duplicate "v5.00.b".
3705 (MSP430 Options): Remove duplicate "and" from the description
3706 of "-mgprel-sec=regexp".
3707 (x86 Options): Remove duplicate copies of "vmldLog102" and
3708 vmlsLog104 from description of "-mveclibabi=type".
3710 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
3712 * internal-fn.h (first_commutative_argument): Declare.
3713 * internal-fn.c (first_commutative_argument): New function.
3714 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
3715 restrictions for pattern statements. Use first_commutative_argument
3716 to look for commutative operands in calls to internal functions.
3718 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
3720 * Makefile.in (wide-int-range.o): New.
3721 * tree-vrp.c: Move all the wide_int_* functions to...
3722 * wide-int-range.cc: ...here.
3723 * tree-vrp.h: Move all the wide_int_* prototypes to...
3724 * wide-int-range.h: ...here.
3726 2018-08-03 Tom de Vries <tdevries@suse.de>
3728 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
3730 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
3731 * except.c (output_function_exception_table): Do early exit if
3732 targetm_common.except_unwind_info (&global_options) == UI_NONE.
3734 2018-08-03 Martin Liska <mliska@suse.cz>
3736 * predict.c (dump_prediction): Change to 2 digits
3739 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
3741 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
3743 (falkor_am_1_gtov_gtov): ... a new insn reservation.
3745 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
3747 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
3748 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
3749 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
3750 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
3751 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
3753 2018-08-02 David Malcolm <dmalcolm@redhat.com>
3755 * diagnostic-show-locus.c (diagnostic_show_locus): Use
3756 pp_take_prefix when saving the existing prefix.
3757 * diagnostic.c (diagnostic_append_note): Likewise.
3758 * langhooks.c (lhd_print_error_function): Likewise.
3759 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
3760 param's type. Free the existing prefix.
3761 (pp_take_prefix): New function.
3762 (pretty_printer::pretty_printer): Drop the prefix parameter.
3763 Rename the length parameter to match the comment.
3764 (pretty_printer::~pretty_printer): Free the prefix.
3765 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
3767 (struct pretty_printer): Drop the "const" from "prefix" field's
3768 type and clarify memory management.
3769 (pp_set_prefix): Drop the "const" from the 2nd param.
3770 (pp_take_prefix): New decl.
3772 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
3774 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
3775 for word_mode_ok here instead of passing as argument.
3776 (expand_block_compare): Change select_block_compare_mode() call.
3777 (expand_strncmp_gpr_sequence): New function.
3778 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
3780 2018-08-02 Jeff Law <law@redhat.com>
3783 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3784 Define to speculation_safe_value_not_needed.
3787 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3788 Define to speculation_safe_value_not_needed.
3790 2018-08-02 Tom de Vries <tdevries@suse.de>
3793 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
3794 function. Return UI_TARGET unconditionally.
3795 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
3796 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
3798 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
3800 * genemit.c (print_overload_test): Fix typo.
3802 2018-08-02 Richard Biener <rguenther@suse.de>
3804 PR tree-optimization/86816
3805 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
3806 which checks for value availability before querying it.
3807 (gvn_uses_equal): Use it.
3808 (same_succ_hash): Likewise.
3809 (gimple_equal_p): Likewise.
3811 2018-08-02 Nick Clifton <nickc@redhat.com>
3814 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3815 Define to speculation_safe_value_not_needed.
3818 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3819 Define to speculation_safe_value_not_needed.
3822 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3823 Define to speculation_safe_value_not_needed.
3826 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3827 Define to speculation_safe_value_not_needed.
3830 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3831 Define to speculation_safe_value_not_needed.
3834 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3835 Define to speculation_safe_value_not_needed.
3838 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3839 Define to speculation_safe_value_not_needed.
3842 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3843 Define to speculation_safe_value_not_needed.
3846 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
3847 speculation_safe_value_not_needed.
3850 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
3851 to speculation_safe_value_not_needed.
3853 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
3855 * doc/md.texi: Expand the documentation of instruction names
3856 to mention port-local uses. Document '@' in pattern names.
3857 * read-md.h (overloaded_instance, overloaded_name): New structs.
3859 (md_reader::handle_overloaded_name): New member function.
3860 (md_reader::get_overloads): Likewise.
3861 (md_reader::m_first_overload): New member variable.
3862 (md_reader::m_next_overload_ptr): Likewise.
3863 (md_reader::m_overloads_htab): Likewise.
3864 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
3865 m_next_overload_ptr and m_overloads_htab.
3866 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
3867 (get_mode_token, get_code_token, get_int_token): New functions.
3868 (map_attr_string): Add an optional argument that passes back
3869 the associated iterator.
3870 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
3871 (md_reader::handle_overloaded_name, add_overload_instance): New
3873 (apply_iterators): Handle '@' names. Report an error if '@'
3874 is used without iterators.
3875 (initialize_iterators): Initialize the new iterator_group fields.
3876 * genopinit.c (handle_overloaded_code_for)
3877 (handle_overloaded_gen): New functions.
3878 (main): Use them to print declarations of maybe_code_for_* and
3879 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
3880 * genemit.c (print_overload_arguments, print_overload_test)
3881 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
3882 (main): Use it to print definitions of maybe_code_for_* and
3883 maybe_gen_* functions.
3884 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
3885 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
3886 instead of explicit mode checks.
3887 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
3888 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
3889 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
3890 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
3891 (aarch64_expand_compare_and_swap): Likewise
3892 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
3893 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
3894 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
3895 (aarch64_constant_pool_reload_icode): Delete.
3896 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
3897 instead of aarch64_constant_pool_reload_icode. Use
3898 code_for_aarch64_reload_mov instead of explicit mode checks.
3899 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
3900 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
3901 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
3902 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
3903 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
3904 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
3905 (aarch64_atomic_load_op_code): Delete.
3906 (aarch64_emit_atomic_load_op): Likewise.
3907 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
3908 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
3909 instead of aarch64_emit_atomic_load_op.
3910 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
3911 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
3912 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
3913 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
3914 character before the pattern name.
3915 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
3916 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
3917 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
3918 (aarch64_frecps<mode>): Likewise.
3919 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
3920 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
3921 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
3922 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
3923 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
3925 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
3927 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
3928 Allow HFmode constants if TARGET_FP_F16INST.
3930 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
3933 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
3934 No longer check last store for clobber of address register.
3936 2018-08-02 Martin Liska <mliska@suse.cz>
3938 PR gcov-profile/86817
3939 * gcov.c (process_all_functions): New function.
3941 (process_file): Move functions processing to
3942 process_all_functions.
3944 2018-08-02 David Malcolm <dmalcolm@redhat.com>
3946 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
3947 "const" to the "gimple *" and "rtx_insn *" parameters.
3948 * dumpfile.h (dump_user_location_t::dump_user_location_t):
3950 (dump_location_t::dump_location_t): Likewise.
3952 2018-08-01 Martin Sebor <msebor@redhat.com>
3954 PR tree-optimization/86650
3955 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
3956 rather than a "gcall *". Directly pass the data of interest
3957 to percent_K_format, rather than building a temporary CALL_EXPR
3959 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
3960 (gimple_fold_builtin_strncat): Adjust.
3961 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
3962 gcall* argument with gimple*.
3963 * gimple-ssa-warn-restrict.c (check_call): Same.
3964 (wrestrict_dom_walker::before_dom_children): Same.
3965 (builtin_access::builtin_access): Same.
3966 (check_bounds_or_overlap): Same
3967 (maybe_diag_overlap): Same.
3968 (maybe_diag_offset_bounds): Same.
3969 * tree-diagnostic.c (default_tree_printer): Move usage of
3970 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
3972 * tree-pretty-print.c (percent_K_format): Add argument.
3973 * tree-pretty-print.h: Add argument.
3974 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
3975 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
3976 (maybe_diag_stxncpy_trunc): Same.
3977 (handle_builtin_stxncpy): Same.
3978 (handle_builtin_strcat): Same.
3980 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3982 * match.pd: Optimise pointer range checks.
3984 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3986 PR tree-optimization/86758
3987 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
3988 to remove pattern statements.
3990 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3992 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
3993 result of dfs_enumerate_from when constructing stmt_vec_infos,
3994 instead of additionally calling get_loop_body.
3996 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3998 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
4000 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
4001 When creating an iv, assert that the step is not known to be zero.
4002 (vect_setup_realignment): Update call accordingly.
4003 * tree-vect-stmts.c (vectorizable_store): Likewise.
4004 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
4006 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4008 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
4009 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
4010 (vectorizable_reduction): Likewise.
4011 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4012 (vect_detect_hybrid_slp_stmts): Likewise.
4013 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
4015 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
4017 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
4018 (wide_int_set_zero_nonzero_bits): ...this.
4019 (zero_nonzero_bits_from_vr): Rename to...
4020 (vrp_set_zero_nonzero_bits): ...this.
4021 (extract_range_from_multiplicative_op_1): Abstract wide int
4023 (wide_int_range_multiplicative_op): ...here.
4024 (extract_range_from_binary_expr_1): Extract wide int binary
4025 operations into their own functions.
4026 (wide_int_range_lshift): New.
4027 (wide_int_range_can_optimize_bit_op): New.
4028 (wide_int_range_shift_undefined_p): New.
4029 (wide_int_range_bit_xor): New.
4030 (wide_int_range_bit_ior): New.
4031 (wide_int_range_bit_and): New.
4032 (wide_int_range_trunc_mod): New.
4033 (extract_range_into_wide_ints): New.
4034 (vrp_shift_undefined_p): New.
4035 (extract_range_from_multiplicative_op): New.
4036 (vrp_can_optimize_bit_op): New.
4037 * tree-vrp.h (value_range::dump): New.
4038 (wide_int_range_multiplicative_op): New.
4039 (wide_int_range_lshift):New.
4040 (wide_int_range_shift_undefined_p): New.
4041 (wide_int_range_bit_xor): New.
4042 (wide_int_range_bit_ior): New.
4043 (wide_int_range_bit_and): New.
4044 (wide_int_range_trunc_mod): New.
4045 (zero_nonzero_bits_from_bounds): Rename to...
4046 (wide_int_set_zero_nonzero_bits): ...this.
4047 (zero_nonzero_bits_from_vr): Rename to...
4048 (vrp_set_zero_nonzero_bits): ...this.
4049 (range_easy_mask_min_max): Rename to...
4050 (wide_int_range_can_optimize_bit_op): this.
4051 * vr-values.c (simplify_bit_ops_using_ranges): Rename
4052 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
4054 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4056 * tree-vectorizer.h (vect_orig_stmt): New function.
4057 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
4058 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
4059 (vect_create_epilog_for_reduction): Likewise.
4060 (vectorizable_live_operation): Likewise.
4061 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
4062 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
4063 * tree-vect-stmts.c (vectorizable_call): Likewise.
4064 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
4066 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4068 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
4070 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
4071 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
4072 (vect_transform_loop): Likewise.
4073 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
4075 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4077 * tree-vectorizer.h (vect_schedule_slp): Return void.
4078 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
4079 (vect_schedule_slp): Likewise.
4081 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4083 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
4085 (vect_transform_loop): Update calls accordingly. Schedule SLP
4086 instances before the main loop, if any exist.
4088 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4090 PR tree-optimization/86749
4091 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
4092 If the lhs is used in a COND_EXPR, check that it is being used
4093 as the "then" or "else" value.
4095 2018-08-01 Tom de Vries <tdevries@suse.de>
4098 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
4099 speculation_safe_value_not_needed.
4101 2018-08-01 Richard Biener <rguenther@suse.de>
4103 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
4106 2018-08-01 Martin Liska <mliska@suse.cz>
4108 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
4109 format how successful transformation is dumped.
4110 (gimple_mod_pow2_value_transform): Likewise.
4111 (gimple_mod_subtract_transform): Likewise.
4112 (gimple_stringops_transform): Likewise.
4114 2018-08-01 Martin Liska <mliska@suse.cz>
4117 * value-prof.c (interesting_stringop_to_profile_p):
4118 Simplify the code and add BUILT_IN_MEMMOVE.
4119 (gimple_stringops_transform): Likewise.
4121 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
4123 * config/aarch64/aarch64-simd.md
4124 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
4125 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
4126 use GPI iterator instead of SI mode.
4128 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
4130 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
4131 rs6000_speculation_barrier.
4132 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
4133 new barrier pattern name.
4135 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
4137 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
4138 (speculation_barrier): New insn.
4140 2018-08-01 Richard Biener <rguenther@suse.de>
4143 * graphite.h: Include isl/id.h and isl/space.h to allow build
4146 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
4149 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4150 mode for COFF targets.
4151 * defaults.h (TARGET_COFF): Define.
4152 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
4153 TARGET_COFF): Define.
4154 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
4155 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
4156 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
4158 2018-07-31 Alexandre Oliva <oliva@adacore.com>
4160 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
4161 * gimple-streamer-out.c (output_bb): Save it.
4162 * lto-streamer-in.c (input_struct_function_base): Restore
4163 instance discriminator if available. Create map on demand.
4164 * lto-streamer-out.c (output_struct_function_base): Save it if
4166 * final.c (decl_to_instance_map): Document LTO strategy.
4168 2018-07-31 Alexandre Oliva <oliva@adacore.com>
4169 Olivier Hainque <hainque@adacore.com>
4171 * debug.h (decl_to_instance_map_t): New type.
4172 (decl_to_instance_map): Declare.
4173 (maybe_create_decl_to_instance_map): New inline function.
4174 * final.c (bb_discriminator, last_bb_discriminator): New statics,
4175 to track basic block discriminators.
4176 (final_start_function_1): Initialize them.
4177 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
4179 (decl_to_instance_map): New variable.
4180 (map_decl_to_instance, maybe_set_discriminator): New functions.
4181 (notice_source_line): Set discriminator.
4183 2018-07-31 Ian Lance Taylor <iant@golang.org>
4185 * targhooks.c (default_have_speculation_safe_value): Add
4188 2018-07-31 David Malcolm <dmalcolm@redhat.com>
4190 * dump-context.h: Include "pretty-print.h".
4191 (dump_context::refresh_dumps_are_enabled): New decl.
4192 (dump_context::emit_item): New decl.
4193 (class dump_context): Add fields "m_test_pp" and
4195 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
4196 (temp_dump_context::get_dumped_text): New decl.
4197 (class temp_dump_context): Add field "m_pp".
4198 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
4199 (dump_context::refresh_dumps_are_enabled): ...and add a test for
4201 (set_dump_file): Update for above change.
4202 (set_alt_dump_file): Likewise.
4203 (dump_loc): New overload, taking a pretty_printer *.
4204 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
4205 to any test pretty-printer.
4206 (make_item_for_dump_gimple_stmt): New function, adapted from
4207 optinfo::add_gimple_stmt.
4208 (dump_context::dump_gimple_stmt): Call it, and use the result,
4209 eliminating the direct usage of dump_file and alt_dump_file in
4210 favor of indirectly using them via emit_item.
4211 (make_item_for_dump_gimple_expr): New function, adapted from
4212 optinfo::add_gimple_expr.
4213 (dump_context::dump_gimple_expr): Call it, and use the result,
4214 eliminating the direct usage of dump_file and alt_dump_file in
4215 favor of indirectly using them via emit_item.
4216 (make_item_for_dump_generic_expr): New function, adapted from
4218 (dump_context::dump_generic_expr): Call it, and use the result,
4219 eliminating the direct usage of dump_file and alt_dump_file in
4220 favor of indirectly using them via emit_item.
4221 (make_item_for_dump_printf_va): New function, adapted from
4222 optinfo::add_printf_va.
4223 (make_item_for_dump_printf): New function.
4224 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
4225 and use the result, eliminating the direct usage of dump_file and
4226 alt_dump_file in favor of indirectly using them via emit_item.
4227 (make_item_for_dump_dec): New function.
4228 (dump_context::dump_dec): Call it, and use the result,
4229 eliminating the direct usage of dump_file and alt_dump_file in
4230 favor of indirectly using them via emit_item.
4231 (make_item_for_dump_symtab_node): New function, adapted from
4232 optinfo::add_symtab_node.
4233 (dump_context::dump_symtab_node): Call it, and use the result,
4234 eliminating the direct usage of dump_file and alt_dump_file in
4235 favor of indirectly using them via emit_item.
4236 (dump_context::begin_scope): Reimplement, avoiding direct usage
4237 of dump_file and alt_dump_file in favor of indirectly using them
4239 (dump_context::emit_item): New member function.
4240 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
4241 Set up test pretty-printer on the underlying context. Call
4242 refresh_dumps_are_enabled.
4243 (temp_dump_context::~temp_dump_context): Call
4244 refresh_dumps_are_enabled.
4245 (temp_dump_context::get_dumped_text): New member function.
4246 (selftest::verify_dumped_text): New function.
4247 (ASSERT_DUMPED_TEXT_EQ): New macro.
4248 (selftest::test_capture_of_dump_calls): Run all tests twice, with
4249 and then without optinfo enabled. Add uses of
4250 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
4251 * dumpfile.h: Update comment for the dump_* API.
4252 * optinfo-emit-json.cc
4253 (selftest::test_building_json_from_dump_calls): Update for new
4254 param for temp_dump_context ctor.
4255 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
4256 and "m_owned" field.
4257 (optinfo_item::~optinfo_item): Likewise.
4258 (optinfo::add_item): New member function.
4259 (optinfo::emit): Update comment.
4260 (optinfo::add_string): Delete.
4261 (optinfo::add_printf): Delete.
4262 (optinfo::add_printf_va): Delete.
4263 (optinfo::add_gimple_stmt): Delete.
4264 (optinfo::add_gimple_expr): Delete.
4265 (optinfo::add_tree): Delete.
4266 (optinfo::add_symtab_node): Delete.
4267 (optinfo::add_dec): Delete.
4268 * optinfo.h (class dump_context): New forward decl.
4269 (optinfo::add_item): New decl.
4270 (optinfo::add_string): Delete.
4271 (optinfo::add_printf): Delete.
4272 (optinfo::add_printf_va): Delete.
4273 (optinfo::add_gimple_stmt): Delete.
4274 (optinfo::add_gimple_expr): Delete.
4275 (optinfo::add_tree): Delete.
4276 (optinfo::add_symtab_node): Delete.
4277 (optinfo::add_dec): Delete.
4278 (optinfo::add_poly_int): Delete.
4279 (optinfo_item::optinfo_item): Remove "owned" param.
4280 (class optinfo_item): Remove field "m_owned".
4282 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4285 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
4286 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
4287 requested variable alignment.
4288 (expand_one_ssa_partition): Likewise.
4289 (expand_one_var): Likewise.
4291 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4293 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
4294 to speculation_safe_value_not_needed.
4296 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4298 * targhooks.h (speculation_safe_value_not_needed): New prototype.
4299 * targhooks.c (speculation_safe_value_not_needed): New function.
4300 * target.def (have_speculation_safe_value): Update documentation.
4301 * doc/tm.texi: Regenerated.
4303 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4305 * config/aarch64/iterators.md (ALLI_TI): New iterator.
4306 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
4308 (despeculate_copy<ALLI:mode>_insn): New insn.
4309 (despeculate_copyti_insn): New insn.
4310 (despeculate_simple<ALLI:mode>): New insn
4311 (despeculate_simpleti): New insn.
4312 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
4314 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
4315 aarch64_speculation_safe_value.
4316 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
4318 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4320 * config/aarch64/aarch64-speculation.cc: New file.
4321 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
4322 before pass_reorder_blocks.
4323 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
4325 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
4326 X14 and X15 when tracking speculation.
4327 * config/aarch64/aarch64.md (register name constants): Add
4328 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
4329 (unspec): Add UNSPEC_SPECULATION_TRACKER.
4330 (speculation_barrier): New insn attribute.
4331 (cmp<mode>): Allow SP in comparisons.
4332 (speculation_tracker): New insn.
4333 (speculation_barrier): Add speculation_barrier attribute.
4334 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
4335 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
4336 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
4338 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4340 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
4341 aarch64_track_speculation is true.
4342 (tb<optab><mode>1): Likewise.
4343 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
4344 generate CB[N]Z when tracking speculation.
4345 (aarch64_split_compare_and_swap): Likewise.
4346 (aarch64_split_atomic_op): Likewise.
4348 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4350 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
4352 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4354 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
4355 (speculation_barrier): New insn.
4357 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4359 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
4360 * config/arm/arm.md (speculation_barrier): New expand.
4361 (speculation_barrier_insn): New pattern.
4363 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4365 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
4366 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
4367 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
4368 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
4370 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
4371 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
4372 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
4373 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
4374 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
4375 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
4376 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
4377 * builtins.c (expand_speculation_safe_value): New function.
4378 (expand_builtin): Call it.
4379 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
4380 * doc/extend.texi: Document __builtin_speculation_safe_value.
4381 * doc/md.texi: Document "speculation_barrier" pattern.
4382 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
4383 TARGET_HAVE_SPECULATION_SAFE_VALUE.
4384 * doc/tm.texi: Regenerated.
4385 * target.def (have_speculation_safe_value, speculation_safe_value): New
4387 * targhooks.c (default_have_speculation_safe_value): New function.
4388 (default_speculation_safe_value): New function.
4389 * targhooks.h (default_have_speculation_safe_value): Add prototype.
4390 (default_speculation_safe_value): Add prototype.
4392 2018-07-31 David Malcolm <dmalcolm@redhat.com>
4394 * dump-context.h (dump_context::dump_loc): New decl.
4395 * dumpfile.c (dump_context::dump_loc): New member function.
4396 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
4397 and dump_gimple_stmt.
4398 (dump_context::dump_gimple_expr_loc): Likewise, using
4400 (dump_context::dump_generic_expr_loc): Likewise, using
4402 (dump_context::dump_printf_loc_va): Likewise, using
4404 (dump_context::begin_scope): Explicitly using the global function
4405 "dump_loc", rather than the member function.
4407 2018-07-31 Martin Sebor <msebor@redhat.com>
4409 PR tree-optimization/86741
4410 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
4412 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
4414 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
4415 depend on whether prefetch instructions will be emitted or not.
4416 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
4417 will be emitted or not.
4418 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
4419 (TARGET_SETMEM_PFD): New macros.
4421 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4423 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
4424 (NULL_STMT_VEC_INFO): Delete.
4425 (stmt_vec_info::operator*): Likewise.
4426 (stmt_vec_info::operator gimple *): Likewise.
4427 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
4428 of NULL_STMT_VEC_INFO.
4429 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4430 (vect_reassociating_reduction_p): Likewise.
4431 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
4432 (vectorizable_store): Likewise.
4433 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
4434 (vec_info::free_stmt_vec_infos): Likewise.
4436 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4438 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
4439 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
4441 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4443 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
4444 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
4445 (vec_info::free_stmt_vec_info): New private member functions.
4446 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
4447 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
4448 * tree-parloops.c (gather_scalar_reductions): Remove calls to
4449 set_stmt_vec_info_vec and free_stmt_vec_infos.
4450 * tree-vect-loop.c (_loop_vec_info): Remove call to
4451 set_stmt_vec_info_vec.
4452 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
4453 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
4454 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
4455 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
4456 (vec_info::free_stmt_vec_info): ...these new functions. Remove
4457 assignments in {vec_info::,}new_stmt_vec_info that are redundant
4458 with the clearing in the xcalloc.
4459 (stmt_vec_info_vec): Delete.
4460 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
4461 (vectorize_loops): Likewise.
4462 (vec_info::~vec_info): Remove argument from call to
4463 free_stmt_vec_infos.
4464 (vec_info::add_stmt): Remove vinfo argument from call to
4467 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4469 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
4470 rather than a gimple stmt.
4471 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
4472 information for pattern statements when passed the original
4473 statement; instead wait to be passed the pattern statement itself.
4474 Don't call set_vinfo_for_stmt here.
4475 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
4476 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
4477 stmt_vec_infos here.
4478 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
4479 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
4480 stmt_vec_infos entry.
4482 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4484 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
4485 * tree-vectorizer.c (vec_info::replace_stmt): New function.
4486 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
4487 * tree-vect-stmts.c (vectorizable_call): Likewise.
4488 (vectorizable_simd_clone_call): Likewise.
4490 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4492 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
4493 * tree-vectorizer.c (vec_info::remove_stmt): New function.
4494 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
4495 * tree-vect-loop.c (vect_transform_loop): Likewise.
4496 * tree-vect-slp.c (vect_schedule_slp): Likewise.
4497 * tree-vect-stmts.c (vect_remove_stores): Likewise.
4499 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4501 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
4502 (vect_dr_stmt): Delete.
4503 * tree-vectorizer.c (vec_info::lookup_dr): New function.
4504 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
4506 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
4507 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
4508 (vect_verify_datarefs_alignment, vect_peeling_supportable)
4509 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
4510 (vect_analyze_data_refs): Likewise.
4511 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
4513 (vect_find_same_alignment_drs): Likewise.
4514 (vect_slp_analyze_node_dependences): Update calls accordingly.
4515 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
4516 instead of DR_VECT_AUX.
4517 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
4518 of a vector data references. Use vec_info::lookup_dr instead of
4520 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
4521 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
4522 instead of DR_VECT_AUX.
4524 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4526 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
4528 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
4530 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4531 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4532 (vect_gen_prolog_loop_niters): Likewise.
4534 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4536 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
4537 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
4538 (vect_known_alignment_in_bytes, vect_dr_behavior)
4539 (vect_get_scalar_dr_size): Take references as dr_vec_infos
4540 instead of data_references. Update calls to other routines for
4541 which the same change has been made.
4542 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
4543 dr_vec_infos instead of stmt_vec_infos.
4544 (vect_analyze_data_ref_dependence): Update call accordingly.
4545 (vect_slp_analyze_data_ref_dependence)
4546 (vect_record_base_alignments): Use DR_VECT_AUX.
4547 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
4548 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
4549 (vector_alignment_reachable_p, vect_get_data_access_cost)
4550 (vect_peeling_supportable, vect_analyze_group_access_1)
4551 (vect_analyze_group_access, vect_analyze_data_ref_access)
4552 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
4553 (vect_compile_time_alias, vect_small_gap_p)
4554 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
4555 (vect_supportable_dr_alignment): Take references as dr_vec_infos
4556 instead of data_references. Update calls to other routines for
4557 which the same change has been made.
4558 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
4559 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
4560 (vect_slp_analyze_and_verify_node_alignment)
4561 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
4562 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
4563 (vect_setup_realignment): Use dr_vec_infos. Update calls after
4565 (_vect_peel_info::dr): Replace with...
4566 (_vect_peel_info::dr_info): ...this new field.
4567 (vect_peeling_hash_get_most_frequent)
4568 (vect_peeling_hash_choose_best_peeling): Update accordingly.
4569 (vect_peeling_hash_get_lowest_cost):
4570 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
4571 routines for which the same change has been made.
4572 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
4574 * tree-vect-loop-manip.c (get_misalign_in_elems)
4575 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
4577 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4578 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
4579 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
4580 (get_group_load_store_type, get_negative_load_store_type)
4581 (vect_get_data_ptr_increment, vectorizable_store)
4582 (vectorizable_load): Likewise.
4583 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
4584 Update calls to other routines for which the same change has been made.
4586 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4588 * tree-vectorizer.h (vec_info::move_dr): New member function.
4589 (dataref_aux): Rename to...
4590 (dr_vec_info): ...this and add "dr" and "stmt" fields.
4591 (_stmt_vec_info::dr_aux): Update accordingly.
4592 (_stmt_vec_info::data_ref_info): Delete.
4593 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
4594 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
4595 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
4596 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
4598 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
4599 (STMT_VINFO_DR_INFO): New macro.
4600 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
4601 (set_dr_misalignment): Update after rename of dataref_aux.
4602 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
4603 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
4604 initialization of STMT_VINFO_DATA_REF.
4605 * tree-vectorizer.c (vec_info::move_dr): New function.
4606 * tree-vect-patterns.c (vect_recog_bool_pattern)
4607 (vect_recog_mask_conversion_pattern)
4608 (vect_recog_gather_scatter_pattern): Use it.
4609 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
4610 the "dr" and "stmt" fields of dr_vec_info instead of
4611 STMT_VINFO_DATA_REF.
4613 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4615 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
4616 (is_pattern_stmt_p): Use it.
4617 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
4618 on pattern statements.
4620 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4622 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
4623 original stmt as a stmt_vec_info rather than a gimple stmt.
4624 (vect_pattern_recog_1): Take the statement directly as a
4625 stmt_vec_info, rather than via a gimple_stmt_iterator.
4626 Update call to vect_mark_pattern_stmts.
4627 (vect_pattern_recog): Update calls accordingly.
4629 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4631 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
4632 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
4633 a vect_def_type for the first argument.
4634 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
4635 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
4636 operand if it isn't defined by a vectorized statement.
4637 (vect_build_gather_load_calls): Remove the mask_dt argument and
4638 update calls to vect_get_vec_def_for_stmt_copy.
4639 (vectorizable_bswap): Likewise the dt argument.
4640 (vectorizable_call): Update calls to vectorizable_bswap and
4641 vect_get_vec_def_for_stmt_copy.
4642 (vectorizable_simd_clone_call, vectorizable_assignment)
4643 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
4644 (vectorizable_comparison): Update calls to
4645 vect_get_vec_def_for_stmt_copy.
4646 (vectorizable_store): Likewise. Remove now-unnecessary calls to
4648 (vect_get_loop_based_defs): Remove dt argument and update call
4649 to vect_get_vec_def_for_stmt_copy.
4650 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
4651 and vect_get_vec_def_for_stmt_copy.
4652 (vectorizable_load): Update calls to vect_build_gather_load_calls
4653 and vect_get_vec_def_for_stmt_copy.
4654 * tree-vect-loop.c (vect_create_epilog_for_reduction)
4655 (vectorizable_reduction, vectorizable_live_operation): Update calls
4656 to vect_get_vec_def_for_stmt_copy.
4658 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4660 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
4661 and gimple stmt arguments with a stmt_vec_info.
4662 (vect_record_base_alignments): Update calls accordingly.
4663 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
4664 and gimple stmt arguments with a stmt_vec_info.
4665 (vect_build_slp_tree_1): Remove vinfo argument and update call
4666 to vect_record_max_nunits.
4667 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
4668 and vect_record_max_nunits.
4670 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4672 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
4673 file and take a stmt_vec_info instead of a gimple stmt.
4674 (supportable_widening_operation, vect_finish_replace_stmt)
4675 (vect_finish_stmt_generation, vect_get_store_rhs)
4676 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
4677 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
4678 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
4679 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
4680 (vect_create_data_ref_ptr, bump_vector_ptr)
4681 (vect_permute_store_chain, vect_setup_realignment)
4682 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
4683 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
4684 (vectorizable_reduction, vectorizable_induction)
4685 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
4686 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
4687 than gimple stmts as arguments.
4688 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
4689 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
4690 (can_group_stmts_p, vect_check_gather_scatter)
4691 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
4692 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
4693 (vect_permute_load_chain, vect_shift_permute_load_chain)
4694 (vect_transform_grouped_load)
4695 (vect_record_grouped_load_vectors): Likewise.
4696 * tree-vect-loop.c (vect_fixup_reduc_chain)
4697 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
4698 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
4699 (vectorizable_reduction, vectorizable_induction)
4700 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
4701 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
4702 (vect_get_load_store_mask): Likewise.
4703 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
4704 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
4705 * tree-vect-stmts.c (vect_mark_relevant)
4706 (is_simple_and_all_uses_invariant)
4707 (exist_non_indexing_operands_for_use_p, process_use)
4708 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
4709 (vect_get_vec_def_for_operand, vect_get_vec_defs)
4710 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
4711 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
4712 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
4713 (get_negative_load_store_type, get_load_store_type)
4714 (vect_check_load_store_mask, vect_check_store_rhs)
4715 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
4716 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4717 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4718 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4719 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
4720 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
4721 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
4722 (supportable_widening_operation): Likewise.
4724 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4726 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
4727 a stmt_vec_info instead of a gcall.
4728 (vect_check_gather_scatter): Update call accordingly.
4729 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
4731 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
4732 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
4733 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
4734 instead of a gimple stmt.
4735 (vect_transform_loop): Update calls accordingly.
4736 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
4737 stmt_vec_infos instead of gimple stmts.
4738 (vect_analyze_slp_instance): Update use accordingly.
4739 * tree-vect-stmts.c (read_vector_array, write_vector_array)
4740 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
4741 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
4742 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
4743 (vect_gen_widened_results_half, vect_get_loop_based_defs)
4744 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
4745 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
4746 down to subroutines.
4748 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4750 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
4751 of the worklist from a vector of gimple stmts to a vector of
4753 * tree-vect-stmts.c (vect_mark_relevant, process_use)
4754 (vect_mark_stmts_to_be_vectorized): Likewise
4756 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4758 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
4759 statement before passing it to vect_analyze_stmt.
4760 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
4761 the chain of phi vector definitions. Track the exit phi via its
4763 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
4764 STMT_VINFO_REDUC_DEF.
4765 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
4766 stmt_vec_infos to handle the statement chains.
4767 (vect_get_slp_defs): Record the first statement in the node
4768 using a stmt_vec_info.
4769 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
4770 statements here and pass their stmt_vec_info down to subroutines.
4771 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
4772 down to vect_finish_stmt_generation.
4773 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
4774 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
4775 stmt_vec_infos to subroutines.
4776 (vect_remove_stores): Use stmt_vec_infos to handle the statement
4779 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4781 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
4782 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
4783 (vect_permute_store_chain, vect_setup_realignment)
4784 (vect_permute_load_chain, vect_shift_permute_load_chain)
4785 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
4786 stmts internally, and when passing values to other vectorizer routines.
4787 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
4788 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
4789 (vect_analyze_loop_operations, get_initial_def_for_reduction)
4790 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
4791 (vectorizable_reduction, vectorizable_induction)
4792 (vectorizable_live_operation, vect_transform_loop_stmt)
4793 (vect_transform_loop): Likewise.
4794 * tree-vect-patterns.c (vect_reassociating_reduction_p)
4795 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
4796 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
4797 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4798 (vect_slp_analyze_node_operations_1): Likewise.
4799 * tree-vect-stmts.c (vect_mark_relevant, process_use)
4800 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
4801 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
4802 (vect_finish_stmt_generation_1, get_group_load_store_type)
4803 (get_load_store_type, vect_build_gather_load_calls)
4804 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4805 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4806 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4807 (vectorizable_store, vectorizable_load, vectorizable_condition)
4808 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
4809 (supportable_widening_operation): Likewise.
4810 (vect_get_vector_types_for_stmt): Likewise.
4811 * tree-vectorizer.h (vect_dr_behavior): Likewise.
4813 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4815 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4816 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
4817 (vect_permute_store_chain, vect_permute_load_chain)
4818 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
4819 repeated stmt_vec_info lookups.
4820 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
4821 (vect_update_ivs_after_vectorizer): Likewise.
4822 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
4823 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
4824 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
4825 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4826 (vect_bb_slp_scalar_cost): Likewise.
4827 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
4829 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4831 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
4832 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
4833 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
4834 to get gassigns and gcalls, rather than operating on generc gimple
4836 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
4837 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
4838 (vectorizable_load, vect_analyze_stmt): Likewise.
4839 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
4841 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4843 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
4844 return stmt_vec_infos rather than gimple stmts. Do not accept
4846 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
4848 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
4849 Update use of get_later_stmt.
4850 (vect_get_constant_vectors): Update call accordingly.
4851 (vect_schedule_slp_instance): Likewise
4852 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
4853 (vect_slp_analyze_instance_dependence): Likewise.
4854 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
4856 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4858 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
4859 (stmt_info_for_cost::stmt_info): ...this new field.
4860 (add_stmt_costs): Update accordingly.
4861 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
4862 (vect_get_known_peeling_cost): Likewise.
4863 (vect_estimate_min_profitable_iters): Likewise.
4864 * tree-vect-stmts.c (record_stmt_cost): Likewise.
4866 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4868 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
4869 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4870 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
4872 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
4874 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4876 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
4877 a gimple stmt to a stmt_vec_info.
4878 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
4880 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4882 * tree-vectorizer.h (vec_info::grouped_stores): Change from
4883 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4884 (_loop_vec_info::reduction_chains): Likewise.
4885 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
4887 * tree-vect-slp.c (vect_analyze_slp): Likewise.
4889 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4891 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
4892 a gimple stmt to a stmt_vec_info.
4893 (_stmt_vec_info::next_element): Likewise.
4894 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
4895 (vect_slp_analyze_and_verify_node_alignment)
4896 (vect_analyze_group_access_1, vect_analyze_group_access)
4897 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
4898 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
4899 (vect_supportable_dr_alignment): Update accordingly.
4900 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
4901 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
4902 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
4903 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
4904 * tree-vect-slp.c (vect_build_slp_tree_1)
4905 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
4906 (vect_split_slp_store_group, vect_analyze_slp_instance)
4907 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
4908 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
4909 (get_group_load_store_type, get_load_store_type)
4910 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
4911 (vect_transform_stmt, vect_remove_stores): Likewise.
4913 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4915 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
4917 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4918 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
4919 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
4920 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
4921 (vector_alignment_reachable_p, vect_get_data_access_cost)
4922 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
4923 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
4924 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
4925 (vect_analyze_group_access_1, vect_analyze_group_access)
4926 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
4927 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
4928 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
4929 result of vect_dr_stmt and use the stmt_vec_info instead of
4930 the associated gimple stmt.
4931 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4932 (vect_gen_prolog_loop_niters): Likewise.
4933 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4935 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4937 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
4938 to a vec<stmt_vec_info>.
4939 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
4940 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
4942 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
4943 to a vec<stmt_vec_info>.
4944 (bst_traits::value_type, bst_traits::value_type): Likewise.
4945 (bst_traits::hash): Update accordingly.
4946 (vect_get_and_check_slp_defs): Change the stmts parameter from
4947 a vec<gimple *> to a vec<stmt_vec_info>.
4948 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
4949 (vect_build_slp_tree): Likewise.
4950 (vect_build_slp_tree_2): Likewise. Update uses of
4951 SLP_TREE_SCALAR_STMTS.
4952 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
4953 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
4954 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
4955 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
4956 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
4957 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
4958 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
4959 (vect_get_constant_vectors, vect_get_slp_defs)
4960 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
4961 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
4962 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
4963 instead of gimple stmts.
4964 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
4965 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
4966 (vect_slp_analyze_instance_dependence): Update uses of
4967 SLP_TREE_SCALAR_STMTS.
4968 (vect_slp_analyze_and_verify_node_alignment): Likewise.
4969 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
4970 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
4971 (get_initial_defs_for_reduction): Likewise.
4972 (vect_create_epilog_for_reduction): Likewise.
4973 (vectorize_fold_left_reduction): Likewise.
4974 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
4975 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
4976 (can_vectorize_live_stmts): Likewise.
4978 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4980 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
4981 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4982 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
4984 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
4986 (gather_scalar_reductions): Update after above interface changes.
4987 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
4988 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
4990 (vect_force_simple_reduction): Likewise.
4991 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
4992 LOOP_VINFO_REDUCTIONS.
4993 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4995 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4997 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
4998 a gimple stmt to a stmt_vec_info.
4999 * tree-vect-loop.c (vect_active_double_reduction_p)
5000 (vect_force_simple_reduction, vectorizable_reduction): Update
5003 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5005 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
5006 vec<gimple *> to a vec<stmt_vec_info>.
5007 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
5008 the reduction_phis argument from a vec<gimple *> to a
5010 (vectorizable_reduction): Likewise the phis local variable that
5011 is passed to vect_create_epilog_for_reduction. Update for new type
5012 of SLP_TREE_VEC_STMTS.
5013 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
5014 (vectorizable_live_operation): Likewise.
5015 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
5016 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
5018 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5020 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
5021 a gimple stmt to a stmt_vec_info.
5022 (vectorizable_condition, vectorizable_live_operation)
5023 (vectorizable_reduction, vectorizable_induction): Pass back the
5024 vectorized statement as a stmt_vec_info.
5025 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
5026 use of STMT_VINFO_VEC_STMT.
5027 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
5028 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
5029 as stmt_vec_infos rather than gimple stmts.
5030 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
5032 (vectorizable_live_operation): Likewise.
5033 (vectorizable_reduction, vectorizable_induction): Likewise,
5034 updating use of STMT_VINFO_VEC_STMT.
5035 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
5036 of STMT_VINFO_VEC_STMT.
5037 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
5038 (vectorizable_simd_clone_call, vectorizable_conversion)
5039 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
5040 (vectorizable_store, vectorizable_load, vectorizable_condition)
5041 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
5042 from a gimple stmt to a stmt_vec_info.
5043 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
5044 pointer to a stmt_vec_info to the vectorizable_* routines.
5046 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5048 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
5049 a gimple stmt to a stmt_vec_info.
5050 (is_pattern_stmt_p): Update accordingly.
5051 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
5052 (vect_record_grouped_load_vectors): Likewise.
5053 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
5054 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
5055 (vect_model_reduction_cost): Likewise.
5056 (vect_create_epilog_for_reduction): Likewise.
5057 (vectorizable_reduction, vectorizable_induction): Likewise.
5058 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5059 Return the stmt_vec_info for the pattern statement.
5060 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
5061 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
5062 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
5063 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
5064 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
5065 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
5066 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
5067 (free_stmt_vec_info, vect_is_simple_use): Likewise.
5069 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5071 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
5072 (vect_finish_stmt_generation): Likewise.
5073 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
5074 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
5075 (vect_build_gather_load_calls): Use the return value of the above
5076 functions instead of a separate call to vinfo_for_stmt. Use narrow
5077 scopes for the input gimple stmt and wider scopes for the associated
5078 stmt_vec_info. Use vec_info::lookup_def when setting these
5079 stmt_vec_infos from an SSA_NAME definition.
5080 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
5081 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
5082 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
5083 (vectorizable_store, vectorizable_load, vectorizable_condition)
5084 (vectorizable_comparison): Likewise.
5085 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
5086 (vectorizable_reduction): Likewise.
5088 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5090 * tree-vectorizer.h (vect_is_simple_use): Add an optional
5091 stmt_vec_info * parameter before the optional gimple **.
5092 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
5093 (process_use, vect_get_vec_def_for_operand_1): Update callers.
5094 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
5095 * tree-vect-loop.c (vectorizable_reduction): Likewise.
5096 (vectorizable_live_operation): Likewise.
5097 * tree-vect-patterns.c (type_conversion_p): Likewise.
5098 (vect_look_through_possible_promotion): Likewise.
5099 (vect_recog_rotate_pattern): Likewise.
5100 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
5102 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5104 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
5105 a typedef to a wrapper class.
5106 (NULL_STMT_VEC_INFO): New macro.
5107 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
5108 (stmt_vec_info::operator*): New function.
5109 (stmt_vec_info::operator gimple *): Likewise.
5110 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
5111 (add_stmt_costs): Likewise.
5112 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
5113 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
5114 (vect_get_known_peeling_cost): Likewise.
5115 (vect_estimate_min_profitable_iters): Likewise.
5116 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5117 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
5118 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
5119 (vectorizable_store, free_stmt_vec_infos): Likewise.
5120 (new_stmt_vec_info): Change return type of xcalloc to
5123 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5125 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
5126 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
5127 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
5128 a single_imm_use-based sequence.
5129 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
5131 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5133 * tree-vectorizer.h (vec_info::lookup_def): Declare.
5134 * tree-vectorizer.c (vec_info::lookup_def): New function.
5135 * tree-vect-patterns.c (vect_get_internal_def): Use it.
5136 (vect_widened_op_tree): Likewise.
5137 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
5138 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
5139 (vectorizable_reduction): Likewise.
5140 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
5142 (vect_is_slp_reduction): Update calls accordingly. Use
5143 vec_info::lookup_def.
5144 (vect_is_simple_reduction): Likewise
5145 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
5147 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5149 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
5150 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
5151 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
5153 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
5154 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
5155 (vect_update_vf_for_slp, vect_analyze_loop_operations)
5156 (vect_is_slp_reduction, vectorizable_induction)
5157 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
5158 * tree-vect-patterns.c (vect_init_pattern_stmt):
5159 (vect_determine_min_output_precision_1, vect_determine_precisions)
5160 (vect_pattern_recog): Likewise.
5161 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
5162 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
5163 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
5164 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
5165 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
5166 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
5167 info field from a loop to a loop_vec_info.
5169 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5171 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
5172 (vec_info::add_stmt): Declare.
5173 * tree-vectorizer.c (vec_info::add_stmt): New function.
5174 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
5175 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
5176 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
5177 (vectorizable_induction): Likewise.
5178 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
5179 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
5180 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
5181 (vectorizable_load): Likewise.
5182 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5183 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
5184 (vect_recog_gather_scatter_pattern): Likewise.
5185 (append_pattern_def_seq): Likewise. Remove a check that is
5186 performed by add_stmt itself.
5188 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5190 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
5191 which make_ssa_name was called with new_stmt before new_stmt
5194 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5196 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
5198 (vect_is_slp_reduction): ...here...
5199 (vect_is_simple_reduction): ...and here. Remove repetition of tests
5200 that are already known to be false.
5202 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5204 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
5205 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
5206 STMT_VINFO_NUM_SLP_USES when it's true.
5207 (vect_free_slp_instance): Add a final_p parameter and pass it to
5209 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
5210 (vect_analyze_slp_instance): Likewise.
5211 (vect_slp_analyze_operations): Likewise.
5212 (vect_slp_analyze_bb_1): Likewise.
5213 * tree-vectorizer.c (vec_info): Likewise.
5214 * tree-vect-loop.c (vect_transform_loop): Likewise.
5216 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5218 * tree-vect-loop.c (vectorizable_reduction): Assert that the
5219 function is not called for second and subsequent members of
5222 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5224 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
5225 cases for nested loops from here to ...
5226 (vect_create_epilog_for_reduction): ...here. Only call
5227 vect_is_simple_use for inner-loop reductions.
5229 2018-07-31 Martin Liska <mliska@suse.cz>
5231 PR gcov-profile/85338
5232 PR gcov-profile/85350
5233 PR gcov-profile/85372
5234 * profile.c (struct location_triplet): New.
5235 (struct location_triplet_hash): Likewise.
5236 (output_location): Do not output a BB that
5237 is already recorded for a line.
5238 (branch_prob): Use streamed_locations.
5240 2018-07-31 Martin Liska <mliska@suse.cz>
5242 PR gcov-profile/85370
5243 * coverage.c (coverage_begin_function): Do not mark target
5244 clones as artificial functions.
5246 2018-07-31 Martin Liska <mliska@suse.cz>
5248 PR gcov-profile/83813
5249 PR gcov-profile/84758
5250 PR gcov-profile/85217
5251 PR gcov-profile/85332
5252 * profile.c (branch_prob): Do not record GOTO expressions
5253 for GIMPLE statements which locations are already streamed.
5255 2018-07-31 Olivier Hainque <hainque@adacore.com>
5257 * gcc.c (handle_spec_function): Accept a soft_matched_part
5258 argument, as do_spec_1. Pass it down to ...
5259 (eval_spec_function): Accept a soft_matched_part argument,
5260 and pass it down to ...
5261 (do_spec_2): Accept a soft_matched_part argument, and pass
5262 it down to do_spec_1.
5263 (do_spec_1): Pass soft_matched_part to handle_spec_function.
5264 (handle_braces): Update call to handle_spec_function.
5265 (driver::set_up_specs): Update calls to do_spec_2.
5266 (compare_debug_dump_opt_spec_function): Likewise.
5267 (compare_debug_self_opt_spec_function): Likewise.
5269 2018-07-31 Olivier Hainque <hainque@adacore.com>
5271 * common.opt (nolibc): New option.
5272 * doc/invoke.texi (Link Options): Document it.
5273 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
5274 * config/alpha/linux.h: Likewise.
5275 * config/arc/elf.h: Likewise.
5276 * config/arm/uclinux-elf.h: Likewise.
5277 * config/arm/unknown-elf.h: Likewise.
5278 * config/avr/avrlibc.h: Likewise.
5279 * config/bfin/bfin.h: Likewise.
5280 * config/bfin/linux.h: Likewise.
5281 * config/bfin/uclinux.h: Likewise.
5282 * config/darwin.h: Likewise.
5283 * config/darwin10.h: Likewise.
5284 * config/darwin12.h: Likewise.
5285 * config/gnu-user.h: Likewise.
5286 * config/lm32/uclinux-elf.h: Likewise.
5287 * config/pa/pa-hpux11.h: Likewise.
5288 * config/pa/pa64-hpux.h: Likewise.
5289 * config/sparc/sparc.h: Likewise.
5291 2018-07-31 Olivier Hainque <hainque@adacore.com>
5293 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
5294 undefined variables.
5296 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
5299 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
5302 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
5304 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
5305 terminated string literal.
5307 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
5309 PR rtl-optimization/85160
5310 * combine.c (is_just_move): New function.
5311 (try_combine): Allow combining two instructions into two if neither of
5312 the original instructions was a move.
5314 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
5317 * doc/extend.texi (Global Register Variables): Discourage use of type
5319 (Local Register Variables): Likewise.
5321 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
5323 PR tree-optimization/86506
5324 * hwint.h (ceil_log2): Resync with hwint.c implementation.
5326 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
5329 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
5330 hard_regno, make sure no insn between `from` and `to` clobbers it.
5332 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
5333 Tom de Vries <tdevries@suse.de>
5335 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
5336 (PTX_DEFAULT_RUNTIME_DIM): ... this.
5337 (nvptx_goacc_validate_dims): Set default worker and gang dims to
5338 PTX_DEFAULT_RUNTIME_DIM.
5339 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
5341 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
5343 * config/pa/pa.c (pa_output_addr_vec): Align address table.
5344 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
5345 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
5347 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
5349 * config/rs6000/constraints.md (wG constraint): Delete, no longer
5351 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
5352 predicate to reflect toc fusion has been deleted.
5353 (toc_fusion_mem_raw): Delete, no longer used.
5354 (toc_fusion_mem_wrapped): Likewise.
5355 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
5357 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
5358 Delete, no longer used.
5359 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
5360 meant to be used for toc fusion.
5361 (rs6000_debug_print_mode): Delete toc fusion debugging.
5362 (rs6000_debug_reg_global): Likewise.
5363 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
5364 fusion and secondary reload support that were never used.
5365 (rs6000_option_override_internal): Delete TOC fusion, that was only
5366 partially defined, and it did not work unless you also used the
5368 (rs6000_legitimate_address_p): Delete TOC fusion support.
5369 (rs6000_opt_masks): Likewise.
5370 (fusion_wrap_memory_address): Delete function, no longer used.
5371 (fusion_split_address); Delete TOC fusion support.
5372 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
5373 longer used with toc fusion being deleted.
5374 (TARGET_TOC_FUSION_FP): Likewise.
5375 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
5377 (toc fusion spliter): Delete TOC fusion support.
5378 (toc_fusionload_<mode>): Likewise.
5379 (toc_fusionload_di): Likewise.
5380 (fusion_gpr_load_<mode>): Delete generator function, this insn no
5381 longer needs to be named. Rename predicate to delete TOC fusion.
5382 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
5383 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
5384 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
5385 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
5386 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
5388 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
5390 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
5391 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
5392 __int128 in built-in function prototypes.
5393 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
5394 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
5396 2018-07-27 Martin Sebor <msebor@redhat.com>
5398 PR tree-optimization/86696
5399 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
5400 types, including enums.
5401 (handle_char_store): Be prepared for the above function to fail.
5403 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
5405 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
5406 when optimization level is lower than 2 or optimize for size.
5408 2018-07-26 Martin Sebor <msebor@redhat.com>
5410 PR tree-optimization/86043
5411 PR tree-optimization/86042
5412 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
5413 (get_string_cst_length): Rename...
5414 (get_min_string_length): ...to this. Add argument.
5415 (handle_char_store): Extend to handle multi-character stores by
5417 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
5418 * tree.h (initializer_zerop): Add argument.
5420 2018-07-26 Jakub Jelinek <jakub@redhat.com>
5423 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
5424 declare target to variables if they have always,{to,from,tofrom} map
5427 2018-07-26 Martin Liska <mliska@suse.cz>
5430 * lto-wrapper.c: Add linker_output as prefix
5431 for ltrans_output_file.
5433 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
5435 PR rtl-optimization/85805
5436 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
5437 value for hard registers if that was written in the same mode.
5439 2018-07-26 Martin Liska <mliska@suse.cz>
5441 PR gcov-profile/86536
5442 * gcov.c (format_gcov): Use printf format %.*f directly
5443 and do not handle special values.
5445 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5447 * common/config/arc/arc-common.c (arc_option_optimization_table):
5448 Update default optimizations for size.
5450 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5452 * config/arc/arc.md (movsf_insn): Add short instruction selection.
5453 * config/arc/constraints.md (CfZ): New constraint.
5454 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
5455 (subsf3_fpu): Likewise.
5456 (cmpsf_fpu): Likewise.
5457 (cmpsf_fpu_uneq): Likewise.
5459 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5461 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
5463 (arc_is_uncached_mem_p): uncached applies to both the variable and
5466 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5468 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
5471 2018-07-25 David Malcolm <dmalcolm@redhat.com>
5473 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
5474 field "m_scopes" from vec to auto_vec.
5476 2018-07-25 Martin Liska <mliska@suse.cz>
5478 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
5481 2018-07-25 Richard Biener <rguenther@suse.de>
5484 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
5485 special wrt context_die late.
5486 (gen_subprogram_die): Re-use DIEs in local scope.
5488 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
5490 PR tree-optimization/86644
5491 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
5493 2018-07-25 Martin Liska <mliska@suse.cz>
5496 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
5497 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
5499 2018-07-25 Martin Liska <mliska@suse.cz>
5502 * params.def: Explain ASan abbreviation and provide
5503 a documentation link.
5505 2018-07-24 Martin Sebor <msebor@redhat.com>
5507 PR tree-optimization/86622
5508 PR tree-optimization/86532
5509 * builtins.h (string_length): Declare.
5510 * builtins.c (c_strlen): Correct handling of non-constant offsets.
5511 (check_access): Be prepared for non-constant length ranges.
5512 (string_length): Make extern.
5513 * expr.c (string_constant): Only handle the minor non-constant
5514 array index. Use string_constant to compute the length of
5515 a generic string constant.
5517 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
5519 PR tree-optimization/86618
5520 * tree-vect-stmts.c (vectorizable_call): Don't take the address
5521 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
5523 2018-07-24 David Malcolm <dmalcolm@redhat.com>
5525 PR tree-optimization/86636
5526 * json.cc (json::object::set): Fix comment. Add assertions.
5527 (json::array::append): Move here from json.h. Add comment and an
5529 (json::string::string): Likewise.
5530 * json.h (json::array::append): Move to json.cc.
5531 (json::string::string): Likewise.
5532 * optinfo-emit-json.cc
5533 (optrecord_json_writer::impl_location_to_json): Assert that we
5534 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
5535 wrapper around it. Expand the location once, rather than three
5537 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
5538 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
5540 (optrecord_json_writer::optinfo_to_json): Likewise, in four
5541 places. Fix some overlong lines.
5543 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
5545 * config/aarch64/aarch64-simd.md
5546 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
5547 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
5548 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
5549 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
5550 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
5551 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
5552 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
5553 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
5554 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
5556 2018-07-24 Jakub Jelinek <jakub@redhat.com>
5559 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
5560 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
5561 and abs_d == d, do the power of two handling if profitable.
5563 2018-07-24 Richard Biener <rguenther@suse.de>
5565 * match.pd: Add BIT_FIELD_REF canonicalizations.
5567 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
5570 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
5572 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
5574 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
5575 terminated STRING_CST object.
5577 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
5579 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
5581 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
5583 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
5584 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
5585 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
5586 the elements into a register.
5587 (rs6000_split_v4si_init_di_reg): Delete.
5588 (rs6000_split_v4si_init): Delete.
5589 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
5590 (vsx_init_v4si): Rewrite as a define_expand.
5592 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
5594 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
5595 zero_extend argument from memory): New.
5597 2018-07-22 Martin Sebor <msebor@redhat.com>
5600 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
5601 diagnosing calls with unknown arguments unless -Walloca-larger-than
5602 is restricted to less than PTRDIFF_MAX bytes.
5604 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
5606 * doc/gcov.texi (Invoking Gcov): Editorial changes.
5608 2018-07-20 David Malcolm <dmalcolm@redhat.com>
5610 * pretty-print.c (text_info::set_location): Remove redundant
5611 "line_table" parameter from call to rich_location::set_range.
5613 2018-07-20 Martin Sebor <msebor@redhat.com>
5616 * builtins.c (expand_builtin_alloca): Adjust.
5617 * calls.c (alloc_max_size): Simplify.
5618 * cgraphunit.c (cgraph_node::expand): Adjust.
5619 * common.opt (larger_than_size, warn_frame_larger_than): Remove
5621 (frame_larger_than_size): Same.
5622 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
5623 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
5625 * doc/invoke.texi (GCC Command Options): Document option arguments.
5626 Explain byte-size arguments and suffixes.
5627 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
5628 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
5629 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
5630 * doc/options.texi (UInteger): Expand.
5631 (Host_Wide_Int, ByteSize): Document new properties.
5632 * final.c (final_start_function_1): Include sizes in an error message.
5633 * function.c (frame_offset_overflow): Same.
5634 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
5635 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
5636 Diagnose unbounded alloca calls only for limits of less than
5638 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
5639 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
5641 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
5642 only for limits of less than PTRDIFF_MAX.
5643 * langhooks-def.h (lhd_handle_option): Change function argument
5645 * langhooks.c (lhd_handle_option): Same.
5646 * langhooks.h (handle_option): Same.
5647 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
5649 (var_type, var_type_struct): Same.
5650 (var_set): Handle ByteSize flag.
5651 * optc-gen.awk: Add comments to output to ease debugging. Make
5652 use of HOST_WIDE_INT where appropriate.
5653 * opts-gen-save.awk: Use %lx to format unsigned long.
5654 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
5655 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
5656 arguments. Parse bytes-size suffixes.
5657 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
5658 (enum_value_to_arg): Same.
5659 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
5660 (handle_option): Adjust.
5661 (generate_option): Change function argument to HOST_WIDE_INT.
5662 (cmdline_handle_error): Adjust.
5663 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
5664 (set_option): Change function argument to HOST_WIDE_INT.
5665 (option_enabled): Handle cl_host_wide_int.
5666 (get_option_state): Handle CLVC_SIZE.
5667 (control_warning_option): Same.
5668 * opts.c (common_handle_option): Change function argument to
5669 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
5670 OPT_Wvla_larger_than_.
5671 * opts.h (enum cl_var_type): Add an enumerator.
5672 * stor-layout.c (layout_decl): Print a more meaningful warning.
5673 * toplev.c (output_stack_usage): Adjust.
5675 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
5677 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
5678 call to inline_expand_builtin_string_cmp.
5679 (expand_builtin_strcmp): Likewise.
5680 (expand_builtin_strncmp): Likewise.
5681 (inline_string_cmp): Delete the last parameter, change char_type_node
5682 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
5684 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
5685 the inlining expansion on target where the type of the call has same or
5686 narrower precision than unsigned char.
5688 2018-07-20 David Malcolm <dmalcolm@redhat.com>
5690 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
5691 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
5692 * common.opt (fsave-optimization-record): New option.
5693 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
5694 * doc/invoke.texi (-fsave-optimization-record): New option.
5695 * dumpfile.c: Include "optinfo-emit-json.h".
5696 (struct kv_pair): Move to coretypes.h.
5697 (optgroup_options): Make non-static.
5698 (dump_context::end_scope): Call
5699 optimization_records_maybe_pop_dump_scope.
5700 * dumpfile.h (optgroup_options): New decl.
5701 * json.cc: New file.
5703 * optinfo-emit-json.cc: New file.
5704 * optinfo-emit-json.h: New file.
5705 * optinfo.cc: Include "optinfo-emit-json.h".
5706 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
5707 (optinfo_enabled_p): Check optimization_records_enabled_p.
5708 (optinfo_wants_inlining_info_p): Likewise.
5709 * optinfo.h: Update comment.
5710 * profile-count.c (profile_quality_as_string): New function.
5711 * profile-count.h (profile_quality_as_string): New decl.
5712 (profile_count::quality): New accessor.
5713 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
5714 and optinfo_emit_json_cc_tests.
5715 * selftest.h (selftest::json_cc_tests): New decl.
5716 (selftest::optinfo_emit_json_cc_tests): New decl.
5717 * toplev.c: Include "optinfo-emit-json.h".
5718 (compile_file): Call optimization_records_finish.
5719 (do_compile): Call optimization_records_start.
5720 * tree-ssa-live.c: Include optinfo.h.
5721 (remove_unused_scope_block_p): Retain inlining information if
5722 optinfo_wants_inlining_info_p returns true.
5724 2018-07-20 Richard Biener <rguenther@suse.de>
5727 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
5728 to cover -flto-partition=none.
5730 2018-07-20 Martin Liska <mliska@suse.cz>
5732 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
5733 (get_decl_source_range): Remove unused function.
5735 2018-07-20 Richard Biener <rguenther@suse.de>
5737 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
5738 (struct vn_phi_s): Likewise.
5739 (struct vn_reference_s): Likewise.
5740 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
5741 for searching the slot of an entry known to be in the hash itself.
5742 (vn_phi_hasher::equal): Likewise.
5743 (vn_reference_hasher::equal): Likewise.
5744 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
5746 (optimistic_info, current_info): Remove, keeping only valid_info.
5747 (vn_reference_lookup_1): Remove fallback lookup.
5748 (vn_reference_lookup_2): Likewise.
5749 (vn_nary_op_lookup_1): Likewise.
5750 (vn_phi_lookup): Likewise.
5751 (vn_nary_build_or_lookup_1): Make sure to not chain the built
5753 (vn_reference_insert): Adjust, chain the inserted hash element
5754 at last_inserted_ref.
5755 (vn_reference_insert_pieces): Likewise.
5756 (visit_reference_op_call): Likewise.
5757 (vn_nary_op_insert_into): Chain the inserted hash element at
5759 (vn_nary_op_insert_pieces): Adjust.
5760 (vn_nary_op_insert): Likewise.
5761 (vn_nary_op_insert_stmt): Likewise.
5762 (vn_phi_insert): Adjust, chain the inserted hash element at
5764 (process_scc): Remove clearing and copying the optimistic
5765 table. Instead remove elements inserted during an optimistic
5766 iteration from the single table we maintain.
5767 (init_scc_vn): Adjust.
5768 (free_scc_vn): Likewise.
5769 (sccvn_dom_walker::record_cond): Likewise.
5770 (sccvn_dom_walker::after_dom_children): Likewise.
5772 2018-07-19 Martin Sebor <msebor@redhat.com>
5774 PR tree-optimization/84047
5775 PR tree-optimization/83776
5776 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
5777 (check_array_bounds): Call it.
5779 2018-07-19 Martin Sebor <msebor@redhat.com>
5781 * align.h (align_flags): Use member initialization.
5783 2018-07-19 David Malcolm <dmalcolm@redhat.com>
5785 * Makefile.in (OBJS): Add optinfo.o.
5786 * coretypes.h (class symtab_node): New forward decl.
5787 (struct cgraph_node): New forward decl.
5788 (class varpool_node): New forward decl.
5789 * dump-context.h: New file.
5790 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
5792 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
5793 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
5794 (set_alt_dump_file): Likewise.
5795 (dump_context::~dump_context): New dtor.
5796 (dump_gimple_stmt): Move implementation to...
5797 (dump_context::dump_gimple_stmt): ...this new member function.
5798 Add the stmt to any pending optinfo, creating one if need be.
5799 (dump_gimple_stmt_loc): Move implementation to...
5800 (dump_context::dump_gimple_stmt_loc): ...this new member function.
5801 Start a new optinfo and add the stmt to it.
5802 (dump_gimple_expr): Move implementation to...
5803 (dump_context::dump_gimple_expr): ...this new member function.
5804 Add the stmt to any pending optinfo, creating one if need be.
5805 (dump_gimple_expr_loc): Move implementation to...
5806 (dump_context::dump_gimple_expr_loc): ...this new member function.
5807 Start a new optinfo and add the stmt to it.
5808 (dump_generic_expr): Move implementation to...
5809 (dump_context::dump_generic_expr): ...this new member function.
5810 Add the tree to any pending optinfo, creating one if need be.
5811 (dump_generic_expr_loc): Move implementation to...
5812 (dump_context::dump_generic_expr_loc): ...this new member
5813 function. Add the tree to any pending optinfo, creating one if
5815 (dump_printf): Move implementation to...
5816 (dump_context::dump_printf_va): ...this new member function. Add
5817 the text to any pending optinfo, creating one if need be.
5818 (dump_printf_loc): Move implementation to...
5819 (dump_context::dump_printf_loc_va): ...this new member function.
5820 Start a new optinfo and add the stmt to it.
5821 (dump_dec): Move implementation to...
5822 (dump_context::dump_dec): ...this new member function. Add the
5823 value to any pending optinfo, creating one if need be.
5824 (dump_context::dump_symtab_node): New member function.
5825 (dump_context::get_scope_depth): New member function.
5826 (dump_context::begin_scope): New member function.
5827 (dump_context::end_scope): New member function.
5828 (dump_context::ensure_pending_optinfo): New member function.
5829 (dump_context::begin_next_optinfo): New member function.
5830 (dump_context::end_any_optinfo): New member function.
5831 (dump_context::s_current): New global.
5832 (dump_context::s_default): New global.
5833 (dump_scope_depth): Delete global.
5834 (dumpfile_ensure_any_optinfo_are_flushed): New function.
5835 (dump_symtab_node): New function.
5836 (get_dump_scope_depth): Reimplement in terms of dump_context.
5837 (dump_begin_scope): Likewise.
5838 (dump_end_scope): Likewise.
5839 (selftest::temp_dump_context::temp_dump_context): New ctor.
5840 (selftest::temp_dump_context::~temp_dump_context): New dtor.
5841 (selftest::verify_item): New function.
5842 (ASSERT_IS_TEXT): New macro.
5843 (ASSERT_IS_TREE): New macro.
5844 (ASSERT_IS_GIMPLE): New macro.
5845 (selftest::test_capture_of_dump_calls): New test.
5846 (selftest::dumpfile_c_tests): Call it.
5847 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
5848 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
5849 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
5850 descriptive comment.
5851 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
5852 (dump_node, dump_bb): Move these unrelated decls.
5853 (class dump_manager): Add leading comment.
5854 * optinfo.cc: New file.
5855 * optinfo.h: New file.
5857 2018-07-19 Michael Collison <michael.collison@arm.com>
5858 Richard Henderson <rth@redhat.com>
5860 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
5861 (subti): Handle op1 zero.
5862 (subvti4, usub4ti4): New.
5863 (*sub<GPI>3_compare1_imm): New.
5864 (sub<GPI>3_carryinCV): New.
5865 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
5866 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
5868 2018-07-19 Michael Collison <michael.collison@arm.com>
5869 Richard Henderson <rth@redhat.com>
5871 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
5872 (addti3): Create simpler code if low part is already known to be 0.
5873 (addvti4, uaddvti4): New.
5874 (*add<GPI>3_compareC_cconly_imm): New.
5875 (*add<GPI>3_compareC_cconly): New.
5876 (*add<GPI>3_compareC_imm): New.
5877 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
5878 handle constants within this pattern..
5879 (*add<GPI>3_compareV_cconly_imm): New.
5880 (*add<GPI>3_compareV_cconly): New.
5881 (*add<GPI>3_compareV_imm): New.
5882 (add<GPI>3_compareV): New.
5883 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
5884 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
5885 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
5886 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
5887 with 'comparison' operator.
5888 (*add<GPI>3_compareV_cconly_imm): Ditto.
5889 (*add<GPI>3_compareV_cconly): Ditto.
5890 (*add<GPI>3_compareV_imm): Ditto.
5891 (add<GPI>3_compareV): Ditto.
5892 (add<mode>3_carryinC): Ditto.
5893 (*add<mode>3_carryinC_zero): Ditto.
5894 (*add<mode>3_carryinC): Ditto.
5895 (add<mode>3_carryinV): Ditto.
5896 (*add<mode>3_carryinV_zero): Ditto.
5897 (*add<mode>3_carryinV): Ditto.
5899 2018-07-19 Michael Collison <michael.collison@arm.com>
5900 Richard Henderson <rth@redhat.com>
5902 * config/aarch64/aarch64-modes.def (CC_V): New.
5903 * config/aarch64/aarch64-protos.h
5904 (aarch64_addti_scratch_regs): Declare
5905 (aarch64_subvti_scratch_regs): Declare.
5906 (aarch64_expand_subvti): Declare.
5907 (aarch64_gen_unlikely_cbranch): Declare
5908 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
5909 for signed overflow using CC_Vmode.
5910 (aarch64_get_condition_code_1): Handle CC_Vmode.
5911 (aarch64_gen_unlikely_cbranch): New function.
5912 (aarch64_addti_scratch_regs): New function.
5913 (aarch64_subvti_scratch_regs): New function.
5914 (aarch64_expand_subvti): New function.
5916 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
5918 * config/aarch64/aarch64-option-extensions.def: New entry for profile
5920 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
5921 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
5924 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
5927 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5928 address check not strict.
5930 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
5932 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
5934 (store_pair_lanes<mode>): Likewise.
5935 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
5936 enum value 'ADDR_QUERY_LDP_STP_N'.
5937 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
5938 (aarch64_print_address_internal): Add declaration.
5939 (aarch64_print_ldpstp_address): Remove.
5940 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
5941 (aarch64_print_operand): Change printing of 'y'.
5942 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
5943 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
5944 'true' rather than '1'.
5945 * config/aarch64/constraints.md (Uml): Likewise.
5946 (Uml): Rename to Umn.
5949 2018-07-19 Richard Biener <rguenther@suse.de>
5951 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
5953 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
5954 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
5955 (vn_reference_hasher): Likewise.
5956 (struct vn_tables_s): Remove obstack and alloc-pool members.
5957 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
5958 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
5959 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
5960 (vn_reference_insert_pieces): Likewise.
5961 (alloc_vn_nary_op_noinit): Adjust.
5962 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
5963 (vn_phi_eq): Adjust.
5964 (shared_lookup_phiargs): Remove.
5965 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
5966 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
5967 (visit_reference_op_call): Likewise.
5968 (copy_nary, copy_phi, copy_reference): Remove.
5969 (process_scc): Rewind the obstack when iterating. Do not
5970 copy the elements to valid_info but just move them from one
5971 hashtable to the other.
5972 (allocate_vn_table): Adjust.
5973 (free_vn_table): Likewise.
5974 (init_scc_vn): Likewise.
5975 (free_scc_vn): Likewise.
5977 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
5980 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
5981 indirect_return as function type attribute.
5982 (ix86_attribute_table): Change indirect_return to function
5984 * doc/extend.texi: Update indirect_return attribute.
5986 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
5988 * wide-int.h (widest2_int): New.
5989 * gimple-fold.c (arith_overflowed_p): Use it.
5990 * tree.h (widest2_int_cst): New.
5991 * tree-vrp.c (wide_int_binop_overflow): Rename from
5992 vrp_int_const_binop.
5993 Rewrite to work on trees.
5994 (extract_range_from_multiplicative_op_1): Abstract code to...
5995 (wide_int_range_min_max): ...here.
5996 (wide_int_range_cross_product): ...and here.
5997 (extract_range_from_binary_expr_1): Abstract overflow code to...
5998 (wide_int_range_mult_wrapping): ...here.
5999 * tree-vrp.h (wide_int_range_cross_product): New.
6000 (wide_int_range_mult_wrapping): New.
6002 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
6003 Julia Koval <julia.koval@intel.com>
6005 * config/i386/x86-tune-costs.h (skylake_memcpy,
6006 skylake_memset): Replace rep_prefix with unrolling for size 512.
6008 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
6011 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
6012 comparision with EQ_EXPR in last stmt.
6014 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
6016 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
6017 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
6018 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
6019 previously known as "PowerPC AltiVec Built-in Functions". Move
6020 some material to new subsubsections "PowerPC AltiVec Built-in
6021 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
6023 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
6024 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
6025 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
6026 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
6028 2018-07-18 Richard Biener <rguenther@suse.de>
6030 PR tree-optimization/86557
6031 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
6034 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
6036 * config/s390/s390.c (s390_function_profiler): Generate CFI.
6038 2018-07-17 Jeff Law <law@redhat.com>
6040 * config/arm/arm.c (get_label_padding): Update for recent
6041 changes to label_to_alignment.
6043 PR tree-optimization/86010
6044 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
6046 * config/mips/mips.c (vr4130_align_insns): Update for recent
6047 changes to label_to_alignment.
6049 * config/frv/frv.c (frv_label_align): Update for recent changes
6050 to label_to_alignment.
6052 * config/nios2/nios2.c (nios2_label_align): Update for recent
6053 changes which dropped ALIGN_LABELS_LOG.
6055 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
6057 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
6058 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
6060 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
6062 * config/arc/arc.c (arc_label_align): Use align_labels instead of
6063 deprecated align_labels_log.
6065 2018-07-17 Richard Biener <rguenther@suse.de>
6068 * dwarf2out.c (init_sections_and_labels): Always generate
6069 a debug_line_str_section for early LTO debug.
6070 (dwarf2out_finish): Reset debug_line_str_hash output early.
6071 Bump counter for extra dwarf5 .debug_loc labels to not conflict
6073 (dwarf2out_early_finish): Output debug_line_str.
6075 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
6077 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
6078 index register on z196 or later.
6080 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
6082 * config/s390/s390.c (s390_default_align): Set default function
6084 (s390_override_options_after_change): Call s390_default align.
6085 (s390_option_override_internal): Call s390_default_align.
6086 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
6088 2018-07-17 Jakub Jelinek <jakub@redhat.com>
6091 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
6092 to _looptemp_ clauses, other than the first two.
6094 2018-07-17 Martin Liska <mliska@suse.cz>
6096 * opts.c: Do not enable OPT_falign_* for -Os.
6098 2018-07-17 Martin Liska <mliska@suse.cz>
6100 * align.h (MAX_CODE_ALIGN): New.
6101 (MAX_CODE_ALIGN_VALUE): New.
6102 * common/config/i386/i386-common.c (ix86_handle_option):
6103 (MAX_CODE_ALIGN): Moved to align.h.
6104 * final.c (MAX_CODE_ALIGN): Likewise.
6105 * opts.c (parse_and_check_align_values):
6106 (MAX_CODE_ALIGN): Likewise.
6107 (MAX_CODE_ALIGN_VALUE): Likewise.
6109 2018-07-17 Martin Liska <mliska@suse.cz>
6111 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
6112 in order to fulfil coding style.
6113 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
6114 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6115 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
6116 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
6117 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
6118 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
6119 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
6120 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
6121 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6123 2018-07-17 Martin Liska <mliska@suse.cz>
6125 * align.h: New file.
6126 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
6128 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
6129 align_flags of label_to_alignment.
6130 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
6132 * config/m68k/m68k.c: Do not use removed align_labels_value and
6134 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
6135 (LOOP_ALIGN): Likewise.
6136 (LABEL_ALIGN): Likewise.
6137 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
6138 Remove not used macro.
6139 (rs6000_loop_align): Change return type to align_flags.
6140 (rs6000_loop_align_max_skip): Remove.
6141 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
6142 Change return type to align_flags.
6143 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
6144 Remove not used macro.
6145 (rs6000_loop_align): Change return type to align_flags.
6146 (rs6000_loop_align_max_skip): Remove.
6147 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
6148 * config/rx/rx-protos.h (rx_align_for_label): Make it
6150 * config/rx/rx.c (rx_align_for_label): Change return type
6152 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
6154 into align_flags class.
6155 (LABEL_ALIGN): Likewise.
6156 (LOOP_ALIGN): Likewise.
6157 * config/s390/s390.c (s390_label_align): Use align_flags
6159 (s390_asm_output_function_label): Likewise.
6160 * config/sh/sh.c (sh_override_options_after_change):
6161 Use align_flags class directly without macros.
6162 (find_barrier): Likewise.
6163 (barrier_align): Likewise.
6164 (sh_loop_align): Likewise.
6165 * config/spu/spu.c (spu_option_override):
6166 Use align_flags_tuple::get_value instead of removed macros.
6167 (spu_sched_init): Likewise.
6168 * config/spu/spu.h (GTY): Likewise.
6169 * config/visium/visium.c (visium_option_override):
6170 Set "8" as default secondary alignment.
6171 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
6172 in order to guarantee secondary alignment of 8.
6173 * coretypes.h: Include align.h header file.
6174 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
6175 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
6176 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
6177 * doc/tm.texi.in: Likewise.
6178 * final.c (struct label_alignment): Remove not used structure.
6179 (LABEL_ALIGN): Change type to align_flags.
6180 (LOOP_ALIGN): Likewise.
6181 (JUMP_ALIGN): Likewise.
6182 (default_loop_align_max_skip): Remove.
6183 (default_label_align_max_skip): Likewise.
6184 (default_jump_align_max_skip): Likewise.
6185 (default_label_align_after_barrier_max_skip):
6186 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
6187 (LABEL_TO_MAX_SKIP): Remove.
6188 (label_to_alignment): Return align_flags type instead of integer.
6189 (label_to_max_skip): Remove.
6190 (align_fuzz): Use align_flags type.
6191 (compute_alignments): Use align_flags type and use align_flags::max
6192 to combine multiple alignments.
6193 (grow_label_align): Grow vec instead of C array.
6194 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
6195 (shorten_branches): Use align_flags type and use align_flags::max
6196 to combine multiple alignments.
6197 (final_scan_insn_1): Remove usage of secondary alignment that comes
6198 from label alignment, but instead use proper secondary alignment
6199 which is computed in grow_label_align.
6200 * flags.h (struct align_flags_tuple): Move to align.h.
6201 (struct align_flags): Likewise.
6202 (state_align_loops): Rename to align_loops.
6203 (state_align_jumps): Rename to align_jumps.
6204 (state_align_labels): Rename to align_labels.
6205 (state_align_functions): Rename to align_functions.
6206 (align_loops_log): Remove.
6207 (align_jumps_log): Remove.
6208 (align_labels_log): Remove.
6209 (align_functions_log): Remove.
6210 (align_loops_max_skip): Remove.
6211 (align_jumps_max_skip): Remove.
6212 (align_labels_max_skip): Remove.
6213 (align_functions_max_skip): Remove.
6214 (align_loops_value): Remove.
6215 (align_jumps_value): Remove.
6216 (align_labels_value): Remove.
6217 (align_functions_value): Remove.
6218 * output.h (label_to_alignment): Change return type to align_flags.
6219 (label_to_max_skip): Remove.
6220 * target.def: Remove loop_align_max_skip, label_align_max_skip,
6221 jump_align_max_skip macros.
6222 * targhooks.h (default_loop_align_max_skip): Remove.
6223 (default_label_align_max_skip): Likewise.
6224 (default_jump_align_max_skip): Likewise.
6225 (default_label_align_after_barrier_max_skip): Remove.
6226 * toplev.c (read_log_maxskip): Use ::normalize function.
6227 (parse_N_M): Remove not used argument and also call ::normalize.
6228 (parse_alignment_opts): Do not pass unused arguments.
6229 * varasm.c (assemble_start_function): Use directly align_functions
6230 instead of removed macros.
6231 * system.h: Do not poison removed macros.
6233 2018-07-17 Jakub Jelinek <jakub@redhat.com>
6236 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
6237 and cond temporaries don't have reference type if iterator has
6238 pointer type. For init use &for_pre_body instead of pre_p if
6239 for_pre_body is non-empty.
6241 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
6243 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
6244 double-double modes to SFmode directly directly.
6245 (trunc<mode>sf2_fprs): Delete.
6247 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
6249 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
6250 for conversions between IFmode and the decimal floating point modes.
6251 (init_float128_ieee): Use the correct names for conversions between
6252 KFmode and the decimal floating point modes.
6254 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
6256 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
6257 for the conversions between TDmode and IFmode.
6258 (init_float128_ieee): Use more correct names for the conversions
6259 between TDmode and KFmode.
6261 2018-07-16 Jakub Jelinek <jakub@redhat.com>
6263 PR tree-optimization/86526
6264 * builtins.c (expand_builtin_memcmp): Formatting fixes.
6265 (inline_expand_builtin_string_cmp): Likewise.
6266 (inline_string_cmp): Likewise. Use c_readstr instead of
6267 builtin_memcpy_read_str. Add unit_mode temporary.
6269 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
6272 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
6273 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
6275 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6277 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
6278 Alphabetize prototypes of built-in functions, separating out
6279 built-in functions that are listed in this section but should be
6280 described elsewhere.
6282 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
6285 * expmed.c (emit_store_flag): Do not emit setcc followed by a
6286 conditional move when trapping comparison was split to a
6287 non-trapping one (and vice versa).
6289 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
6291 * config/s390/s390.c (s390_function_profiler): Generate nops
6292 instead of profiler call sequences.
6293 * config/s390/s390.opt: Add the new option.
6295 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
6297 * config/s390/s390.c (s390_function_profiler): Generate
6298 __mcount_loc section.
6299 * config/s390/s390.opt: Add the new option.
6301 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
6303 * common.opt: Add the new warning.
6304 * config/s390/s390.c (s390_function_profiler): Emit "brasl
6305 %r0,__fentry__" when -mfentry is specified.
6306 (s390_option_override_internal): Disallow -mfentry for 31-bit
6308 * config/s390/s390.opt: Add the new option.
6310 2018-07-16 Richard Biener <rguenther@suse.de>
6313 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
6314 for function-local FUNCTION_DECL and RESULT_DECL immediately.
6316 2018-07-16 Martin Liska <mliska@suse.cz>
6319 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
6322 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
6324 * config/arc/arcHS.md: Update ARCHS scheduling rules.
6326 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
6328 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
6330 * config/arc/arc-cpus.def (hs4x): New cpu.
6332 * config/arc/arc-tables.opt: Regenerate.
6333 * config/arc/arc.c (arc_sched_issue_rate): New function.
6334 (TARGET_SCHED_ISSUE_RATE): Define.
6335 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
6336 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
6338 (attr tune): Add ARCHS4x tune values.
6339 (attr tune_dspmpy): Define.
6340 (*tst): Correct instruction type.
6341 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
6342 * config/arc/arcHS4x.md: New file.
6343 * config/arc/fpu.md: Update instruction type attributes.
6344 * config/arc/t-multilib: Regenerate.
6346 2018-07-16 Tom de Vries <tdevries@suse.de>
6349 * var-tracking.c (vt_initialize): Fix pre_dec handling.
6351 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6353 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
6356 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
6358 PR tree-optimization/86514
6359 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
6360 conversion to a boolean type from a type with greater precision.
6362 2018-07-16 Tom de Vries <tdevries@suse.de>
6364 * var-tracking.c (vt_initialize): Print adjusted insn slim if
6365 dump_flags request TDF_SLIM.
6367 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
6369 * fold-const.c (int_const_binop_1): Abstract...
6370 (wide_int_binop): ...wide int code here.
6371 (poly_int_binop): ...poly int code here.
6372 Abstract the rest of int_const_binop_1 into int_const_binop.
6373 * fold-const.h (wide_int_binop): New.
6374 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
6375 Remove useless PLUS/MINUS_EXPR case.
6376 (zero_nonzero_bits_from_vr): Move wide int code...
6377 (zero_nonzero_bits_from_bounds): ...here.
6378 (extract_range_from_binary_expr_1): Move mask optimization code...
6379 (range_easy_mask_min_max): ...here.
6380 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
6381 (range_easy_mask_min_max): New.
6383 2018-07-15 Jeff Law <law@redhat.com>
6386 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
6389 2018-07-14 Jim Wilson <jimw@sifive.com>
6391 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
6393 2018-07-14 Paul Koning <ni1d@arrl.net>
6395 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
6397 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
6399 * lto-streamer-out.c (copy_function_or_variable): Dump info about
6402 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
6403 Steve Munroe <munroesj52@gmail.com>
6405 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
6406 (_mm_andnot_si128): Likewise.
6407 (_mm_or_si128): Likewise.
6408 (_mm_xor_si128): Likewise.
6410 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
6413 * builtins.c (expand_builtin_memcmp): Inline the calls first
6414 when result_eq is false.
6415 (expand_builtin_strcmp): Inline the calls first.
6416 (expand_builtin_strncmp): Likewise.
6417 (inline_string_cmp): New routine. Expand a string compare
6418 call by using a sequence of char comparison.
6419 (inline_expand_builtin_string_cmp): New routine. Inline expansion
6420 a call to str(n)cmp/memcmp.
6421 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
6423 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
6425 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
6427 * config/arm/driver-arm.c: Include arm-native.h.
6428 (host_detect_local_cpu): Use auto-generated data tables.
6429 (vendors, arm_cpu_table): Delete. Move part information to ...
6430 * config/arm/arm-cpus.in: ... here.
6431 * config/arm/parsecpu.awk (gen_native): New function.
6432 (vendor, part): New CPU fields.
6433 (END): Add support for building the native CPU detection tables.
6434 * config/arm/t-arm (arm-native.h): Add build rule.
6435 (driver-arm.o): Add dependency on arm-native.h.
6437 2018-07-13 Richard Biener <rguenther@suse.de>
6440 * match.pd (addr1 - addr2): Allow either of the operand to
6443 2018-07-13 Tom de Vries <tdevries@suse.de>
6445 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
6448 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
6450 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
6451 arrays instead of numbered variables.
6453 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
6455 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
6456 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
6458 2018-07-13 Richard Biener <rguenther@suse.de>
6461 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
6462 instead of get_context_die.
6464 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
6465 Richard Biener <rguenther@suse.de>
6468 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
6469 that the loop latch destination where phi is defined.
6471 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
6473 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
6474 (riscv_expand_epilogue): Add assertion to check interrupt mode.
6475 (riscv_set_current_function): Extract getting interrupt type to new
6477 (riscv_get_interrupt_type): New function.
6478 (riscv_merge_decl_attributes): New function, checking interrupt type is
6480 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
6482 2018-07-12 Paul Koning <ni1d@arrl.net>
6484 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
6487 2018-07-12 Paul Koning <ni1d@arrl.net>
6489 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
6490 zero reference, add doloop_end instead.
6491 * doc/md.texi (decrement_and_branch_until_zero): Remove.
6492 (Looping patterns): Remove decrement_and_branch_until_zero. Add
6493 detail for doloop_end.
6495 2018-07-12 Martin Sebor <msebor@redhat.com>
6498 * attribs.c (decl_attributes): Reject conflicting attributes before
6499 calling attribute handlers.
6501 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
6503 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
6505 (gcc::dump_manager::get_dump_file_name): likewise.
6506 (dump_begin): Likewise.
6507 * dumpfile.h (dump_begin): Update prototype.
6508 (gcc::dump_manager::get_dump_file_name,
6509 gcc::dump_manager::get_dump_file_name): Update prototype.
6511 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6513 * internal-fn.h (vectorizable_internal_fn_p): New function.
6514 * tree-vect-slp.c (compatible_calls_p): Likewise.
6515 (vect_build_slp_tree_1): Remove nops argument. Handle calls
6516 to internal functions.
6517 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
6519 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6521 * fold-const.h (inverse_conditions_p): Declare.
6522 * fold-const.c (inverse_conditions_p): New function.
6523 * match.pd: Use inverse_conditions_p. Add folds of view_converts
6524 that test the inverse condition of a conditional internal function.
6525 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
6526 * internal-fn.c (internal_fn_mask_index): Handle conditional
6528 (vectorized_internal_fn_supported_p): New function.
6529 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
6530 (any_pred_load_store): Replace with...
6531 (need_to_predicate): ...this new variable.
6532 (redundant_ssa_names): New variable.
6533 (ifcvt_can_use_mask_load_store): Move initial checks to...
6534 (ifcvt_can_predicate): ...this new function. Handle tree codes
6535 for which a conditional internal function exists.
6536 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
6537 instead of ifcvt_can_use_mask_load_store. Update after variable
6539 (predicate_load_or_store): New function, split out from
6540 predicate_mem_writes.
6541 (check_redundant_cond_expr): New function.
6542 (value_available_p): Likewise.
6543 (predicate_rhs_code): Likewise.
6544 (predicate_mem_writes): Rename to...
6545 (predicate_statements): ...this. Use predicate_load_or_store
6546 and predicate_rhs_code.
6547 (combine_blocks, tree_if_conversion): Update after above name changes.
6548 (ifcvt_local_dce): Handle redundant_ssa_names.
6549 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
6550 general conditional functions.
6551 * tree-vect-stmts.c (vectorizable_call): Likewise.
6553 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6554 Alan Hayward <alan.hayward@arm.com>
6555 David Sherwood <david.sherwood@arm.com>
6557 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
6558 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
6559 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
6560 plus and minus and convert them into IFN_COND_FMA-based sequences.
6561 (convert_mult_to_fma): Handle conditional plus and minus.
6563 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6565 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
6566 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
6567 (cond_fnms_optab): New optabs.
6568 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
6570 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
6571 * internal-fn.h (get_conditional_internal_fn): Declare.
6572 (get_unconditional_internal_fn): Likewise.
6573 * internal-fn.c (cond_ternary_direct): New macro.
6574 (expand_cond_ternary_optab_fn): Likewise.
6575 (direct_cond_ternary_optab_supported_p): Likewise.
6576 (FOR_EACH_COND_FN_PAIR): Likewise.
6577 (get_conditional_internal_fn): New function.
6578 (get_unconditional_internal_fn): Likewise.
6579 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
6580 (gimple_match_op::gimple_match_op): Add a new overload for 5
6582 (gimple_match_op::set_op): Likewise.
6583 (gimple_resimplify5): Declare.
6584 * genmatch.c (decision_tree::gen): Generate simplifications for
6586 * gimple-match-head.c (gimple_simplify): Define an overload for
6587 5 operands. Handle calls with 5 arguments in the top-level overload.
6588 (convert_conditional_op): Handle conversions from unconditional
6589 internal functions to conditional ones.
6590 (gimple_resimplify5): New function.
6591 (build_call_internal): Pass a fifth operand.
6592 (maybe_push_res_to_seq): Likewise.
6593 (try_conditional_simplification): Try converting conditional
6594 internal functions to unconditional internal functions.
6595 Handle 3-operand unconditional forms.
6596 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
6597 Define ternary equivalents of the current rules for binary conditional
6599 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
6601 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
6602 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
6603 (optab): Handle them.
6604 (SVE_COND_FP_TERNARY): New int iterator.
6605 (sve_fmla_op, sve_fmad_op): New int attributes.
6606 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
6607 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
6608 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
6610 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6612 * target.def (preferred_else_value): New target hook.
6613 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
6614 * doc/tm.texi: Regenerate.
6615 * targhooks.h (default_preferred_else_value): Declare.
6616 * targhooks.c (default_preferred_else_value): New function.
6617 * internal-fn.h (conditional_internal_fn_code): Declare.
6618 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
6619 (get_conditional_internal_fn): Use it.
6620 (conditional_internal_fn_code): New function.
6621 * gimple-match.h (gimple_match_cond): New struct.
6622 (gimple_match_op): Add a cond member function.
6623 (gimple_match_op::gimple_match_op): Update all forms to take a
6625 * genmatch.c (expr::gen_transform): Use the same condition as res_op
6626 for the suboperation, but don't specify a particular else_value.
6627 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
6628 (visit_nary_op, visit_reference_op_load): Pass
6629 gimple_match_cond::UNCOND to the gimple_match_op constructor.
6630 * gimple-match-head.c: Include tree-eh.h
6631 (convert_conditional_op): New function.
6632 (maybe_resimplify_conditional_op): Likewise.
6633 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
6634 (gimple_resimplify2): Likewise.
6635 (gimple_resimplify3): Likewise.
6636 (gimple_resimplify4): Likewise.
6637 (maybe_push_res_to_seq): Return null for conditional operations.
6638 (try_conditional_simplification): New function.
6639 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
6641 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
6643 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
6645 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
6647 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
6649 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
6651 (hash_tree): Do not hash DECL_FCONTEXT
6652 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
6653 Do not stream DECL_FCONTEXT.
6654 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
6655 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
6657 2018-07-12 Richard Biener <rguenther@suse.de>
6660 * dwarf2out.c (gen_block_die): Only output blocks when they have
6661 at least one !DECL_IGNORED_P variable.
6663 2018-07-12 Richard Biener <rguenther@suse.de>
6666 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
6667 Remove -mieee-fp handling.
6669 2018-07-12 Richard Biener <rguenther@suse.de>
6671 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
6672 left-over from last patch.
6674 2018-07-12 Jakub Jelinek <jakub@redhat.com>
6676 PR tree-optimization/86492
6677 * gimple-ssa-store-merging.c
6678 (imm_store_chain_info::coalesce_immediate_stores): Call
6679 check_no_overlap even for the merge_overlapping case. Formatting fix.
6681 2018-07-12 Richard Biener <rguenther@suse.de>
6684 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
6685 move possibly trapping operations into the conditional.
6687 2018-07-12 Richard Biener <rguenther@suse.de>
6689 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
6690 (vn_lookup_simplify_result): Remove recursion limit applied
6692 (vn_nary_build_or_lookup_1): Adjust.
6693 (try_to_simplify): Likewise.
6694 * gimple-match-head.c (gimple_resimplify1): Instead apply one
6696 (gimple_resimplify2): Likewise.
6697 (gimple_resimplify3): Likewise.
6698 (gimple_resimplify4): Likewise.
6700 2018-07-11 Jakub Jelinek <jakub@redhat.com>
6702 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
6703 Use __mmask64 type instead of __mmask8 for __M argument.
6704 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
6705 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
6707 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
6708 __mmask16 for __M argument.
6709 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
6710 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
6711 to __mmask16 instead of __mmask8.
6712 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
6713 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
6714 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
6715 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
6716 instead of __mmask16 for __U argument.
6717 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
6718 __mmask16 instead of __mmask8 for __U argument.
6719 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
6721 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
6723 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
6725 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
6727 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
6729 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
6731 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
6733 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
6735 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
6737 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
6739 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
6740 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
6741 return type as well as __M argument type and all casts from __mmask8
6743 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
6744 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
6745 return type as well as __M argument type and all casts from __mmask8
6747 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
6748 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
6749 return type as well as __M argument type and all casts from __mmask8
6751 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
6752 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
6753 return type as well as __M argument type and all casts from __mmask8
6755 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
6756 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
6759 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
6761 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
6762 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
6765 2018-07-11 Paul Koning <ni1d@arrl.net>
6767 * doc/md.texi (define_subst): Document how multiple occurrences of
6768 the same argument in the replacement pattern are handled.
6770 2018-07-11 Paul Koning <ni1d@arrl.net>
6772 * doc/extend.texi (Common Variable Attributes): Move "mode" into
6774 (Common Type Attributes): Add "mode" attribute.
6776 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
6778 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
6779 stream DECL_ORIGINAL_TYPE.
6780 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
6781 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
6782 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
6783 Do not walk original type.
6784 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
6786 (write_ts_decl_non_common_tree_pointers): Do not stream
6788 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
6789 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
6791 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
6793 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
6794 thread twice from the same starting edge.
6796 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
6798 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
6799 * gimple.c (gimple_call_nonnull_result_p): ...here...
6800 (gimple_call_nonnull_arg): ...and here.
6801 * gimple.h (gimple_call_nonnull_result_p): New.
6802 (gimple_call_nonnull_arg): New.
6804 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
6806 * config/arm/arm-cpus.in: Move information from fpu field of each
6807 cpu definition to the isa field.
6808 * config/arm/parsecpu.awk (fpu): Delete match rule.
6809 (gen_comm_data): Don't add bits from the CPU's FPU entry.
6811 2018-07-11 Richard Biener <rguenther@suse.de>
6814 * dwarf2out.c (init_sections_and_labels): Use
6815 output_asm_line_debug_info consistently.
6816 (dwarf2out_early_finish): Likewise.
6817 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
6820 2018-07-11 Richard Biener <rguenther@suse.de>
6822 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
6823 Rework father_bb setting in a way to avoid propagating constants
6824 multiple times on a loop body.
6826 2018-07-10 Mark Wielaard <mark@klomp.org>
6829 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
6831 2018-07-10 Richard Biener <rguenther@suse.de>
6833 * hash-map.h (hash_map::iterator::operator*): Return
6834 references to key and value.
6836 2018-07-10 Jakub Jelinek <jakub@redhat.com>
6839 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
6840 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
6841 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
6842 (gimplify_omp_for): For composite loops, move outer
6843 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
6844 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
6845 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
6846 TREE_LIST for both the original class iterator and the "last" helper
6847 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
6848 loop, remember has_decl_expr from outer composite loops for the
6849 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
6851 2018-07-09 Martin Sebor <msebor@redhat.com>
6855 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
6856 accessing implicitly initialized array elements.
6857 * expr.c (string_constant): Handle string initializers of
6858 character arrays within aggregates.
6859 * gimple-fold.c (fold_array_ctor_reference): Add argument.
6860 Store element offset. As a special case, handle zero size.
6861 (fold_nonarray_ctor_reference): Same.
6862 (fold_ctor_reference): Add argument. Store subobject offset.
6863 * gimple-fold.h (fold_ctor_reference): Add argument.
6865 2018-07-09 Paul Koning <ni1d@arrl.net>
6867 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
6868 (pdp11_insn_cost): New function.
6869 (pdp11_md_asm_adjust): New function.
6870 (TARGET_INVALID_WITHIN_DOLOOP): Define.
6871 (pdp11_rtx_costs): Update to match machine better.
6872 (output_addr_const_pdp11): Correct format mismatch warnings.
6873 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
6874 * config/pdp11/pdp11.md: General change to add base_cost and/or
6875 length attributes for use by new pdp11_insn_cost function.
6876 (MIN_BRANCH): Correct definition.
6878 (doloop_end): Use standard pattern name for looping pattern.
6879 (doloop_end_nocc): New.
6880 (movsf): Add another constraint alternative.
6881 (zero_extendqihi2): Add constraint alternatives for not in place
6883 (zero_extendhisi2): Remove.
6884 (shift patterns): Add CC handling variants.
6888 (define_peephole2): New peephole to recognize mov that sets CC for
6891 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6893 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
6894 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
6897 2018-07-09 Jakub Jelinek <jakub@redhat.com>
6900 * real.c (real_nextafter): Return true if result is denormal.
6902 2018-07-09 Martin Liska <mliska@suse.cz>
6904 * common.opt: Add back wrongly removed attribute.
6906 2018-07-09 Richard Biener <rguenther@suse.de>
6909 * dwarf2out.c (gen_block_die): For an early generated DIE
6910 always output high/low PC attributes.
6912 2018-07-09 Tom de Vries <tdevries@suse.de>
6914 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
6915 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
6916 onto VAR_DECL with abstract origin.
6918 2018-07-07 Jim Wilson <jimw@sifive.com>
6920 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
6922 2018-07-07 Tom de Vries <tdevries@suse.de>
6924 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
6926 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
6928 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
6930 (combine_bound): Use wide-int overflow calculation instead of
6932 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
6934 * fold-const.c (int_const_binop_2): Same.
6935 (extract_muldiv_1): Same.
6936 (fold_div_compare): Same.
6937 (fold_abs_const): Same.
6939 * poly-int.h (add): Same.
6943 * predict.c (predict_iv_comparison): Same.
6944 * profile-count.c (slow_safe_scale_64bit): Same.
6945 * simplify-rtx.c (simplify_const_binary_operation): Same.
6946 * tree-chrec.c (tree_fold_binomial): Same.
6947 * tree-data-ref.c (split_constant_offset_1): Same.
6948 * tree-if-conv.c (idx_within_array_bound): Same.
6949 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
6950 * tree-ssa-phiopt.c (minmax_replacement): Same.
6951 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
6952 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
6953 * vr-values.c (vr_values::adjust_range_with_scev): Same.
6954 * wide-int.cc (wi::add_large): Same.
6955 (wi::mul_internal): Same.
6956 (wi::sub_large): Same.
6957 (wi::divmod_internal): Same.
6958 * wide-int.h: Change overflow type to overflow_type for neg, add,
6959 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
6960 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
6961 mul_internal, divmod_internal.
6962 (overflow_type): New enum.
6963 (accumulate_overflow): New.
6965 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
6967 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
6968 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
6970 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
6972 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
6973 argument is checked for zero before entering loop, avoid checking again.
6975 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
6977 * gimplify.h (generic_expr_could_trap_p): Set as global function.
6978 * gimplify.h (generic_expr_could_trap_p): Likwise.
6979 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
6981 2018-07-06 Jakub Jelinek <jakub@redhat.com>
6983 PR tree-optimization/86401
6984 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
6985 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
6986 (fold_bit_and_mask): ... here. New helper function for match.pd.
6987 * fold-const.h (fold_bit_and_mask): Declare.
6988 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
6990 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
6993 * target.def (translate_mode_attribute): New hook.
6994 * targhooks.h (default_translate_mode_attribute): Declare.
6995 * targhooks.c (default_translate_mode_attribute): New function.
6996 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
6997 * doc/tm.texi: Regenerate.
6998 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
6999 (rs6000_translate_mode_attribute): New function.
7001 2018-07-06 Paul Koning <ni1d@arrl.net>
7003 * doc/md.texi (define_split): Document DONE and FAIL.
7004 (define_peephole2): Ditto.
7006 2018-07-05 Jeff Law <law@redhat.com>
7008 PR tree-optimization/86010
7009 * tree-ssa-dse.c (compute_trims): More aggressively trim at
7010 both the head and tail of mem* and str* calls.
7012 2018-07-05 Jim Wilson <jimw@sifive.com>
7014 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
7016 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
7018 * config/aarch64/aarch64-simd.md: correct flags text for
7019 MIN_EXPR replacement.
7021 2018-07-05 James Clarke <jrtc27@jrtc27.com>
7023 * configure: Regenerated.
7025 2018-07-05 Carl Love <cel@us.ibm.com>
7027 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
7028 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
7029 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
7030 VSX_BUILTIN_DOUBLEL_V4SF.
7032 2018-07-05 Martin Sebor <msebor@redhat.com>
7035 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
7036 than its domain to compute its the upper bound of a char array.
7038 2018-07-05 Nathan Sidwell <nathan@acm.org>
7040 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
7041 * doc/cpp.texi: Update comment.
7042 * doc/tm.texi: Rebuilt.
7043 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
7044 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
7045 * doc/extend.texi (Backwards Compatibility): Clarify it is system
7046 headers affected by extern "C".
7047 * system.h: Poison NO_IMPLICIT_EXTERN_C.
7048 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
7049 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
7050 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
7051 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
7052 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
7053 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
7054 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
7055 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
7056 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
7057 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
7058 config/sparc/sp64-elf.h, config/spu/spu.h,
7059 config/stormy16/stormy16.h, config/v850/v850.h,
7060 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
7061 define NO_IMPLICIT_EXTERN_C.
7062 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
7064 2018-07-05 Tamar Christina <tamar.christina@arm.com>
7067 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
7068 instead of GET_MODE_SIZE when comparing Units.
7070 2018-07-05 Tamar Christina <tamar.christina@arm.com>
7073 * rtlanal.c (set_noop_p): Constrain on mode change,
7074 include hard-reg-set.h
7076 2018-07-05 Tamar Christina <tamar.christina@arm.com>
7078 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
7080 2018-07-05 Jakub Jelinek <jakub@redhat.com>
7083 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
7086 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
7088 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
7090 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
7093 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
7095 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
7097 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
7100 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
7102 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
7104 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
7105 neg_*_op* variables.
7107 2018-07-04 Martin Liska <mliska@suse.cz>
7109 * tree-switch-conversion.c: Define
7110 max_ratio_for_speed and max_ratio_for_size constants.
7112 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
7113 Martin Liska <mliska@suse.cz>
7118 * common.opt: Rename align options with 'str_' prefix.
7119 * common/config/i386/i386-common.c (set_malign_value): New
7121 (ix86_handle_option): Use it to set -falign-* options/
7122 * config/aarch64/aarch64-protos.h (struct tune_params): Change
7123 type from int to string.
7124 * config/aarch64/aarch64.c: Update default values from int
7126 * config/alpha/alpha.c (alpha_override_options_after_change):
7128 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
7129 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7130 max skip conditionally.
7131 * config/i386/freebsd.h (SUBALIGN_LOG): New.
7132 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7133 max skip conditionally.
7134 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7135 max skip conditionally.
7136 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
7137 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7138 max skip conditionally.
7139 * config/i386/i386.c (struct ptt): Change type from int to
7141 (ix86_default_align): Set default values.
7142 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
7143 max skip conditionally.
7144 * config/i386/iamcu.h (SUBALIGN_LOG): New.
7145 (ASM_OUTPUT_MAX_SKIP_ALIGN):
7146 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
7147 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7148 max skip conditionally.
7149 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
7150 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
7151 * config/i386/x86-64.h (SUBALIGN_LOG): New.
7152 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7153 max skip conditionally.
7154 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
7155 * config/ia64/ia64.c (ia64_option_override): Set default values
7156 for alignment options.
7157 * config/m68k/m68k.c: Handle new str_align_* options.
7158 * config/mips/mips.c (mips_set_compression_mode): Change
7160 (mips_option_override): Set default values for options.
7161 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
7163 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7165 * config/rx/rx.c (rx_option_override): Likewise.
7166 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
7167 (LABEL_ALIGN): Use align_labels_log.
7168 (LOOP_ALIGN): Use align_loops_align.
7169 * config/s390/s390.c (s390_asm_output_function_label): Use new
7171 * config/sh/sh.c (sh_override_options_after_change):
7172 Change type of constants.
7173 * config/spu/spu.c (spu_sched_init): Likewise.
7174 * config/sparc/sparc.c (sparc_option_override): Set default
7176 * config/visium/visium.c (visium_option_override): Likewise.
7177 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
7178 emit p2align format with last argument if it's not needed.
7179 * doc/invoke.texi: Document extended format of -falign-*.
7180 * final.c: Use align_labels alignment.
7181 * flags.h (struct target_flag_state): Change type to use
7183 (struct align_flags_tuple): New.
7184 (struct align_flags): Likewise.
7185 (align_loops_log): Redefine macro to use new types.
7186 (align_loops_max_skip): Redefine macro to use new types.
7187 (align_jumps_log): Redefine macro to use new types.
7188 (align_jumps_max_skip): Redefine macro to use new types.
7189 (align_labels_log): Redefine macro to use new types.
7190 (align_labels_max_skip): Redefine macro to use new types.
7191 (align_functions_log): Redefine macro to use new types.
7192 (align_loops): Redefine macro to use new types.
7193 (align_jumps): Redefine macro to use new types.
7194 (align_labels): Redefine macro to use new types.
7195 (align_functions): Redefine macro to use new types.
7196 (align_functions_max_skip): Redefine macro to use new types.
7197 (align_loops_value): New macro.
7198 (align_jumps_value): New macro.
7199 (align_labels_value): New macro.
7200 (align_functions_value): New macro.
7201 * function.c (invoke_set_current_function_hook): Propagate
7202 alignment values from flags to global variables default in
7204 * ipa-icf.c (sem_function::equals_wpa): Use
7205 cl_optimization_option_eq instead of memcmp.
7206 * lto-streamer.h (cl_optimization_stream_out): Support streaming
7208 (cl_optimization_stream_in): Likewise.
7209 * optc-save-gen.awk: Support strings in cl_optimization.
7210 * opth-gen.awk: Likewise.
7211 * opts.c (finish_options): Remove error checking of invalid
7213 (MAX_CODE_ALIGN): Remove.
7214 (MAX_CODE_ALIGN_VALUE): Likewise.
7215 (parse_and_check_align_values): New function.
7216 (check_alignment_argument): Likewise.
7217 (common_handle_option): Use check_alignment_argument.
7218 * opts.h (parse_and_check_align_values): Declare.
7219 * toplev.c (init_alignments): Remove.
7220 (read_log_maxskip): New.
7221 (parse_N_M): Likewise.
7222 (parse_alignment_opts): Likewise.
7223 (backend_init_target): Remove usage of init_alignments.
7224 * toplev.h (parse_alignment_opts): Declare.
7225 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
7227 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
7228 * tree.c (cl_option_hasher::equal): New.
7229 * varasm.c: Use new global macros.
7231 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
7233 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
7234 Use a simpler align directive also if MAXSKIP = ALIGN-1.
7235 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7236 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7237 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7238 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
7239 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
7240 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
7241 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
7242 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7243 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7244 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7245 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7247 2018-07-04 Martin Liska <mliska@suse.cz>
7248 Jonathan Wakely <jwakely@redhat.com>
7250 * coverage.c: Use correct type.
7251 * doc/invoke.texi: Language correction.
7253 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
7256 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
7257 ENDBRANCH for non-tail call which may return via indirect branch.
7258 * doc/extend.texi: Document indirect_return attribute.
7260 2018-07-03 Martin Sebor <msebor@redhat.com>
7262 PR tree-optimization/86274
7263 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
7265 (format_floating): Correct handling of infinities and NaNs.
7267 2018-07-03 Martin Sebor <msebor@redhat.com>
7269 * print-tree.c (print_real_cst): New function.
7270 (print_node_brief): Call it.
7271 (print_node): Ditto.
7273 2018-07-03 Jeff Law <law@redhat.com>
7275 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
7276 into a single pattern.
7278 * config/h8300/h8300.md (ors code_iterator): New.
7279 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
7280 a single pattern and single splitter.
7281 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
7282 (iorqi3_1, xorqi3_1): Likewise.
7283 (iorqi3, xorqi3 expanders): Similarly.
7285 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
7286 (movmd_internal) into a single pattern using the P mode iterator.
7287 (movmd splitters): Similarly.
7288 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
7289 (movsd splitters): Similarly.
7291 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
7292 ADDB, ADDW and ADDL into a single ADD attribute which selects the
7293 right table based on the size of the operand.
7294 * config/h8300/h8300.md (length_table): Corresponding changes. All
7295 references to "addb", "addw" and "addl" changed to "add".
7296 (btst patterns): Merge two variants into a single pattern.
7297 (tstqi, tsthi): Likewise.
7298 (addhi3_incdec, addsi3_incdec): Likewise.
7299 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
7300 (mulhi3, mulsi3): Likewise.
7301 (udivhi3, udivsi3): Likewise.
7302 (divhi3, divsi3): Likewise.
7303 (andorqi3, andorhi3, andorsi3): Likewise.
7305 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
7308 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
7309 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
7311 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7313 PR tree-optimization/85694
7314 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
7315 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
7317 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
7318 (<u>avg<mode>3_ceil): New patterns.
7320 2018-07-03 David Malcolm <dmalcolm@redhat.com>
7322 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
7323 scan-tree-dump directive.
7324 * gcc.dg/vect/slp-perm-2.c: Likewise.
7325 * gcc.dg/vect/slp-perm-3.c: Likewise.
7326 * gcc.dg/vect/slp-perm-5.c: Likewise.
7327 * gcc.dg/vect/slp-perm-6.c: Likewise.
7328 * gcc.dg/vect/slp-perm-7.c: Likewise.
7329 * gcc.dg/vect/slp-perm-8.c: Likewise.
7331 2018-07-03 Marek Polacek <polacek@redhat.com>
7334 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
7336 2018-07-03 Richard Biener <rguenther@suse.de>
7339 * tree-ssa-structalias.c (find_func_clobbers): Properly
7340 handle indirect calls.
7342 2018-07-03 Jeff Law <law@redhat.com>
7344 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
7345 (shifts): New code iterator.
7346 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
7347 expander. Fix HImode handling on H8/SX.
7348 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
7349 (subqi3, subhi3, subsi3 expanders): Likewise.
7350 (andqi3, andhi3, andsi3 expanders): Likewise.
7351 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
7352 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
7353 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
7354 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
7355 (zero_extendqihi2, zero_extendqisi2): Likewise.
7356 (extendqihi2, extendqisi2): Likewise.
7357 (rotlqi3, rotlhi3, rotlsi3): Likewise.
7358 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
7359 (rotlqi3_1, rotlhi3_1): Likewise.
7360 (logicalhi3_sn, logicalsi3_sn): Likewise.
7361 (logicalhi3, logicalsi3): Likewise.
7363 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7365 * tree-vect-patterns.c (vect_recog_rotate_pattern)
7366 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
7367 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
7368 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
7369 type to append_pattern_def_seq instead of creating a stmt_vec_info
7371 (build_mask_conversion): Likewise. Remove vinfo argument.
7372 (vect_add_conversion_to_patterm): Likewise, renaming to...
7373 (vect_add_conversion_to_pattern): ...this.
7374 (vect_recog_mask_conversion_pattern): Update call to
7375 build_mask_conversion. Pass the vector type to
7376 append_pattern_def_seq here too.
7377 (vect_recog_gather_scatter_pattern): Update call to
7378 vect_add_conversion_to_pattern.
7380 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7382 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
7383 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7384 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
7385 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
7386 STMT_VINFO_PATTERN_DEF_SEQ to null here.
7387 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
7388 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
7389 append_pattern_def_seq instead of new_pattern_def_seq.
7390 (vect_recog_divmod_pattern): Do both of the above.
7391 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
7394 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7396 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
7397 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
7398 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
7399 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
7400 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
7401 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
7402 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
7403 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
7404 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
7405 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
7406 parameter with a single stmt_vec_info.
7407 (vect_recog_func_ptr): Likewise.
7408 (vect_recog_gather_scatter_pattern): Likewise, folding in...
7409 (vect_try_gather_scatter_pattern): ...this.
7410 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
7411 the stmt_vec_info of the statement to be matched. Don't clear
7412 STMT_VINFO_RELATED_STMT.
7413 (vect_pattern_recog): Update call accordingly.
7415 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7417 PR tree-optimization/85694
7418 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
7419 (uavgM3_ceil): Document new optabs.
7420 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
7421 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
7423 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
7424 (savg_ceil_optab): New optabs.
7425 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
7426 (vect_vect_recog_func_ptrs): Add it.
7427 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
7428 constant directly from the associated lhs.
7430 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7432 * tree-vect-patterns.c (vect_split_statement): New function.
7433 (vect_convert_input): Use it to try to split an existing cast.
7435 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7437 * poly-int.h (print_hex): New function.
7438 * dumpfile.h (dump_dec, dump_hex): Declare.
7439 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
7440 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
7441 min_input_precision, operation_precision and operation_sign.
7442 * tree-vect-patterns.c (vect_get_range_info): New function.
7443 (vect_same_loop_or_bb_p, vect_single_imm_use)
7444 (vect_operation_fits_smaller_type): Delete.
7445 (vect_look_through_possible_promotion): Add an optional
7446 single_use_p parameter.
7447 (vect_recog_over_widening_pattern): Rewrite to use new
7448 stmt_vec_info infomration. Handle one operation at a time.
7449 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
7450 (vect_truncatable_operation_p, vect_set_operation_type)
7451 (vect_set_min_input_precision): New functions.
7452 (vect_determine_min_output_precision_1): Likewise.
7453 (vect_determine_min_output_precision): Likewise.
7454 (vect_determine_precisions_from_range): Likewise.
7455 (vect_determine_precisions_from_users): Likewise.
7456 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
7457 (vect_vect_recog_func_ptrs): Put over_widening first.
7459 (vect_pattern_recog): Call vect_determine_precisions.
7461 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7463 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
7464 statements that have been replaced by further pattern statements.
7465 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
7467 2018-07-03 Richard Biener <rguenther@suse.de>
7469 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
7470 always set *dt. Dump vectype in vectype overload.
7471 * dumpfile.h (dump_gimple_expr): New function.
7472 (dump_gimple_expr_loc): Likewise.
7473 * dumpfile.c (dump_gimple_expr): New function.
7474 (dump_gimple_expr_loc): Likewise.
7476 2018-07-02 Jeff Law <law@redhat.com>
7478 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
7479 the H8/300, H8/300H and H8/S variants into a single pattern.
7480 (movhi_h8300, movqi_h8300hs): Similarly.
7481 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
7482 (QHI mode iterator): New.
7484 * config/h8300/h8300.md: Remove trailing whitespace.
7486 2018-07-02 Jim Wilson <jimw@sifive.com>
7488 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
7489 instead of emit_insn for interrupt returns.
7490 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
7491 (riscv_sret, riscv_uret): Likewise.
7493 2018-07-02 David Malcolm <dmalcolm@redhat.com>
7495 * pretty-print.c (selftest::test_pp_format): Move save and restore
7496 of quotes to class auto_fix_quotes, and add an instance.
7497 * selftest.c: Include "intl.h".
7498 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
7499 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
7500 * selftest.h (selftest::auto_fix_quotes): New class.
7502 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
7504 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
7505 (aarch64_sve_prepare_conditional_op): Remove.
7506 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
7507 Allow aarch64_simd_reg_or_zero as select operand; remove
7508 the aarch64_sve_prepare_conditional_op call.
7509 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
7510 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
7511 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
7512 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
7513 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
7514 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
7515 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
7516 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
7517 and a splitters to match all of the *_any patterns.
7518 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
7520 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
7521 (SVE_COND_FP_BINARY_REV): Remove.
7522 (sve_int_op_rev, sve_fp_op_rev): New.
7523 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
7524 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
7525 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
7526 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
7527 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
7528 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
7529 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
7530 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
7531 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
7533 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
7534 Remove match_dup 1 from the inner unspec.
7535 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
7537 * config/aarch64/aarch64.md (movprfx): New attr.
7538 (length): Default movprfx to 8.
7539 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
7540 (*madd<SVE_I>, *msub<SVE_I): Likewise.
7541 (*<su>mul<SVE_I>3_highpart): Likewise.
7542 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
7543 (*v<ASHIFT><SVE_I>3): Likewise.
7544 (*<su><MAXMIN><SVE_I>3): Likewise.
7545 (*<su><MAXMIN><SVE_F>3): Likewise.
7546 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
7547 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
7548 (*div<SVE_F>4): Likewise.
7550 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
7552 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
7555 2018-07-02 Richard Biener <rguenther@suse.de>
7557 PR tree-optimization/86363
7558 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
7559 memset argument refers to a non-variable address.
7561 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
7563 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
7564 {PLUS,MINUS}_EXPR code to...
7565 (adjust_symbolic_bound): ...here,
7566 (combine_bound): ...here,
7567 (set_value_range_with_overflow): ...and here.
7569 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
7571 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
7573 (extract_range_from_abs_expr): ...here.
7575 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
7577 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
7578 -fno-omit-frame-pointer when not optimizing.
7580 2018-07-02 Martin Liska <mliska@suse.cz>
7583 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
7584 (propagate_nothrow): Likewise.
7586 2018-07-02 Martin Liska <mliska@suse.cz>
7589 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
7591 2018-07-02 David Malcolm <dmalcolm@redhat.com>
7593 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
7594 function in r262149, changing "loc" param from source_location to
7595 const dump_location_t &.
7596 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
7597 declaration, as above.
7599 2018-07-01 Paul Koning <ni1d@arrl.net>
7601 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
7602 -munit-asm, -mgnu-asm, -mdec-asm.
7603 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
7604 (pdp11_output_labelref): New.
7605 (pdp11_output_def): New.
7606 (pdp11_output_addr_vec_elt): New.
7607 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
7608 %# and %@ format codes.
7609 (pdp11_option_override): New.
7610 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
7611 (pdp11_output_ident): New.
7612 (pdp11_asm_named_section): New.
7613 (pdp11_asm_init_sections): New.
7614 (pdp11_file_start): New.
7615 (pdp11_file_end): New.
7616 (output_ascii): Use .ascii/.asciz for -mdec-asm.
7617 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
7618 %o, like %c but octal.
7619 (pdp11_option_override): New.
7620 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
7622 (DATA_SECTION_ASM_OP): Ditto.
7623 (READONLY_DATA_SECTION_ASM_OP): New.
7624 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
7625 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
7626 (ASM_OUTPUT_LABELREF): Ditto.
7627 (ASM_OUTPUT_DEF): Ditto.
7628 (ASM_OUTPUT_EXTERNAL): New.
7629 (ASM_OUTPUT_SOURCE_FILENAME): New.
7630 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
7631 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
7632 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
7633 %# and %@ format codes.
7634 * config/pdp11/pdp11.opt (mgnu-asm): New.
7635 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
7636 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
7637 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
7639 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
7641 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
7642 dereferencing path[] beyond its length.
7644 (debug_all_paths): New.
7645 (rewire_first_differing_edge): New.
7646 (adjust_paths_after_duplication): New.
7647 (duplicate_thread_path): Call adjust_paths_after_duplication.
7649 (thread_through_all_blocks): Add new argument to
7650 duplicate_thread_path.
7652 2018-06-30 Jim Wilson <jimw@sifive.com>
7654 * config/riscv/predicates.md (p2m1_shift_operand): New.
7655 (high_mask_shift_operand): New.
7656 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
7657 pattern using p2m1_shift_operand.
7658 (lshsi3_zero_extend_3+2): New combiner pattern using
7659 high_mask_shift_operand.
7661 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7663 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
7665 (vect_recog_rotate_pattern): ...here.
7666 (vect_convert_input): Try to insert casts of invariants in the
7668 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
7669 preheader to be empty.
7671 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7673 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
7674 vector type. If given, install it in the new statement's
7676 (vect_element_precision): New function.
7677 (vect_unpromoted_value): New struct.
7678 (vect_unpromoted_value::vect_unpromoted_value): New function.
7679 (vect_unpromoted_value::set_op): Likewise.
7680 (vect_look_through_possible_promotion): Likewise.
7681 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
7682 (vect_widened_op_tree, vect_convert_input): Likewise.
7683 (vect_convert_inputs, vect_convert_output): Likewise.
7684 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
7685 to handle the optional cast of the multiplication result and
7686 vect_widened_op_tree to detect the widened multiplication itself.
7687 Do not require the input and output of promotion casts to have
7688 the same sign, but base the signedness of the operation on the
7689 input rather than the result. If the pattern includes two
7690 promotions, check that those promotions have the same sign.
7691 Do not restrict the MULT_EXPR handling to a double-width result;
7692 handle quadruple-width results and wider. Use vect_convert_inputs
7693 to convert the inputs to the common type.
7694 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
7695 to handle the optional cast of the ABS result. Also allow a sign
7696 change or a sign extension between the ABS and MINUS.
7697 Use vect_widened_op_tree to detect the widened subtraction and use
7698 vect_convert_inputs to convert the inputs to the common type.
7699 (vect_handle_widen_op_by_const): Delete.
7700 (vect_recog_widen_op_pattern): New function.
7701 (vect_recog_widen_mult_pattern): Use it.
7702 (vect_recog_widen_shift_pattern): Likewise.
7703 (vect_recog_widen_sum_pattern): Use
7704 vect_look_through_possible_promotion to handle the promoted
7707 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7709 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
7710 the containing gimple_seq *.
7711 * gimple-iterator.h (gsi_for_stmt): Declare it.
7712 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
7713 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
7714 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
7715 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
7716 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
7718 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
7720 (vect_mark_pattern_stmts): ...here. Handle cases in which the
7721 statement being replaced is part of an existing pattern
7722 definition sequence, inserting the new pattern statements before
7724 (vect_pattern_recog_1): Don't return a bool. If the statement
7725 is already part of a pattern, instead apply pattern matching
7726 to the pattern definition statements. Don't clear the
7727 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
7728 (vect_pattern_recog): Don't break after the first match;
7729 continue processing the pattern definition statements instead.
7730 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
7732 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7734 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7735 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7736 (vect_recog_widen_sum_pattern): Use it.
7738 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7740 * tree-vect-loop.c (vectorizable_reduction): Assert that the
7741 phi is not a pattern statement and has not been replaced by
7742 a pattern statement.
7743 * tree-vect-patterns.c (type_conversion_p): Don't check
7744 STMT_VINFO_IN_PATTERN_P.
7745 (vect_recog_vector_vector_shift_pattern): Likewise.
7746 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
7747 the pattern statement rather than the original statement; check
7748 directly for a WIDEN_MULT_EXPR here.
7749 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
7750 vect_is_simple_use to return the pattern statement rather
7751 than the original statement; use is_pattern_stmt_p to check
7752 for such a pattern statement.
7753 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
7754 to return the pattern statement rather than the original statement;
7755 don't do the same transformation here.
7756 (vect_is_simple_use): If the defining statement has been replaced
7757 by a pattern statement, return the pattern statement instead.
7758 Remove the corresponding (local) transformation from the vectype
7761 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7763 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
7764 end and default to null.
7765 * tree-vect-loop.c (vect_create_epilog_for_reduction)
7766 (vectorizable_reduction): Update calls accordingly, dropping the
7767 gimple ** argument if the passed-back statement isn't needed.
7768 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
7769 (vect_recog_rotate_pattern): Likewise.
7770 (vect_recog_mask_conversion_pattern): Likewise.
7771 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
7772 (vect_mask_constant_operand_p): Likewise.
7773 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
7774 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
7775 (get_group_load_store_type, get_load_store_type): Likewise.
7776 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
7777 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
7778 (vectorizable_conversion, vectorizable_assignment): Likewise.
7779 (vectorizable_shift, vectorizable_operation): Likewise.
7780 (vectorizable_store, vect_is_simple_cond): Likewise.
7781 (vectorizable_condition, vectorizable_comparison): Likewise.
7782 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
7783 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
7784 and move it to the end. Cope with null def_stmt_outs.
7786 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
7788 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
7790 2018-06-29 Jeff Law <law@redhat.com>
7792 * config/v850/v850.c (v850_legitimate_address_p): Handle large
7793 displacements for TARGET_V850E2V3 and newer.
7794 (TARGET_LRA_P): Remove. Defaults to LRA now.
7795 * config/v850/v850.md (sign23byte_load): Remove.
7796 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
7797 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
7799 2018-06-29 Martin Liska <mliska@suse.cz>
7802 * coverage.c (coverage_init): Mangle full path name.
7803 * doc/invoke.texi: Document the change.
7804 * gcov-io.c (mangle_path): New.
7805 * gcov-io.h (mangle_path): Likewise.
7806 * gcov.c (mangle_name): Use mangle_path for path mangling.
7808 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7810 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
7811 if starting source register is not even.
7813 2018-06-29 Martin Liska <mliska@suse.cz>
7815 PR tree-optimization/86263
7816 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
7817 Make edge redirection.
7819 2018-06-29 David Malcolm <dmalcolm@redhat.com>
7821 * dumpfile.c (dump_loc): Add indentation based on scope depth.
7822 (dump_scope_depth): New variable.
7823 (get_dump_scope_depth): New function.
7824 (dump_begin_scope): New function.
7825 (dump_end_scope): New function.
7826 * dumpfile.h (get_dump_scope_depth): New declaration.
7827 (dump_begin_scope): New declaration.
7828 (dump_end_scope): New declaration.
7829 (class auto_dump_scope): New class.
7830 (AUTO_DUMP_SCOPE): New macro.
7831 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
7834 2018-06-29 Richard Biener <rguenther@suse.de>
7836 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
7837 compute_all_dependences succeeds.
7838 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
7839 exceed --param loop-max-datarefs-for-datadeps.
7841 2018-06-29 Jakub Jelinek <jakub@redhat.com>
7843 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
7845 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
7848 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
7849 alternative 0 in preferred_for_speed attribute.
7851 2018-06-28 Paul Koning <ni1d@arrl.net>
7853 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
7854 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
7855 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
7856 * config/pdp11/pdp11.md: Correct "length" attribute calculation
7857 for shift insn patterns.
7859 2018-06-28 David Malcolm <dmalcolm@redhat.com>
7861 * cgraph.c (cgraph_node::get_body): Replace assignments to
7862 "dump_file" with calls to set_dump_file.
7863 * dumpfile.c (alt_dump_file): Make static, and group with...
7864 (alt_flags): ...this definition.
7865 (dumps_are_enabled): New variable.
7866 (refresh_dumps_are_enabled): New function.
7867 (set_dump_file): New function.
7868 (set_alt_dump_file): New function.
7869 (gcc::dump_manager::dump_start): Replace assignments to
7870 "dump_file" and "alt_dump_file" with calls to set_dump_file and
7872 (gcc::dump_manager::dump_finish): Likewise.
7873 * dumpfile.h (alt_dump_file): Delete decl.
7874 (dumps_are_enabled): New variable decl.
7875 (set_dump_file): New function decl.
7876 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
7878 * tree-nested.c (lower_nested_functions): Replace assignments to
7879 "dump_file" with calls to set_dump_file.
7881 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
7883 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
7884 goto_locus of each outgoing edge of each basic block.
7886 2018-06-28 Richard Biener <rguenther@suse.de>
7888 * dwarf2out.c (decl_scope_table): Remove.
7889 (push_decl_scope): Likewise.
7890 (pop_decl_scope): Likewise.
7891 (gen_type_die_for_member): Do not call push/pop_decl_scope.
7892 (gen_struct_or_union_type_die): Likewise.
7893 (gen_tagged_type_die): Likewise.
7894 (dwarf2out_init): Do not initialize decl_scope_table.
7895 (dwarf2out_c_finalize): Do not free it.
7897 2018-06-28 Richard Biener <rguenther@suse.de>
7899 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
7900 deciding whether to not re-use a DIE.
7902 2018-06-28 Richard Biener <rguenther@suse.de>
7904 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
7905 DW_AT_abstract_origin attribute.
7907 2018-06-28 Martin Liska <mliska@suse.cz>
7909 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
7910 Use newly introduced constants.
7911 * tree-switch-conversion.h (struct jump_table_cluster):
7912 Define max_ratio_for_size and max_ratio_for_speed.
7914 2018-06-28 Martin Liska <mliska@suse.cz>
7916 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
7917 Add new checking assert to catch invalid state.
7918 (jump_table_cluster::can_be_handled): Handle single case
7920 (jump_table_cluster::is_beneficial): Bail out for such case.
7921 (bit_test_cluster::find_bit_tests):
7922 Add new checking assert to catch invalid state.
7923 (bit_test_cluster::can_be_handled): Handle single case
7925 (bit_test_cluster::is_beneficial): Bail out for such case.
7926 (switch_decision_tree::analyze_switch_statement):
7929 2018-06-28 Martin Liska <mliska@suse.cz>
7931 * common.opt: Introduce -completion option.
7932 * gcc.c (driver_handle_option): Handle it.
7933 (driver::main): Print completions if completion
7935 * opt-suggestions.c (option_proposer::get_completions):
7937 (option_proposer::suggest_completion): Likewise.
7938 (option_proposer::find_param_completions): Likewise.
7939 (verify_autocompletions): Likewise.
7940 (test_completion_valid_options): Likewise.
7941 (test_completion_valid_params): Likewise.
7942 (in_completion_p): Likewise.
7943 (empty_completion_p): Likewise.
7944 (test_completion_partial_match): Likewise.
7945 (test_completion_garbage): Likewise.
7946 (opt_proposer_c_tests): Likewise.
7947 * opt-suggestions.h: Declare new functions.
7948 * opts.c (common_handle_option): Handle OPT__completion_.
7949 * selftest-run-tests.c (selftest::run_tests): Add
7950 opt_proposer_c_tests.
7951 * selftest.c (assert_str_startswith): New.
7952 * selftest.h (assert_str_startswith): Likewise.
7953 (opt_proposer_c_tests): New.
7954 (ASSERT_STR_STARTSWITH): Likewise.
7956 2018-06-28 Martin Liska <mliska@suse.cz>
7958 * Makefile.in: Add opt-suggestions.o.
7959 * gcc-main.c: Include opt-suggestions.h.
7960 * gcc.c (driver::driver): Likewise.
7961 (driver::~driver): Remove m_option_suggestions.
7962 (driver::build_option_suggestions): Moved to option_proposer.
7963 (driver::suggest_option): Likewise.
7964 (driver::handle_unrecognized_options): Use option_proposer.
7965 * gcc.h (class driver): Add new memver m_option_proposer.
7966 * opt-suggestions.c: New file.
7967 * opt-suggestions.h: New file.
7969 2018-06-28 Martin Liska <mliska@suse.cz>
7971 * vec.h (class auto_string_vec): New (moved from auto_argvec).
7972 (auto_string_vec::~auto_string_vec): Likewise.
7974 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
7976 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
7977 prevent_decl_creation_for_types fields up and add reset_location field.
7978 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
7979 statement if id->reset_location is true.
7980 (copy_edges_for_bb): Do not set goto_locus on the new edges if
7981 id->reset_location is true.
7982 (copy_phis_for_bb): Force input_location on the arguments if
7983 id->reset_location is true.
7984 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
7985 is set on the function to be inlined.
7987 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
7989 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
7991 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
7993 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
7994 registers for Pmode.
7995 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
7996 hard registers for the clobbered pseudo.
7998 2018-06-27 Paul Koning <ni1d@arrl.net>
8000 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
8001 mutually exclusive options.
8002 * config/pdp11/constraints.md (h): New constraint.
8003 (O): Update definition to match shift code generation.
8004 (D): New constraint.
8005 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
8007 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
8009 (output_jump): Change arguments.
8010 (pdp11_fixed_cc_regs): New function.
8011 (pdp11_cc_mode): Ditto.
8012 (pdp11_expand_shift): Ditto.
8013 (pdp11_assemble_shift): Ditto.
8014 (pdp11_small_shift): Ditto.
8015 (pdp11_branch_cost): Remove.
8016 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
8018 (pdp11_register_move_cost): Update for CC registers.
8019 (pdp11_rtx_costs): Add case for LSHIFTRT.
8020 (pdp11_output_jump): Add CCNZ mode conditional branches.
8021 (notice_update_cc_on_set): Remove.
8022 (pdp11_cc_mode): New function.
8023 (simple_memory_operand): Correct pre/post decrement case.
8024 (no_side_effect_operand): New function.
8025 (pdp11_regno_reg_class): Add CC_REGS class.
8026 (pdp11_fixed_cc_regs): New function.
8027 (pdp11_small_shift): New function.
8028 (pdp11_expand_shift): New function to expand shift insns.
8029 (pdp11_assemble_shift): New function to output shifts.
8030 (pdp11_branch_cost): Remove.
8031 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
8032 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
8033 (WCHAR_TYPE): Ditto.
8034 (PTRDIFF_TYPE): Ditto.
8035 (ADJUST_INSN_LENGTH): New macro.
8036 (FIXED_REGISTERS): Add CC registers.
8037 (CALL_USED_REGISTERS): Ditto.
8039 (REG_CLASS_NAMES): Ditto.
8040 (REG_CLASS_CONTENTS): Ditto.
8041 (SELECT_CC_MODE): Use new function.
8042 (TARGET_FLAGS_REGNUM): New macro.
8043 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
8044 (cc0_reg_rtx): Remove.
8045 (CC_STATUS_MDEP): Remove.
8046 (CC_STATUS_MDEFP_INIT): Remove.
8047 (CC_IN_FPU): Remove.
8048 (NOTICE_UPDATE_CC): Remove.
8049 (REGISTER_NAMES): Add CC registers.
8050 (BRANCH_COST): Change to constant 1.
8051 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
8053 * config/pdp11/pdp11.opt (mbcopy): Remove.
8054 (mbcopy-builtin): Remove.
8055 (mbranch-cheap): Remove.
8056 (mbranch-expensive): Remove.
8057 * config/pdp11/predicates.md (expand_shift_operand): Update to
8058 match shift code generation.
8059 (ccnz_operator): New predicate.
8060 * doc/invoke.texi (PDP-11 Options): Remove deleted options
8061 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
8062 Remove non-existent option -mabshi, -mno-abshi. Document mutually
8064 * doc/md.texi (PDP-11): Document new D and h constraints. Update
8065 description of O constraint.
8067 2018-06-27 Jeff Law <law@redhat.com>
8068 Austin Law <austinklaw@gmail.com>
8070 * config/v850/v850.md (addsi3_set_flags): New pattern.
8071 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
8072 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
8073 (zero_extendhisi2_v850_set_flags): Likewise.
8074 (zero_extendqisi2_v850_set_flags): Likewise.
8075 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
8076 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
8077 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
8079 * config/v850/v850-protos.h (notice_update_cc): Remove.
8080 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
8081 (v850_print_operand): Handle 'D' and "d".
8082 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
8083 Add handling of arithmetic/logical operations compared against zero.
8084 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
8085 Do not look at v850_compare_op, instead get mode from last argument.
8086 (v850_gen_compare): Remove
8087 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
8088 after reload for prologue insns.
8089 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
8091 (construct_save_jarl): Likewise.
8092 (TARGET_FLAGS_REGNUM): Define.
8093 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
8094 (NOTICE_UPDATE_CC): Remove.
8095 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
8096 than cc0. Conditionalize on reload_completed.
8097 (cmpsi_insn, setfcc_insn): Likewise.
8098 (tst1 splitter): Turn into define_and_split which sets the flags
8100 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
8101 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
8102 (cstoresf4, cstoredf4): Clobber the flags.
8103 (cmpsi, cmpsf, cmpdf): Remove expanders.
8104 (setf_insn): Remove pattern.
8105 (addsi3): Turn into define_and_split which clobbers the flags after
8106 reload and a suitable pattern (addsi3_clobber_flags) for use after
8108 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
8109 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
8110 (ashrsi3, ashrsi3_v850e2): Likewise.
8111 (bins): Clobber the flags.
8112 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
8113 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
8114 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
8115 (call_value_internal_short, call_value_internal_long): Likewise.
8116 (callt_save_interrupt, callt_return_interrupt): Likewise.
8117 (save_interrupt, return_interrupt): Likewise.
8118 (callt_save_all_interrupt, save_all_interrupt): Likewise.
8119 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
8120 (restore_all_interrupt, _restore_all_interrupt): Likewise.
8121 (All FP comparisons): Only allow after reload has completed.
8123 (divh, divhu): Tweak output template.
8124 (branch_z_normal, branch_z_invert): Remove
8125 (branch_nz_normal, branch_nz_invert): Likewise.
8126 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
8128 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
8129 * config/v850/v850.c (notice_update_cc): Remove.
8130 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
8131 (CC_NO_CARRY): Likewise.
8132 (NOTICE_UPDATE_CC): Define to nothing.
8133 * config/v850/v850.md: Remove block comment on cc0 handling
8134 Remove "cc" attribute from all patterns. Remove cc_status handling
8135 from all patterns. Minor formatting fixes.
8137 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8139 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
8140 (cortex-a76.cortex-a55): Likewise.
8141 * config/aarch64/aarch64-tune.md: Regenerate.
8142 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
8143 cortex-a76.cortex-a55.
8145 2018-06-27 Jeff Law <law@redhat.com>
8147 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
8148 (MULTILIB_DIRNAMES): Similarly.
8150 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
8152 * gimple.h (gimple_return_retbnd): Delete.
8153 (gimple_return_set_retbnd): Likewise.
8154 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
8155 gimple_return_set_retbnd.
8156 * gimple-pretty-print.c (dump_gimple_return): Remove call to
8157 gimple_return_retbnd and adjust.
8158 * tree-inline.h (struct copy_body_data): Remove retbnd field.
8159 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
8160 Explicitly return NULL in a couple more cases. Move assertion
8161 on debug statements and remove unreachable code.
8162 (reset_debug_binding): Do not test id->retbnd.
8163 (expand_call_inline): Do not set it.
8165 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
8167 * configure.ac: Add --disable-gcov option.
8168 * configure: Regenerate.
8169 * Makefile.in: Honour @enable_gcov@.
8170 * doc/install.texi: Document --disable-gcov.
8172 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8174 * config/arm/arm-cpus.in (cortex-a76): New entry.
8175 (cortex-a76.cortex-a55): Likewise.
8176 * config/arm/arm-tables.opt: Regenerate.
8177 * config/arm/arm-tune.md: Likewise.
8178 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
8179 * doc/invoke.texi (ARM Options): Document cortex-a76 and
8180 cortex-a76.cortex-a55.
8182 2018-06-27 Tamar Christina <tamar.christina@arm.com>
8185 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
8187 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
8189 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
8191 (EPILOGUE_USES): Likewise.
8193 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
8195 * tree-inline.c (remap_location): New function extracted from...
8196 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
8197 (copy_phis_for_bb): ...here. Call remap_location.
8198 (copy_cfg_body): Adjust call to copy_edges_for_bb.
8200 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
8202 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
8203 unaligned vsx for 16B memset.
8205 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
8208 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
8209 ieee128_float_type_node to long_double_type_node unless
8210 TARGET_LONG_DOUBLE_128 is set.
8212 2018-06-26 David Malcolm <dmalcolm@redhat.com>
8214 * cfgloop.c (get_loop_location): Convert return type from
8215 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
8216 by implicit construction from rtx_insn *, and using
8217 dump_user_location_t::from_function_decl for the fallback case.
8218 * cfgloop.h (get_loop_location): Convert return type from
8219 location_t to dump_user_location_t.
8220 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
8221 dump_printf_loc to pass in a dump_location_t rather than a
8222 location_t, via the gimple stmt.
8223 * coverage.c (get_coverage_counts): Update calls to
8224 dump_printf_loc to pass in dump_location_t rather than a
8226 * doc/optinfo.texi (Dump types): Convert example of
8227 dump_printf_loc from taking "locus" to taking "insn". Update
8228 description of the "_loc" calls to cover dump_location_t.
8229 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
8231 (dump_user_location_t::dump_user_location_t): New constructors,
8232 from gimple *stmt and rtx_insn *.
8233 (dump_user_location_t::from_function_decl): New function.
8234 (dump_loc): Make static.
8235 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
8236 const dump_location_t &.
8237 (dump_generic_expr_loc): Delete.
8238 (dump_printf_loc): Convert param "loc" from location_t to
8239 const dump_location_t &.
8240 (selftest::test_impl_location): New function.
8241 (selftest::dumpfile_c_tests): New function.
8242 * dumpfile.h: Include "profile-count.h".
8243 (class dump_user_location_t): New class.
8244 (struct dump_impl_location_t): New struct.
8245 (class dump_location_t): New class.
8246 (dump_printf_loc): Convert 2nd param from source_location to
8247 const dump_location_t &.
8248 (dump_generic_expr_loc): Delete.
8249 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
8250 const dump_location_t &.
8251 * gimple-fold.c (fold_gimple_assign): Update call to
8252 dump_printf_loc to pass in a dump_location_t rather than a
8253 location_t, via the gimple stmt.
8254 (gimple_fold_call): Likewise.
8255 * gimple-loop-interchange.cc
8256 (loop_cand::analyze_iloop_reduction_var): Update for change to
8257 check_reduction_path.
8258 (tree_loop_interchange::interchange): Update for change to
8260 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
8261 change in return-type of find_loop_location.
8262 (graphite_regenerate_ast_isl): Likewise.
8263 * graphite-optimize-isl.c (optimize_isl): Likewise.
8264 * graphite.c (graphite_transform_loops): Likewise.
8265 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
8266 pass in a dump_location_t rather than a location_t, via the
8268 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
8269 * ipa.c (walk_polymorphic_call_targets): Likewise.
8270 * loop-unroll.c (report_unroll): Convert "locus" param from
8271 location_t to dump_location_t.
8272 (decide_unrolling): Update for change to get_loop_location's
8274 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
8275 location_t to dump_user_location_t.
8276 (grid_find_single_omp_among_assignments_1): Updates calls to
8277 dump_printf_loc to pass in a dump_location_t rather than a
8278 location_t, via the gimple stmt.
8279 (grid_parallel_clauses_gridifiable): Convert "tloc" from
8280 location_t to dump_location_t. Updates calls to dump_printf_loc
8281 to pass in a dump_location_t rather than a location_t, via the
8283 (grid_inner_loop_gridifiable_p): Likewise.
8284 (grid_dist_follows_simple_pattern): Likewise.
8285 (grid_gfor_follows_tiling_pattern): Likewise.
8286 (grid_target_follows_gridifiable_pattern): Likewise.
8287 (grid_attempt_target_gridification): Convert initialization
8288 of local "grid" from memset to zero-initialization; FIXME: does
8289 this require C++11? Update call to dump_printf_loc to pass in a
8290 optinfo_location rather than a location_t, via the gimple stmt.
8291 * profile.c (read_profile_edge_counts): Updates call to
8292 dump_printf_loc to pass in a dump_location_t rather than a
8294 (compute_branch_probabilities): Likewise.
8295 * selftest-run-tests.c (selftest::run_tests): Call
8297 * selftest.h (dumpfile_c_tests): New decl.
8298 * tree-loop-distribution.c (pass_loop_distribution::execute):
8299 Update for change in return type of find_loop_location.
8300 * tree-parloops.c (parallelize_loops): Likewise.
8301 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
8302 "locus" from location_t to dump_user_location_t.
8303 (canonicalize_loop_induction_variables): Likewise.
8304 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
8305 for change in return type of find_loop_location.
8306 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
8307 to dump_printf_loc to pass in a dump_location_t rather than a
8308 location_t, via the stmt.
8309 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
8311 * tree-vect-loop-manip.c (find_loop_location): Convert return
8312 type from source_location to dump_user_location_t.
8313 (vect_do_peeling): Update for above change.
8314 (vect_loop_versioning): Update for change in type of
8316 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
8317 from location_t to dump_user_location_t.
8318 (vect_estimate_min_profitable_iters): Update for change in type
8320 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
8321 location_t to dump_location_t.
8322 (vect_slp_bb): Update for change in type of vect_location.
8323 * tree-vectorizer.c (vect_location): Convert from source_location
8324 to dump_user_location_t.
8325 (try_vectorize_loop_1): Update for change in vect_location's type.
8326 (vectorize_loops): Likewise.
8327 (increase_alignment): Likewise.
8328 * tree-vectorizer.h (vect_location): Convert from source_location
8329 to dump_user_location_t.
8330 (find_loop_location): Convert return type from source_location to
8331 dump_user_location_t.
8332 (check_reduction_path): Convert 1st param from location_t to
8333 dump_user_location_t.
8334 * value-prof.c (check_counter): Update call to dump_printf_loc to
8335 pass in a dump_user_location_t rather than a location_t; update
8336 call to error_at for change in type of "locus".
8337 (check_ic_target): Update call to dump_printf_loc to
8338 pass in a dump_user_location_t rather than a location_t, via the
8341 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
8343 * config/s390/s390.h (enum processor_flags): Do not use
8344 default tune parameter when -march was specified.
8346 2018-06-26 Jakub Jelinek <jakub@redhat.com>
8349 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
8350 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
8353 2018-06-26 Richard Biener <rguenther@suse.de>
8355 PR tree-optimization/86287
8357 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
8358 (vect_analyze_loop): Initialize n_stmts.
8360 2018-06-26 Richard Biener <rguenther@suse.de>
8363 * fold-const.c (fold_convertible_p): Pointer extension
8366 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
8369 * dwarf2out.c (loc_list_has_views): Adjust comments.
8370 (dw_loc_list): Split single cross-partition range with
8373 2018-06-25 Jeff Law <law@redhat.com>
8375 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
8376 on -mbig-switch by default.
8378 * config/v850/predicates.md (const_float_1_operand): Fix match_code
8380 (const_float_0_operand): Remove unused predicate.
8381 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
8382 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
8383 (recipsf2): New expander. Original pattern now called
8385 (recipdf2, recipdf2_insn): Similarly.
8386 (rsqrtsf2, rsqrtsf2_insn): Similarly
8387 (rsqrtdf2, rsqrtdf2_insn): Similarly
8389 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
8391 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
8392 Simplify logic for FreeBSD (twice).
8394 2018-06-25 Martin Sebor <msebor@redhat.com>
8396 PR tree-optimization/86204
8397 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
8398 a strnlen result if it's less than the length of the string.
8400 2018-06-25 Martin Sebor <msebor@redhat.com>
8402 PR tree-optimization/85700
8403 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
8404 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
8405 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
8407 2018-06-25 Martin Sebor <msebor@redhat.com>
8409 * doc/extend.texi (Zero-length arrays): Update and clarify.
8411 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
8413 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
8414 added IEEE/IBM long double multilib support on PowerPC little
8415 endian Linux systems.
8416 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
8417 (MULTILIB_DEFAULTS): Likewise.
8418 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8420 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
8421 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
8422 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
8424 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
8427 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
8428 (REORDER_45): Likewise.
8430 2018-06-25 Jeff Law <law@redhat.com>
8432 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
8433 dividend to 32 bits. Adjust length.
8434 (udivmodhi4): Cleanup output template. Fix length.
8436 2018-06-25 Carl Love <cel@us.ibm.com>
8438 * config/rs6000/vsx.md: Change word selector to prefered location.
8440 2018-06-25 Richard Biener <rguenther@suse.de>
8442 PR tree-optimization/86304
8443 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
8444 epilogue-if-converted loops as well.
8446 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
8448 * lto-section-out.c (lto_begin_section): Do not print section
8449 name for noaddr and unnumbered dumps.
8451 2018-06-25 Richard Biener <rguenther@suse.de>
8453 * tree-vectorizer.h (struct vec_info_shared): New structure
8454 with parts split out from struct vec_info and loop_nest from
8455 struct _loop_vec_info.
8456 (struct vec_info): Adjust accordingly.
8457 (struct _loop_vec_info): Likewise.
8458 (LOOP_VINFO_LOOP_NEST): Adjust.
8459 (LOOP_VINFO_DATAREFS): Likewise.
8460 (LOOP_VINFO_DDRS): Likewise.
8461 (struct _bb_vec_info): Likewise.
8462 (BB_VINFO_DATAREFS): Likewise.
8463 (BB_VINFO_DDRS): Likewise.
8464 (struct _stmt_vec_info): Add dr_aux member.
8465 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
8466 (DR_MISALIGNMENT_UNINITIALIZED): New.
8467 (set_dr_misalignment): Adjust.
8468 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
8469 (vect_analyze_loop): Adjust prototype.
8470 (vect_analyze_loop_form): Likewise.
8471 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
8472 Compute dependences lazily.
8473 (vect_record_base_alignments): Use shared datarefs/ddrs.
8474 (vect_verify_datarefs_alignment): Likewise.
8475 (vect_analyze_data_refs_alignment): Likewise.
8476 (vect_analyze_data_ref_accesses): Likewise.
8477 (vect_analyze_data_refs): Likewise.
8478 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
8479 constructor parameter for shared part.
8480 (vect_analyze_loop_form): Pass in shared part and adjust.
8481 (vect_analyze_loop_2): Pass in storage for the number of
8482 stmts. Move loop nest finding to the caller. Compute
8484 (vect_analyze_loop): Pass in shared part.
8485 (vect_transform_loop): Verify shared datarefs are unchanged.
8486 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
8487 constructor parameter for shared part.
8488 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
8489 (vect_slp_bb): Verify shared datarefs are unchanged before
8491 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
8493 (new_stmt_vec_info): Initialize DR_AUX misalignment to
8494 DR_MISALIGNMENT_UNINITIALIZED.
8495 * tree-vectorizer.c (vec_info::vec_info): Add constructor
8496 parameter for shared part.
8497 (vec_info::~vec_info): Adjust.
8498 (vec_info_shared::vec_info_shared): New.
8499 (vec_info_shared::~vec_info_shared): Likewise.
8500 (vec_info_shared::save_datarefs): Likewise.
8501 (vec_info_shared::check_datarefs): Likewise.
8502 (try_vectorize_loop_1): Construct shared part live for analyses
8503 of a single loop for multiple vector sizes.
8504 * tree-parloops.c (gather_scalar_reductions): Adjust.
8506 2018-06-25 Richard Biener <rguenther@suse.de>
8508 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
8509 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
8510 (vect_analyze_data_refs): Remove similar code from here and
8511 simplify accordingly.
8513 2018-06-25 Richard Biener <rguenther@suse.de>
8515 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
8516 for reverse storage order accesses rather than asserting
8517 they cannot happen here.
8519 2018-06-25 Tom de Vries <tdevries@suse.de>
8522 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
8523 Use data16 instead of .byte for insn prefix.
8525 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
8528 * parser.c (make_char_string_pack): Pass this literal chars
8529 through cpp_interpret_string.
8530 (cp_parser_userdef_numeric_literal): Check the result of
8531 make_char_string_pack.
8533 2018-06-24 Maya Rashish <coypu@sdf.org>
8535 * ginclude/stddef.h: Simplify conditions around avoiding
8536 re-definition of __size_t.
8538 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
8540 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
8541 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
8543 2018-06-22 Maya Rashish <coypu@sdf.org>
8545 * doc/invoke.texi (mno-fancy-math-387): Update for changes
8546 made to OpenBSD and NetBSD through the years.
8548 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8550 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
8551 behavior of vec_pack (vector double, vector double) to match
8552 behavior of vec_float2 (vector double, vector double).
8554 2018-06-22 Olivier Hainque <hainque@adacore.com>
8556 * gimplify.c (gimplify_function_tree): Prevent creation
8557 of a trampoline for the address of the current function
8558 passed to entry/exit instrumentation hooks.
8560 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
8563 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
8566 2018-06-22 Martin Liska <mliska@suse.cz>
8568 PR tree-optimization/86263
8569 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
8570 Bail out if is_enabled is false.
8571 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
8573 (jump_table_cluster::is_enabled): New function.
8575 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
8577 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
8578 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
8579 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
8580 (lto_input_ts_binfo_tree_pointers): Likewise.
8581 * tree-streamer-out.c (streamer_write_tree_bitfields,
8582 write_ts_binfo_tree_pointers): Likewise.
8583 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
8585 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
8587 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
8589 2018-06-22 Martin Liska <mliska@suse.cz>
8591 * symbol-summary.h (get): Make it pure and inline move
8592 functionality from ::get function.
8593 (get): Remove and inline into ::get and ::get_create.
8594 (get_create): Move code from ::get function.
8596 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8599 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
8600 -x assembler-with-cpp.
8602 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8604 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
8605 _FILE_OFFSET_BITS=64 for C++.
8607 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
8609 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
8610 conversion insn that shows up when pr85657-3.c is compiled using
8611 IEEE 128-bit long double.
8612 (neg<mode>2_internal): Use the correct mode to check whether the
8613 mode is IBM extended.
8614 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
8615 multiply and divide external functions from being created more
8618 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8620 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
8622 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
8623 the edge can be forwarded.
8624 (cfg_layout_merge_blocks): Likewise.
8626 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8628 * except.c (finish_eh_generation): Commit edge insertions only after
8629 the EH edges have been redirected from post-landing to landing pads.
8631 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8633 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
8634 create_tmp_var_for to create the FRAME decl.
8635 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
8637 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8639 * tree-inline.c (copy_edges_for_bb): Minor tweak.
8640 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
8641 debug statement when resetting its value.
8642 (expand_call_inline): Copy the locus of the call onto the assignment
8643 of the return value, if any. Use local variable in more cases.
8645 2018-06-21 Martin Liska <mliska@suse.cz>
8647 * ipa-pure-const.c (propagate_nothrow): Use
8648 funct_state_summaries->get.
8649 (dump_malloc_lattice): Likewise.
8650 (propagate_malloc): Likewise.
8652 2018-06-21 Richard Biener <rguenther@suse.de>
8654 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
8655 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
8656 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
8658 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
8659 BLOCK_ABSTRACT_ORIGIN unconditionally.
8661 2018-06-21 David Malcolm <dmalcolm@redhat.com>
8663 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
8665 * ipa-reference.c (ipa_reference_c_finalize): Delete
8666 ipa_ref_opt_sum_summaries and set it to NULL.
8668 2018-06-21 Tom de Vries <tdevries@suse.de>
8670 PR tree-optimization/85859
8671 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
8672 test with comment from bb_no_side_effects_p.
8674 2018-06-21 Richard Biener <rguenther@suse.de>
8676 PR tree-optimization/86232
8677 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
8678 max for constant niter.
8680 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
8682 * config/aarch64/aarch64-simd.md
8683 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
8685 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
8687 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
8688 Make opernads of the unspec commutative.
8690 2018-06-21 Richard Biener <rguenther@suse.de>
8692 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
8693 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
8694 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
8695 (vect_analyze_data_ref_dependence): Re-order checks to deal with
8697 (vect_record_base_alignments): Do not record base alignment
8698 for gathers or scatters.
8699 (vect_compute_data_ref_alignment): Drop return value that is always
8700 true. Bail out early for gathers or scatters.
8701 (vect_enhance_data_refs_alignment): Bail out early for gathers
8703 (vect_find_same_alignment_drs): Likewise.
8704 (vect_analyze_data_refs_alignment): Remove dead code.
8705 (vect_slp_analyze_and_verify_node_alignment): Likewise.
8706 (vect_analyze_data_refs): For possible gathers or scatters do
8707 not create an alternate DR, just check their possible validity
8708 and mark them. Adjust DECL_NONALIASED handling to not rely
8710 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
8711 update inits of gathers or scatters.
8712 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
8713 Also copy gather/scatter flag to pattern vinfo.
8715 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
8717 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
8718 behavior of vec_packsu (vector unsigned long long, vector unsigned
8719 long long) to match behavior of vec_packs with same signature.
8721 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
8722 Thomas Schwinge <thomas@codesourcery.com>
8723 Cesar Philippidis <cesar@codesourcery.com>
8725 * gimplify.c (gimplify_scan_omp_clauses): Add support for
8726 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
8727 (gimplify_adjust_omp_clauses): Likewise.
8728 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
8729 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
8730 (gimplify_omp_target_update): Update handling of acc update and
8732 * omp-low.c (install_var_field): Remove unused parameter
8733 base_pointers_restrict.
8734 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
8735 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
8737 (omp_target_base_pointers_restrict_p): Delete.
8738 (scan_omp_target): Update call to scan_sharing_clauses.
8739 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
8741 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
8742 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
8743 (convert_local_omp_clauses): Likewise.
8744 * tree-pretty-print.c (dump_omp_clause): Likewise.
8745 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
8747 (omp_clause_code_name): Likewise.
8749 2018-06-20 Jakub Jelinek <jakub@redhat.com>
8752 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
8755 PR tree-optimization/86231
8756 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
8757 anti-range don't overwrite *vr0min before using it to compute *vr0max.
8759 2018-06-20 Tom de Vries <tdevries@suse.de>
8761 PR tree-optimization/86097
8762 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
8763 iv type if signedness of iv type is not the same as that of *nit.
8765 2018-06-20 Jakub Jelinek <jakub@redhat.com>
8767 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
8768 EDGE_EH edges, verify they are all EDGE_EH.
8770 2018-06-20 Maya Rashish <coypu@sdf.org>
8772 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
8774 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8776 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
8777 * config/aarch64/aarch64.c (xgene1_tunings): Add
8778 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
8779 (aarch64_mode_valid_for_sched_fusion_p):
8780 Allow 16-byte modes.
8781 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
8782 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
8784 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
8786 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
8787 * config/aarch64/iterators.md (VQ2): New mode iterator.
8789 2018-06-20 Martin Liska <mliska@suse.cz>
8791 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
8792 Change default ratio from 10 to 8.
8794 2018-06-20 Martin Liska <mliska@suse.cz>
8796 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
8798 (bit_test_cluster::find_bit_tests): Likewise.
8799 (switch_decision_tree::analyze_switch_statement): Find clusters.
8800 * tree-switch-conversion.h (struct jump_table_cluster): Document
8803 2018-06-20 Martin Liska <mliska@suse.cz>
8805 * tree-switch-conversion.c (switch_conversion::collect):
8806 Record m_uniq property.
8807 (switch_conversion::expand): Bail out for special conditions.
8808 (group_cluster::~group_cluster): New.
8809 (group_cluster::group_cluster): Likewise.
8810 (group_cluster::dump): Likewise.
8811 (jump_table_cluster::emit): New.
8812 (switch_decision_tree::fix_phi_operands_for_edges): New.
8813 (struct case_node): Remove struct.
8814 (jump_table_cluster::can_be_handled): New.
8815 (case_values_threshold): Moved to header.
8816 (reset_out_edges_aux): Likewise.
8817 (jump_table_cluster::is_beneficial): New.
8818 (bit_test_cluster::can_be_handled): Likewise.
8819 (add_case_node): Remove.
8820 (bit_test_cluster::is_beneficial): New.
8821 (case_bit_test::cmp): New.
8822 (bit_test_cluster::emit): New.
8823 (expand_switch_as_decision_tree_p): Remove.
8824 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
8825 (fix_phi_operands_for_edge): Likewise.
8826 (switch_decision_tree::analyze_switch_statement): New.
8827 (compute_cases_per_edge): Move ...
8828 (switch_decision_tree::compute_cases_per_edge): ... here.
8829 (try_switch_expansion): Likewise.
8830 (switch_decision_tree::try_switch_expansion): Likewise.
8831 (record_phi_operand_mapping): Likewise.
8832 (switch_decision_tree::record_phi_operand_mapping): Likewise.
8833 (emit_case_decision_tree): Likewise.
8834 (switch_decision_tree::emit): Likewise.
8835 (balance_case_nodes): Likewise.
8836 (switch_decision_tree::balance_case_nodes): Likewise.
8837 (dump_case_nodes): Likewise.
8838 (switch_decision_tree::dump_case_nodes): Likewise.
8839 (emit_jump): Likewise.
8840 (switch_decision_tree::emit_jump): Likewise.
8841 (emit_cmp_and_jump_insns): Likewise.
8842 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
8843 (emit_case_nodes): Likewise.
8844 (switch_decision_tree::emit_case_nodes): Likewise.
8845 (conditional_probability): Remove.
8846 * tree-switch-conversion.h (enum cluster_type): New.
8848 (struct cluster): Likewise.
8849 (cluster::cluster): Likewise.
8850 (struct simple_cluster): Likewise.
8851 (simple_cluster::simple_cluster): Likewise.
8852 (struct group_cluster): Likewise.
8853 (struct jump_table_cluster): Likewise.
8854 (struct bit_test_cluster): Likewise.
8855 (struct min_cluster_item): Likewise.
8856 (struct case_tree_node): Likewise.
8857 (case_tree_node::case_tree_node): Likewise.
8858 (jump_table_cluster::case_values_threshold): Likewise.
8859 (struct case_bit_test): Likewise.
8860 (struct switch_decision_tree): Likewise.
8861 (struct switch_conversion): Likewise.
8862 (switch_decision_tree::reset_out_edges_aux): Likewise.
8864 2018-06-20 Martin Liska <mliska@suse.cz>
8866 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
8867 (hoist_edge_and_branch_if_true): Likewise.
8868 (expand_switch_using_bit_tests_p): Likewise.
8869 (struct case_bit_test): Likewise.
8870 (case_bit_test_cmp): Likewise.
8871 (emit_case_bit_tests): Likewise.
8872 (switch_conversion::switch_conversion): New class.
8873 (struct switch_conv_info): Remove old struct.
8874 (collect_switch_conv_info): More to ...
8875 (switch_conversion::collect): ... this.
8876 (check_range): Likewise.
8877 (switch_conversion::check_range): Likewise.
8878 (check_all_empty_except_final): Likewise.
8879 (switch_conversion::check_all_empty_except_final): Likewise.
8880 (check_final_bb): Likewise.
8881 (switch_conversion::check_final_bb): Likewise.
8882 (create_temp_arrays): Likewise.
8883 (switch_conversion::create_temp_arrays): Likewise.
8884 (free_temp_arrays): Likewise.
8885 (gather_default_values): Likewise.
8886 (switch_conversion::gather_default_values): Likewise.
8887 (build_constructors): Likewise.
8888 (switch_conversion::build_constructors): Likewise.
8889 (constructor_contains_same_values_p): Likewise.
8890 (switch_conversion::contains_same_values_p): Likewise.
8891 (array_value_type): Likewise.
8892 (switch_conversion::array_value_type): Likewise.
8893 (build_one_array): Likewise.
8894 (switch_conversion::build_one_array): Likewise.
8895 (build_arrays): Likewise.
8896 (switch_conversion::build_arrays): Likewise.
8897 (gen_def_assigns): Likewise.
8898 (switch_conversion::gen_def_assigns): Likewise.
8899 (prune_bbs): Likewise.
8900 (switch_conversion::prune_bbs): Likewise.
8901 (fix_phi_nodes): Likewise.
8902 (switch_conversion::fix_phi_nodes): Likewise.
8903 (gen_inbound_check): Likewise.
8904 (switch_conversion::gen_inbound_check): Likewise.
8905 (process_switch): Use the newly created class.
8906 (switch_conversion::expand): New.
8907 (switch_conversion::~switch_conversion): New.
8908 * tree-switch-conversion.h: New file.
8910 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8912 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
8913 tree-vect-patterns.c.
8914 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
8915 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
8916 (vect_recog_sad_pattern): Likewise.
8917 (vect_recog_widen_sum_pattern): Likewise.
8918 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
8919 (vect_recog_widen_shift_pattern): Remove the type_in argument.
8920 (vect_recog_rotate_pattern): Likewise.
8921 (vect_recog_mult_pattern): Likewise.
8922 (vect_recog_vector_vector_shift_pattern): Likewise.
8923 (vect_recog_divmod_pattern): Likewise.
8924 (vect_recog_mixed_size_cond_pattern): Likewise.
8925 (vect_recog_bool_pattern): Likewise.
8926 (vect_recog_mask_conversion_pattern): Likewise.
8927 (vect_try_gather_scatter_pattern): Likewise.
8928 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
8929 (vect_recog_over_widening_pattern): Likewise.
8930 (vect_recog_gather_scatter_pattern): Likewise.
8931 (vect_recog_func_ptr): Move from tree-vectorizer.h
8932 (vect_vect_recog_func_ptrs): Move further down the file.
8933 (vect_recog_func): Likewise. Remove the third argument.
8934 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
8935 (vect_pattern_recog_1): Expect the pattern function to do any
8936 necessary target tests. Also expect it to provide a vector type.
8937 Remove the type_in handling.
8939 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8941 * tree-vect-patterns.c (vect_pattern_detected): New function.
8942 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
8943 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
8944 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
8945 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
8946 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
8947 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
8948 (vect_recog_mask_conversion_pattern)
8949 (vect_try_gather_scatter_pattern): Likewise.
8951 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8953 * tree-vect-patterns.c (vect_get_internal_def): New function.
8954 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
8955 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
8956 (search_type_for_mask_1): Use it.
8958 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8960 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
8961 redundant WIDEN_SUM_EXPR handling.
8962 (vect_recog_sad_pattern): Likewise.
8964 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8966 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
8967 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
8968 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
8969 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
8970 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
8972 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8974 * tree-vect-stmts.c (vectorizable_call): Make sure that we
8975 use the stmt_vec_info of the original bb statement for the
8976 new zero assignment, even if the call is part of a pattern.
8978 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8980 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
8981 that the sequence is attached to the original statement rather
8982 than the pattern statement.
8983 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
8984 PATTERN_DEF_SEQ from the original statement rather than
8985 the main pattern statement.
8986 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
8987 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
8988 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
8990 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8992 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
8993 definition statements before the early exit for statements that aren't
8995 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
8997 (vect_transform_loop): ...here. Process pattern definition
8998 statements without first checking whether the main pattern
8999 statement is live or relevant.
9001 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
9003 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
9004 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
9006 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
9008 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
9009 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
9010 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
9011 (expand_block_compare): Change select_block_compare_mode call.
9012 (expand_strncmp_align_check): Use new functions, fix comment.
9013 (emit_final_str_compare_gpr): New function.
9014 (expand_strn_compare): Refactor and clean up code.
9015 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
9017 2018-06-19 Tony Reix <tony.reix@atos.com>
9018 Damien Bergamini <damien.bergamini@atos.com>
9019 David Edelsohn <dje.gcc@gmail.com>
9021 * collect2.c (static_obj): New variable.
9022 (static_libs): New variable.
9023 (is_in_list): Uncomment declaration.
9024 (main): Track AIX libraries linked statically.
9025 (is_in_list): Uncomment definition.
9026 (scan_prog_file): Don't add AIX shared libraries initializer
9027 to constructor list if linking statically.
9029 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
9031 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
9033 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
9035 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
9037 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
9040 2018-06-19 Martin Liska <mliska@suse.cz>
9042 * config/i386/i386.c (ix86_can_inline_p): Do not use
9043 ipa_fn_summaries::get_create.
9044 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
9046 (devirtualization_time_bonus): Likewise.
9047 (ipcp_propagate_stage): Likewise.
9048 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
9049 (edge_set_predicate): Likewise.
9050 (evaluate_conditions_for_known_args): Likewise.
9051 (evaluate_properties_for_edge): Likewise.
9052 (ipa_call_summary::reset): Tranform to ...
9053 (ipa_call_summary::~ipa_call_summary): ... this.
9054 (ipa_fn_summary::reset): Transform to ...
9055 (ipa_fn_summary::~ipa_fn_summary): ... this.
9056 (ipa_fn_summary_t::remove): Rename to ...
9057 (ipa_fn_summary_t::remove_callees): ... this.
9058 (ipa_fn_summary_t::duplicate): Use placement new
9059 instead of memory copy.
9060 (ipa_call_summary_t::duplicate): Likewise.
9061 (ipa_call_summary_t::remove): Remove.
9062 (dump_ipa_call_summary): Change get_create to get.
9063 (ipa_dump_fn_summary): Dump only when summary exists.
9064 (analyze_function_body): Use symbol_summary::get instead
9066 (compute_fn_summary): Likewise.
9067 (estimate_edge_devirt_benefit): Likewise.
9068 (estimate_edge_size_and_time): Likewise.
9069 (inline_update_callee_summaries): Likewise.
9070 (remap_edge_change_prob): Likewise.
9071 (remap_edge_summaries): Likewise.
9072 (ipa_merge_fn_summary_after_inlining): Likewise.
9073 (write_ipa_call_summary): Likewise.
9074 (ipa_fn_summary_write): Likewise.
9075 (ipa_free_fn_summary): Likewise.
9076 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
9077 (struct ipa_call_summary): Likewise.
9078 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
9080 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
9081 (estimate_size_after_inlining): Likewise.
9082 (estimate_growth): Likewise.
9083 (growth_likely_positive): Likewise.
9084 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
9085 (inline_call): Likewise.
9086 * ipa-inline.c (caller_growth_limits): Likewise.
9087 (can_inline_edge_p): Likewise.
9088 (can_inline_edge_by_limits_p): Likewise.
9089 (compute_uninlined_call_time): Likewise.
9090 (compute_inlined_call_time): Likewise.
9091 (want_inline_small_function_p): Likewise.
9092 (edge_badness): Likewise.
9093 (update_caller_keys): Likewise.
9094 (update_callee_keys): Likewise.
9095 (inline_small_functions): Likewise.
9096 (inline_to_all_callers_1): Likewise.
9097 (dump_overall_stats): Likewise.
9098 (early_inline_small_functions): Likewise.
9099 (early_inliner): Likewise.
9100 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
9101 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
9102 * ipa-pure-const.c (malloc_candidate_p): Likewise.
9103 * ipa-split.c (execute_split_functions): Likewise.
9104 * symbol-summary.h: Likewise.
9105 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
9107 2018-06-19 Richard Biener <rguenther@suse.de>
9109 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
9110 (vectorize_loops): ... here. Fix dbgcnt handling.
9111 (try_vectorize_loop): Wrap try_vectorize_loop_1.
9113 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
9116 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
9117 ieee128 argument takes up only one (vector) register, not two (floating
9120 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
9122 * gimplify.c (gimplify_init_constructor): Really never clear for an
9123 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
9125 2018-06-19 Richard Biener <rguenther@suse.de>
9127 PR tree-optimization/86179
9128 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
9129 after failed recognition.
9131 2018-06-18 Martin Sebor <msebor@redhat.com>
9134 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
9135 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
9136 Handle integer subtraction.
9137 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
9138 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
9140 2018-06-18 David Malcolm <dmalcolm@redhat.com>
9142 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
9143 param from rtx to rtx_insn *.
9144 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
9146 (frv_ifcvt_modify_insn): Likwise.
9147 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
9148 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
9149 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
9150 as_a <rtx_insn *> cast to local "unprotected_region" once
9151 it's been established that it's not NULL or pc_rtx.
9152 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
9153 param "sethi" from rtx to rtx_insn *.
9154 (nds32_group_float_insns): Likewise for param "insn".
9155 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
9157 (vax_output_int_subtract): Likewise.
9158 * config/vax/vax.c (vax_output_int_add): Likewise for param
9160 (vax_output_int_subtract): Likewise.
9161 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
9162 (emit_pattern_after): Likewise for param "after".
9163 (emit_insn_after): Likewise.
9164 (emit_jump_insn_after): Likewise.
9165 (emit_call_insn_after): Likewise.
9166 (emit_debug_insn_after): Likewise.
9167 (emit_pattern_before): Likewise for param "before".
9168 (emit_insn_before): Likewise.
9169 (emit_jump_insn_before): Likewise.
9170 * final.c (get_insn_template): Likewise for param "insn", removing
9172 * output.h (get_insn_template): Likewise for 2nd param.
9173 * rtl.h (emit_insn_before): Likewise.
9174 (emit_jump_insn_before): Likewise.
9175 (emit_debug_insn_before_noloc): Likewise.
9176 (emit_insn_after): Likewise.
9177 (emit_jump_insn_after): Likewise.
9178 (emit_call_insn_after): Likewise.
9179 (emit_debug_insn_after): Likewise.
9180 (set_insn_deleted): Likewise for param.
9182 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
9185 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
9186 floating point modes, so that IFmode is numerically greater than
9187 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
9188 to declare the ordering. This prevents IFmode from being
9189 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
9190 machine. Include rs6000-modes.h to share the fractional values
9191 between genmodes* and the rest of the compiler.
9195 * config/rs6000/rs6000-modes.h: New file.
9196 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
9197 meaning of rs6000_long_double_size so that 126..128 selects an
9198 appropriate 128-bit floating point type.
9199 (rs6000_option_override_internal): Likewise.
9200 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
9201 (TARGET_LONG_DOUBLE_128): Change the meaning of
9202 rs6000_long_double_size so that 126..128 selects an appropriate
9203 128-bit floating point type.
9204 (LONG_DOUBLE_TYPE_SIZE): Update comment.
9205 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
9206 source and destination to match the standard usage.
9207 (truncifkf2): Likewise.
9208 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
9209 ISA 2.07 to use an explicit clobber, instead of passing in a
9211 (copysign<mode>3_soft): Likewise.
9213 2018-06-18 David Malcolm <dmalcolm@redhat.com>
9215 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
9216 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
9217 (vect_slp_analyze_instance_dependence): Likewise.
9218 (vect_enhance_data_refs_alignment): Likewise.
9219 (vect_analyze_data_refs_alignment): Likewise.
9220 (vect_slp_analyze_and_verify_instance_alignment
9221 (vect_analyze_data_ref_accesses): Likewise.
9222 (vect_prune_runtime_alias_test_list): Likewise.
9223 (vect_analyze_data_refs): Likewise.
9224 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
9225 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9226 (vect_analyze_scalar_cycles_1): Likewise.
9227 (vect_get_loop_niters): Likewise.
9228 (vect_analyze_loop_form_1): Likewise.
9229 (vect_update_vf_for_slp): Likewise.
9230 (vect_analyze_loop_operations): Likewise.
9231 (vect_analyze_loop): Likewise.
9232 (vectorizable_induction): Likewise.
9233 (vect_transform_loop): Likewise.
9234 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
9235 * tree-vect-slp.c (vect_analyze_slp): Likewise.
9236 (vect_make_slp_decision): Likewise.
9237 (vect_detect_hybrid_slp): Likewise.
9238 (vect_slp_analyze_operations): Likewise.
9239 (vect_slp_bb): Likewise.
9240 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
9241 (vectorizable_bswap): Likewise.
9242 (vectorizable_call): Likewise.
9243 (vectorizable_simd_clone_call): Likewise.
9244 (vectorizable_conversion): Likewise.
9245 (vectorizable_assignment): Likewise.
9246 (vectorizable_shift): Likewise.
9247 (vectorizable_operation): Likewise.
9248 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
9250 2018-06-18 Martin Sebor <msebor@redhat.com>
9252 PR tree-optimization/81384
9253 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
9254 * builtins.c (expand_builtin_strnlen): New function.
9255 (expand_builtin): Call it.
9256 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
9257 * builtins.def (BUILT_IN_STRNLEN): New.
9258 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
9259 Warn for bounds in excess of maximum object size.
9260 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
9261 single-value ranges. Handle strnlen.
9262 (handle_builtin_strlen): Handle strnlen.
9263 (strlen_check_and_optimize_stmt): Same.
9264 * doc/extend.texi (Other Builtins): Document strnlen.
9266 2018-06-18 Maya Rashish <coypu@sdf.org>
9268 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
9269 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
9270 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
9272 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
9274 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
9276 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9278 * tree.c (escaped_string::escape): Replace cast to char * by
9279 const_cast<char *> (unescaped).
9281 2018-06-18 Nick Clifton <nickc@redhat.com>
9284 * tree.c (escaped_string): New class. Converts an unescaped
9285 string into its escaped equivalent.
9286 (warn_deprecated_use): Use the new class to convert the
9287 deprecation message, if present.
9288 (test_escaped_strings): New self test.
9289 (test_c_tests): Add test_escaped_strings.
9290 * doc/extend.texi (deprecated): Add a note that the
9291 deprecation message is affected by the -fmessage-length
9292 option, and that control characters will be escaped.
9293 (#pragma GCC error): Document this pragma.
9294 (#pragma GCC warning): Likewise.
9295 * doc/invoke.texi (-fmessage-length): Document this option's
9296 effect on the #warning and #error preprocessor directives and
9297 the deprecated attribute.
9299 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
9301 * tree.c (decl_value_expr_lookup): Revert latest change.
9302 (decl_value_expr_insert): Likewise.
9304 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
9306 * gimplify.c (nonlocal_vlas): Delete.
9307 (nonlocal_vla_vars): Likewise.
9308 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
9310 (gimplify_body): Do not create and destroy nonlocal_vlas.
9311 * tree-nested.c: Include diagnostic.h.
9312 (use_pointer_in_frame): Tweak.
9313 (lookup_field_for_decl): Add assertion and declare the transformation.
9314 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
9315 internal error when the reference is in a wrong context. Do not
9316 create a debug decl by default.
9317 (note_nonlocal_block_vlas): Delete.
9318 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
9319 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
9320 create a debug decl by default.
9321 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
9322 call is in a wrong context.
9323 (fixup_vla_decls): New function.
9324 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
9325 debug variables were created.
9326 * tree.c (decl_value_expr_lookup): Add checking assertion.
9327 (decl_value_expr_insert): Likewise.
9329 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
9332 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
9333 * tree-scalar-evolution.c (interpret_expr): Likewise.
9334 (expression_expensive_p): Likewise.
9335 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
9336 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
9337 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
9338 (ssa_defined_by_minus_one_stmt_p): New.
9340 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
9343 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
9344 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
9345 * dojump.c (do_jump): Likewise.
9346 * expr.c (expand_expr_real_2): Check operand type's sign.
9347 * fold-const.c (const_unop): Handle ABSU_EXPR.
9348 (fold_abs_const): Likewise.
9349 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
9350 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
9351 (strip_sign_op_1): Likesise.
9352 * match.pd: Add new pattern to generate ABSU_EXPR.
9353 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
9354 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
9355 * tree-eh.c (operation_could_trap_helper_p): Likewise.
9356 * tree-inline.c (estimate_operator_cost): Likewise.
9357 * tree-pretty-print.c (dump_generic_node): Likewise.
9358 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
9359 * tree.def (ABSU_EXPR): New.
9361 2018-06-16 Jakub Jelinek <jakub@redhat.com>
9364 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
9365 documented as preserved for backward compatibility only.
9366 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
9368 PR rtl-optimization/86108
9369 * bb-reorder.c (create_forwarder_block): Renamed to ...
9370 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
9371 jump from new landing pad to the second part.
9372 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
9375 2018-06-15 Jakub Jelinek <jakub@redhat.com>
9378 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
9379 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
9380 Only call store_expr for halves if the mode is the same.
9383 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
9384 Fix up comment formatting.
9386 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
9388 * typed-splay-tree.h (typed_splay_tree::remove): New function.
9389 (typed_splay_tree::closure,
9390 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
9391 (typed_splay_tree::typed_splay_tree,
9392 typed_splay_tree::operator =): Declared private.
9393 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
9394 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
9395 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
9396 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
9397 typed_splay_tree::splay_tree_splay,
9398 typed_splay_tree::splay_tree_foreach_helper,
9399 typed_splay_tree::splay_tree_insert,
9400 typed_splay_tree::splay_tree_remove,
9401 typed_splay_tree::splay_tree_lookup,
9402 typed_splay_tree::splay_tree_predecessor,
9403 typed_splay_tree::splay_tree_successor,
9404 typed_splay_tree::splay_tree_min,
9405 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
9406 (typed_splay_tree::root, typed_splay_tree::comp,
9407 typed_splay_tree::delete_key,
9408 typed_splay_tree::delete_value): New data members.
9409 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
9410 typed_splay_tree::remove.
9412 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
9414 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
9415 -mginv and -mno-ginv to the assembler.
9416 * config/mips/mips.opt (-mcrc): New option.
9418 * doc/invoke.text (-mcrc): Document.
9421 2018-06-15 Nick Clifton <nickc@redhat.com>
9424 * tree.c (escaped_string): New class. Converts an unescaped
9425 string into its escaped equivalent.
9426 (warn_deprecated_use): Use the new class to convert the
9427 deprecation message, if present.
9428 (test_escaped_strings): New self test.
9429 (test_c_tests): Add test_escaped_strings.
9430 * doc/extend.texi (deprecated): Add a note that the
9431 deprecation message is affected by the -fmessage-length
9432 option, and that control characters will be escaped.
9433 (#pragma GCC error): Document this pragma.
9434 (#pragma GCC warning): Likewise.
9435 * doc/invoke.texi (-fmessage-length): Document this option's
9436 effect on the #warning and #error preprocessor directives and
9437 the deprecated attribute.
9439 2018-06-15 Richard Biener <rguenther@suse.de>
9441 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
9442 here, also noting vector size used.
9443 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
9444 size used in MSG_OPTIMIZED_LOCATIONS dump.
9445 (pass_slp_vectorize::execute): Adjust.
9447 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
9450 * config/arc/arc.c (arc_return_address_register): Fix
9453 2018-06-15 Richard Biener <rguenther@suse.de>
9456 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
9457 leave useless conversion stripping to force_gimple_operand_gsi.
9458 (gimplify_build2): Likewise.
9459 (gimplify_build1): Likewise.
9461 2018-06-15 Richard Biener <rguenther@suse.de>
9464 * tree-cfg.c (move_stmt_op): unshare invariant addresses
9465 before adjusting their block.
9467 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
9469 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
9470 multilibs for *-*-rtems*.
9471 * config/riscv/t-rtems: New file.
9473 2018-06-14 Jakub Jelinek <jakub@redhat.com>
9476 * match.pd ((A +- CST1) +- CST2): Punt if last resort
9477 unsigned_type_for returns NULL.
9480 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
9481 subregs of multi-word pseudos unless the float mode has word size.
9483 2018-06-14 Richard Biener <rguenther@suse.de>
9486 * tree-vect-generic.c (build_word_mode_vector_type): Remove
9487 duplicate and harmful type_hash_canon.
9488 * tree.c (type_hash_canon): Assert we didn't find ourselves.
9490 2018-06-14 Richard Biener <rguenther@suse.de>
9493 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
9496 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
9498 * config/rtems.h (STDINT_LONG32): Define.
9500 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
9501 Prachi Godbole <prachi.godbole@imgtec.com>
9503 * config/mips/mips-cpus.def: Define P6600.
9504 * config/mips/mips-tables.opt: Regenerate.
9505 * config/mips/mips.c (mips_ucbranch_type): New enum.
9506 (mips_rtx_cost_data): Add support for P6600.
9507 (mips_issue_rate): Likewise.
9508 (mips_multipass_dfa_lookahead): Likewise.
9509 (mips_avoid_hazard): Likewise.
9510 (mips_reorg_process_insns): Likewise.
9511 (mips_classify_branch_p6600): New function.
9512 * config/mips/mips.h (TUNE_P6600): New define.
9513 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
9514 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
9515 * config/mips/mips.md: Include p6600.md.
9516 (processor): Add p6600.
9517 * config/mips/p6600.md: New file.
9518 * doc/invoke.texi: Add p6600 to supported architectures.
9520 2018-06-13 Martin Sebor <msebor@redhat.com>
9522 PR tree-optimization/86114
9523 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
9525 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
9527 2018-06-13 Richard Biener <rguenther@suse.de>
9529 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
9530 Properly set vector type of the intermediate stmt.
9531 * tree-vect-stmts.c (vectorizable_operation): The destination
9532 var always has vectype_out type.
9534 2018-06-13 Jeff Law <law@redhat.com>
9536 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
9537 integer 0 for argument to print_rtl_with_bb.
9538 (rl78_reorg): Likewise.
9540 2018-06-13 David Malcolm <dmalcolm@redhat.com>
9542 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
9543 from rtx to rtx_insn *.
9544 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
9546 (add_sched_insns_for_speculation): Likewise for local "target",
9547 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
9548 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
9549 from rtx_insn ** to rtx_code_label **.
9550 (reorg_emit_nops): Likewise.
9551 (c6x_reorg): Likewise for local "call_labels".
9552 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
9554 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
9555 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
9556 the loops over LABEL_REFS.
9557 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
9559 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
9560 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
9561 (split_branches): Strengthen local "olabel" from rtx to
9562 rtx_insn *, adding a safe_as_a cast.
9563 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
9565 (add_insn_after): Likewise for first two params.
9566 (add_insn_before): Likewise.
9567 (remove_insn): Likewise for param.
9568 (emit_pattern_before_noloc): Likewise for second and third params.
9569 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
9570 (emit_call_insn_before_noloc): Likewise.
9571 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
9573 (emit_barrier_before): Likewise.
9574 (emit_label_before): Strengthen "label" param from "rtx" to
9575 "rtx_code_label *". Strengthen "before" param from "rtx" to
9577 (emit_insn_after_1): Strengthen "after" param from "rtx" to
9579 (emit_pattern_after_noloc): Likewise.
9580 (emit_insn_after_noloc): Likewise.
9581 (emit_jump_insn_after_noloc): Likewise.
9582 (emit_call_insn_after_noloc): Likewise.
9583 (emit_debug_insn_after_noloc): Likewise.
9584 (emit_barrier_after): Likewise.
9585 (emit_label_after): Likewise for both params.
9586 (emit_pattern_after_setloc): Likewise for "after" param. Convert
9587 "loc" param from "int" to "location_t".
9588 (emit_insn_after_setloc): Likewise.
9589 (emit_jump_insn_after_setloc): Likewise.
9590 (emit_call_insn_after_setloc): Likewise.
9591 (emit_debug_insn_after_setloc): Likewise.
9592 (emit_pattern_before_setloc): Likewise for "before" param. Convert
9593 "loc" param from "int" to "location_t".
9594 (emit_pattern_before): Convert NULL_RTX to NULL.
9595 (emit_insn_before_setloc): Convert "loc" param from "int" to
9597 (emit_jump_insn_before_setloc): Likewise.
9598 (emit_call_insn_before_setloc): Likewise.
9599 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
9600 rtx_insn *. Convert "loc" param from "int" to "location_t".
9601 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
9602 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
9603 Convert 3rd param from "int" to "location_t".
9604 (emit_barrier_before, emit_barrier_after, next_real_insn):
9605 Strengthen param from rtx to rtx_insn *.
9606 (emit_label_before): Strengthen 1st param from "rtx" to
9607 "rtx_code_label *". Strengthen 2nd param from "rtx" to
9609 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
9610 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
9611 Strengthen 2nd param from "rtx" to "rtx_insn *".
9612 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
9613 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
9614 Likewise. Convert 3rd param from "int" to "location_t".
9615 (emit_label_after): Strengthen 1st param from "rtx" to
9617 (next_real_insn, remove_insn): Strengthen param from "rtx" to
9619 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
9620 from "rtx" to "rtx_insn *".
9622 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
9624 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
9625 bodies streamed in with -Q.
9626 * dumpfile.c (dump_files): Add lto-stream-out dump file.
9627 * dumpfile.h (tree_dump_index): Add lto_stream_out.
9628 * gimple-streamer-out.c: Include gimple-pretty-print.h
9629 (output_bb): Dump stmts streamed.
9630 * lto-section-out.c: Include print-tree.h
9631 (lto_begin_section): Dump sections created.
9632 (lto_output_decl_index): Dump decl encoded.
9633 * lto-streamer-out.c: Include print-tree.h
9634 (create_output_block): Dump output block created.
9635 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
9636 (output_function): Dump function output.
9637 (output_constructor): Dump constructor streamed.
9638 (write_global_stream): Output indexes encoded.
9639 (produce_asm_for_decls): Dump streams encoded.
9640 * lto-streamer.c (streamer_dump_file): New global var.
9641 * lto-streamer.h (streamer_dump_file): Declare.
9642 * passes.c (ipa_write_summaries): Initialize streamer dump.
9643 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
9646 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
9649 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
9650 offsets for register save directives. Emit a second batch of save
9651 directives, if need be, when the function accesses prior frames.
9653 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9655 * config/arc/fpu.md (fmasf4): Force operand to register.
9656 (fnmasf4): Likewise.
9658 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9660 * config/arc/arc-protos.h (arc_pad_return): Remove.
9661 * config/arc/arc.c (machine_function): Remove force_short_suffix
9663 (arc_print_operand): Adjust printing of '&'.
9664 (arc_verify_short): Remove conditional printing of short suffix.
9665 (arc_final_prescan_insn): Remove reference to size_reason.
9666 (pad_return): New function.
9667 (arc_reorg): Call pad_return.
9668 (arc_pad_return): Remove.
9669 (arc_init_machine_status): Remove reference to force_short_suffix.
9670 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
9671 (attr length): When attribute iscompact is true force to 2
9672 regardless; in the case of maybe check if we want to force the
9673 instruction to have 4 bytes length.
9674 (nopv): Change it to generate 4 byte long nop as well.
9675 (blockage): New pattern.
9676 (simple_return): Remove call to arc_pad_return.
9677 (p_return_i): Likewise.
9679 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9681 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
9683 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9685 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
9688 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9690 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
9691 for ARC700 and ARCv2.
9693 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
9696 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
9697 operands[2] instead of operands[1].
9700 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
9702 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
9703 case, check whether the outer register overlaps an unallocatable
9704 register, not just whether it fits the required class.
9706 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
9708 * poly-int.h (can_div_trunc_p): Add new overload in which all values
9710 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
9711 (memrefs_conflict_p): Likewise.
9712 (init_alias_analysis): Likewise.
9713 * cfgexpand.c (expand_debug_expr): Likewise.
9714 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
9715 * cse.c (fold_rtx): Likewise.
9716 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
9717 * expr.c (emit_block_move_hints): Likewise.
9718 (clear_storage_hints, push_block, emit_push_insn): Likewise.
9719 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
9720 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
9721 (emit_group_store): Likewise.
9722 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
9723 to read the PRE/POST_MODIFY increment.
9724 * calls.c (store_one_arg): Use strip_offset.
9725 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
9727 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
9729 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
9730 (simplify_binary_operation_1): Extend CONST_INT handling to
9732 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
9733 than a HOST_WIDE_INT.
9734 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
9736 (adjust_mems, add_stores): Update accodingly.
9737 (vt_canonicalize_addr): Track polynomial offsets.
9738 (emit_note_insn_var_location): Likewise.
9739 (vt_add_function_parameter): Likewise.
9740 (vt_initialize): Likewise.
9742 2018-06-12 Jeff Law <law@redhat.com>
9744 * config.gcc (alpha*-*-freebsd*): Remove.
9745 * config/alpha/freebsd.h: Remove.
9747 2018-06-12 David Malcolm <dmalcolm@redhat.com>
9750 * spellcheck-tree.c (levenshtein_distance): Rename to...
9751 (get_edit_distance): ...this, and update for underlying renaming.
9752 * spellcheck-tree.h (levenshtein_distance): Rename to...
9753 (get_edit_distance): ...this.
9754 * spellcheck.c (levenshtein_distance): Rename to...
9755 (get_edit_distance): ...this. Convert from Levenshtein distance
9756 to Damerau-Levenshtein distance by supporting transpositions of
9757 adjacent characters. Rename "v1" to "v_next" and "v0" to
9759 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
9760 (selftest::test_edit_distance_unit_test_oneway): ...this, and
9761 update for underlying renaming.
9762 (selftest::levenshtein_distance_unit_test): Rename to...
9763 (selftest::test_get_edit_distance_unit): ...this, and update for
9764 underlying renaming.
9765 (selftest::test_find_closest_string): Add example from PR 69968
9766 where transposition helps
9767 (selftest::test_metric_conditions): Update for renaming.
9768 (selftest::test_metric_conditions): Likewise.
9769 (selftest::spellcheck_c_tests): Likewise.
9770 * spellcheck.h (levenshtein_distance): Rename both overloads to...
9771 (get_edit_distance): ...this.
9772 (best_match::consider): Update for renaming.
9774 2018-06-12 Martin Sebor <msebor@redhat.com>
9776 PR tree-optimization/85259
9777 * builtins.c (compute_objsize): Handle constant offsets.
9778 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
9779 true iff a warning has been issued.
9780 * gimple.h (gimple_nonartificial_location): New function.
9781 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
9782 gimple_nonartificial_location and handle -Wno-system-headers.
9783 (handle_builtin_stxncpy): Same.
9785 2018-06-12 Martin Sebor <msebor@redhat.com>
9788 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
9790 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
9792 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
9793 BUILTIN_VEC_XST entries for pointer to double and long long.
9795 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
9798 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
9800 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
9803 2018-06-12 Martin Liska <mliska@suse.cz>
9805 * doc/options.texi: Document IntegerRange.
9807 2018-06-12 Martin Liska <mliska@suse.cz>
9809 * config/i386/i386.opt: Make MPX-related options as Deprecated.
9810 * opt-functions.awk: Handle Deprecated flag.
9811 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
9813 (read_cmdline_option): Report warning for a deprecated option.
9814 * opts.h (struct cl_option): Add new field cl_deprecated.
9815 (CL_ERR_DEPRECATED): New.
9817 2018-06-12 Martin Liska <mliska@suse.cz>
9819 * doc/options.texi: Document Deprecated option flag.
9821 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9823 * config/arc/arc-arch.h (arc_extras): New enum.
9824 (arc_cpu_t):Add field extra.
9825 (arc_cpu_types): Consider the extras.
9826 * config/arc/arc-cpus.def: Add extras info.
9827 * config/arc/arc-opts.h (processor_type): Consider extra field.
9828 * config/arc/arc.c (arc_override_options): Handle extra field.
9830 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9832 * config/arc/arc-arch.h: Update ARC_OPTX macro.
9833 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
9835 * config/arc/arc.c (arc_init): Update pic warning.
9836 (irq_range): Update irq range parsing warnings.
9837 (arc_override_options): Update various warning messages.
9838 (arc_handle_aux_attribute): Likewise.
9840 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
9842 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
9844 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9846 * doc/sourcebuild.texi: Document usage of line number 0 in verify
9847 compiler messages directives.
9849 2018-06-12 Matthew Fortune <mfortune@gmail.com>
9851 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
9852 * config/mips/mips-tables.opt: Regenerate.
9853 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
9855 * doc/invoke.texi: Document -march=i6500.
9857 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
9859 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
9860 (i6400_gpmul): Add cpu_unit.
9861 (i6400_gpdiv): Likewise.
9862 (i6400_msa_add_d): Update reservations.
9863 (i6400_msa_int_add) Likewise.
9864 (i6400_msa_short_logic3) Likewise.
9865 (i6400_msa_short_logic2) Likewise.
9866 (i6400_msa_short_logic) Likewise.
9867 (i6400_msa_move) Likewise.
9868 (i6400_msa_cmp) Likewise.
9869 (i6400_msa_short_float2) Likewise.
9870 (i6400_msa_div_d) Likewise.
9871 (i6400_msa_long_logic1) Likewise.
9872 (i6400_msa_long_logic2) Likewise.
9873 (i6400_msa_mult) Likewise.
9874 (i6400_msa_long_float2) Likewise.
9875 (i6400_msa_long_float4) Likewise.
9876 (i6400_msa_long_float5) Likewise.
9877 (i6400_msa_long_float8) Likewise.
9878 (i6400_fpu_fadd): Include frint type.
9879 (i6400_fpu_store): New define_insn_reservation.
9880 (i6400_fpu_load): Likewise.
9881 (i6400_fpu_move): Likewise.
9882 (i6400_fpu_fcmp): Likewise.
9883 (i6400_fpu_fmadd): Likewise.
9884 (i6400_int_mult): Include imul3nc type and update reservation.
9885 (i6400_int_div): Include idiv3 type and update reservation.
9886 (i6400_int_load): Update to check type not move_type.
9887 (i6400_int_store): Likewise.
9888 (i6400_int_prefetch): Set zero latency.
9890 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
9892 * gcc.c: Document new %@{...} sequence.
9893 (LINK_COMMAND_SPEC): Use it for the -L switches.
9894 (cpp_unique_options): Use it for the -I switches.
9895 (at_file_argbuf): New global variable.
9896 (in_at_file): Likewise.
9897 (alloc_args): Create at_file_argbuf.
9898 (clear_args): Truncate at_file_argbuf.
9899 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
9900 (open_at_file): New function.
9901 (close_at_file): Likewise.
9902 (create_at_file): Delete.
9903 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
9906 (validate_switches_from_spec): Deal with %@{...} sequence.
9907 (validate_switches): Likewise.
9908 (driver::finalize): Call clear_args.
9910 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
9912 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
9914 2018-06-11 Martin Sebor <msebor@redhat.com>
9916 * doc/invoke.texi (-Wall): List -Wc++17-compat.
9917 (Wno-class-memaccess): Add @opindex.
9918 (Wno-templates, Wno-multiple-inheritance): Same.
9919 (Wno-virtual-inheritance, Wno-namespaces): Same.
9920 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
9921 (Wno-format-overflow, Wno-format-truncation): Same.
9922 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
9923 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
9924 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
9925 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
9926 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
9927 (Wno-misspelled-isr): Same.
9929 2018-06-11 Martin Sebor <msebor@redhat.com>
9931 * PR tree-optimization/86083
9932 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
9934 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
9936 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
9938 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
9941 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
9942 on the correct operand.
9943 (*movdi_internal64): Ditto.
9945 2018-06-11 Martin Liska <mliska@suse.cz>
9947 PR tree-optimization/86089
9948 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
9950 2018-06-11 Julia Koval <julia.koval@intel.com>
9952 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
9953 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
9954 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
9956 2018-06-11 Olivier Hainque <hainque@adacore.com>
9958 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
9959 for Ada with strict dwarf2.
9961 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
9964 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
9967 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
9969 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
9971 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
9973 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
9976 2018-06-08 Martin Liska <mliska@suse.cz>
9978 * tree-cfg.h (debug_function): Fix argument type to match
9981 2018-06-08 Martin Liska <mliska@suse.cz>
9983 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
9984 Remove usage of MPX-related (and removed) fields.
9985 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
9987 2018-06-08 David Malcolm <dmalcolm@redhat.com>
9989 * cfg.c (debug): Use TDF_NONE rather than 0.
9990 * cfghooks.c (debug): Likewise.
9991 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
9992 (struct dump_option_value_info): Convert to...
9993 (struct kv_pair): ...this template type.
9994 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
9996 (optinfo_verbosity_options): Likewise.
9997 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
9999 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
10000 than int for "optgroup_flags" param.
10001 (dump_generic_expr_loc): Use dump_flags_t rather than int for
10003 (dump_dec): Likewise.
10004 (dump_finish): Use TDF_NONE rather than 0.
10005 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
10006 rather than int for "optgroup_flags" param. Use TDF_NONE rather
10007 than 0. Update for change to option_ptr.
10008 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
10009 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
10010 0. Update for changes to optinfo_verbosity_options and
10012 (opt_info_switch_p): Convert optgroup_flags from int to
10014 (dump_basic_block): Use dump_flags_t rather than int
10015 for "dump_kind" param.
10016 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
10017 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
10018 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
10019 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
10020 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
10021 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
10022 TDF_NONE): Convert from macros to...
10023 (enum dump_flag): ...this new enum.
10024 (dump_flags_t): Update to use enum.
10025 (operator|, operator&, operator~, operator|=, operator&=):
10026 Implement for dump_flags_t.
10027 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
10028 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
10029 Convert from macros to...
10030 (enum optgroup_flag): ...this new enum.
10031 (optgroup_flags_t): New typedef.
10032 (operator|, operator|=): Implement for optgroup_flags_t.
10033 (struct dump_file_info): Convert field "alt_flags" to
10034 dump_flags_t. Convert field "optgroup_flags" to
10036 (dump_basic_block): Use dump_flags_t rather than int for param.
10037 (dump_generic_expr_loc): Likewise.
10038 (dump_dec): Likewise.
10039 (dump_register): Convert param "optgroup_flags" to
10041 (opt_info_enable_passes): Likewise.
10042 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
10044 * gimple-pretty-print.c (debug): Likewise.
10045 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
10046 (merged_store_group::apply_stores): Likewise.
10047 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
10048 * gimple.c (verify_gimple_pp): Likewise.
10049 * graphite-poly.c (print_pbb_body): Likewise.
10050 * passes.c (pass_manager::register_one_dump_file): Convert
10051 local "optgroup_flags" to optgroup_flags_t.
10052 * print-tree.c (print_node): Use TDF_NONE rather than 0.
10054 (debug_body): Likewise.
10055 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
10056 to optgroup_flags_t.
10057 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
10059 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
10060 (convert_mult_to_fma): Likewise.
10061 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
10062 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
10063 * tree-vect-data-refs.c (dump_lower_bound): Convert param
10064 "dump_kind" to dump_flags_t.
10066 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
10068 * config/rs6000/rs6000.c (min, max): Delete.
10070 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
10072 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
10073 -mabi=spe and -mabi=no-spe.
10075 2018-06-08 Martin Liska <mliska@suse.cz>
10077 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
10078 where we expect an existing summary.
10080 2018-06-08 Martin Liska <mliska@suse.cz>
10082 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
10083 * ipa-inline.h (estimate_edge_growth): Likewise.
10085 2018-06-08 Martin Liska <mliska@suse.cz>
10087 * cgraph.c (function_version_hasher::hash): Use
10088 cgraph_node::get_uid ().
10089 (function_version_hasher::equal):
10090 * cgraph.h (cgraph_node::get_uid): New method.
10091 * ipa-inline.c (update_caller_keys): Use
10092 cgraph_node::get_uid ().
10093 (update_callee_keys): Likewise.
10094 * ipa-utils.c (searchc): Likewise.
10095 (ipa_reduced_postorder): Likewise.
10096 * lto-cgraph.c (input_node): Likewise.
10097 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
10098 * symbol-summary.h (symtab_insertion): Likewise.
10099 (symtab_removal): Likewise.
10100 (symtab_duplication): Likewise.
10101 * tree-pretty-print.c (dump_function_header): Likewise.
10102 * tree-sra.c (convert_callers_for_node): Likewise.
10104 2018-06-08 Martin Liska <mliska@suse.cz>
10106 * cgraph.c (symbol_table::create_edge): Always assign a new
10108 (symbol_table::free_edge): Do not recycle numbers.
10109 * cgraph.h (cgraph_edge::get): New method.
10110 * symbol-summary.h (symtab_removal): Use it.
10111 (symtab_duplication): Likewise.
10112 (call_summary::hashable_uid): Remove.
10114 2018-06-08 Martin Liska <mliska@suse.cz>
10116 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
10117 (initialize_growth_caches): Remove.
10118 (free_growth_caches): Likewise.
10119 (do_estimate_edge_time): Use edge_growth_cache.
10120 (do_estimate_edge_size): Likewise.
10121 (do_estimate_edge_hints): Likewise.
10122 * ipa-inline.c (reset_edge_caches): Likewise.
10123 (recursive_inlining): Likewise.
10124 (inline_small_functions): Likewise.
10125 * ipa-inline.h (initialize_growth_caches): Remove.
10126 (estimate_edge_size): Likewise.
10127 (estimate_edge_time): Likewise.
10128 (estimate_edge_hints): Likewise.
10129 (reset_edge_growth_cache): Likewise.
10130 * symbol-summary.h (call_summary::remove): New method.
10132 2018-06-08 Martin Liska <mliska@suse.cz>
10134 * ipa-cp.c (class edge_clone_summary): New summary.
10135 (grow_edge_clone_vectors): Remove.
10136 (ipcp_edge_duplication_hook): Remove.
10137 (class edge_clone_summary_t): New call_summary class.
10138 (ipcp_edge_removal_hook): Remove.
10139 (edge_clone_summary_t::duplicate): New function.
10140 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
10141 (create_specialized_node): Likewise.
10142 (ipcp_driver): Initialize edge_clone_summaries and do not
10145 2018-06-08 Martin Liska <mliska@suse.cz>
10147 * symbol-summary.h (get): New function.
10148 (call_summary::m_initialize_when_cloning): New class member.
10150 2018-06-08 Martin Liska <mliska@suse.cz>
10152 * cgraph.c (cgraph_node::remove): Do not recycle uid.
10153 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
10154 (symbol_table::allocate_cgraph_symbol): Do not set uid.
10155 * passes.c (uid_hash_t): Record removed_nodes by their uids.
10156 (remove_cgraph_node_from_order): Use the removed_nodes set.
10157 (do_per_function_toporder): Likwise.
10158 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
10159 instead of summary_uid.
10160 (symtab_removal): Likewise.
10161 (symtab_duplication): Likewise.
10163 2018-06-08 Martin Liska <mliska@suse.cz>
10165 * ipa-cp.c (ipcp_store_bits_results): Use
10166 ipcp_transformation_sum.
10167 (ipcp_store_vr_results): Likewise.
10168 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
10170 (ipcp_transformation_initialize): ... this.
10171 (ipa_set_node_agg_value_chain):
10172 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
10173 (write_ipcp_transformation_info): Likewise.
10174 (read_ipcp_transformation_info): Likewise.
10175 (ipcp_update_bits): Likewise.
10176 (ipcp_update_vr): Likewise.
10177 (ipcp_transform_function): Likewise.
10178 * ipa-prop.h: Rename ipcp_transformation_summary to
10179 ipcp_transformation.
10180 (class ipcp_transformation_t): New function summary.
10181 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
10182 (ipa_get_agg_replacements_for_node): Likewise.
10184 2018-06-08 Martin Liska <mliska@suse.cz>
10186 * ipa-pure-const.c (struct funct_state_d): Do it class instead
10188 (class funct_state_summary_t): New function_summary class.
10189 (has_function_state): Remove.
10190 (get_function_state): Likewise.
10191 (set_function_state): Likewise.
10192 (add_new_function): Likewise.
10193 (funct_state_summary_t::insert): New function.
10194 (duplicate_node_data): Remove.
10195 (remove_node_data): Remove.
10196 (funct_state_summary_t::duplicate): New function.
10197 (register_hooks): Create new funct_state_summaries.
10198 (pure_const_generate_summary): Use it.
10199 (pure_const_write_summary): Likewise.
10200 (pure_const_read_summary): Likewise.
10201 (propagate_pure_const): Likewise.
10202 (propagate_nothrow): Likewise.
10203 (dump_malloc_lattice): Likewise.
10204 (propagate_malloc): Likewise.
10205 (execute): Do not register hooks, just remove summary
10207 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
10210 2018-06-08 Martin Liska <mliska@suse.cz>
10212 * ipa-reference.c (remove_node_data): Remove.
10213 (duplicate_node_data): Likewise.
10214 (class ipa_ref_var_info_summary_t): New class.
10215 (class ipa_ref_opt_summary_t): Likewise.
10216 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
10217 (get_reference_optimization_summary): Use
10218 ipa_ref_opt_sum_summaries.
10219 (set_reference_vars_info): Remove.
10220 (set_reference_optimization_summary): Likewise.
10221 (ipa_init): Create summaries.
10222 (init_function_info): Use function summary.
10223 (ipa_ref_opt_summary_t::duplicate): New function.
10224 (ipa_ref_opt_summary_t::remove): New function.
10225 (get_read_write_all_from_node): Fix GNU coding style.
10226 (propagate): Use function summary.
10227 (write_node_summary_p): Fix GNU coding style.
10228 (stream_out_bitmap): Likewise.
10229 (ipa_reference_read_optimization_summary): Use function summary.
10230 (ipa_reference_c_finalize): Do not release hooks.
10232 2018-06-08 Martin Liska <mliska@suse.cz>
10234 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
10235 (analyze_function_body): Extract multiple calls of get_create.
10236 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
10237 * ipa-inline.c (recursive_inlining): Use ::get method.
10238 * ipa-inline.h (estimate_edge_growth): Likewise.
10240 2018-06-08 Martin Liska <mliska@suse.cz>
10242 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
10244 (hsa_function_summary::hsa_function_summary): Use the new enum
10246 (hsa_gpu_implementation_p): Use hsa_summaries::get.
10247 * hsa-gen.c (hsa_get_host_function): Likewise.
10248 (get_brig_function_name): Likewise.
10249 * ipa-hsa.c (process_hsa_functions): Likewise.
10250 (ipa_hsa_write_summary): Likewise.
10251 * symbol-summary.h (symtab_duplication): Use ::get function/
10252 (get): New function.
10254 2018-06-08 Martin Liska <mliska@suse.cz>
10256 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
10258 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
10259 (hsa_register_kernel): Likewise.
10260 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
10261 * hsa-gen.c (hsa_get_host_function): Likewise.
10262 (get_brig_function_name): Likewise.
10263 (generate_hsa): Likewise.
10264 (pass_gen_hsail::execute): Likewise.
10265 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
10266 (devirtualization_time_bonus): Likewise.
10267 (ipcp_propagate_stage): Likewise.
10268 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
10269 (edge_set_predicate): Likewise.
10270 (evaluate_conditions_for_known_args): Likewise.
10271 (evaluate_properties_for_edge): Likewise.
10272 (ipa_fn_summary::reset): Likewise.
10273 (ipa_fn_summary_t::duplicate): Likewise.
10274 (dump_ipa_call_summary): Likewise.
10275 (ipa_dump_fn_summary): Likewise.
10276 (analyze_function_body): Likewise.
10277 (compute_fn_summary): Likewise.
10278 (estimate_edge_devirt_benefit): Likewise.
10279 (estimate_edge_size_and_time): Likewise.
10280 (estimate_calls_size_and_time): Likewise.
10281 (estimate_node_size_and_time): Likewise.
10282 (inline_update_callee_summaries): Likewise.
10283 (remap_edge_change_prob): Likewise.
10284 (remap_edge_summaries): Likewise.
10285 (ipa_merge_fn_summary_after_inlining): Likewise.
10286 (ipa_update_overall_fn_summary): Likewise.
10287 (read_ipa_call_summary): Likewise.
10288 (inline_read_section): Likewise.
10289 (write_ipa_call_summary): Likewise.
10290 (ipa_fn_summary_write): Likewise.
10291 (ipa_free_fn_summary): Likewise.
10292 * ipa-hsa.c (process_hsa_functions): Likewise.
10293 (ipa_hsa_write_summary): Likewise.
10294 (ipa_hsa_read_section): Likewise.
10295 * ipa-icf.c (sem_function::merge): Likewise.
10296 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
10297 (do_estimate_edge_time): Likewise.
10298 (estimate_size_after_inlining): Likewise.
10299 (estimate_growth): Likewise.
10300 (growth_likely_positive): Likewise.
10301 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
10302 (inline_call): Likewise.
10303 * ipa-inline.c (caller_growth_limits): Likewise.
10304 (can_inline_edge_p): Likewise.
10305 (can_inline_edge_by_limits_p): Likewise.
10306 (compute_uninlined_call_time): Likewise.
10307 (compute_inlined_call_time): Likewise.
10308 (want_inline_small_function_p): Likewise.
10309 (edge_badness): Likewise.
10310 (update_caller_keys): Likewise.
10311 (update_callee_keys): Likewise.
10312 (recursive_inlining): Likewise.
10313 (inline_small_functions): Likewise.
10314 (inline_to_all_callers_1): Likewise.
10315 (dump_overall_stats): Likewise.
10316 (early_inline_small_functions): Likewise.
10317 (early_inliner): Likewise.
10318 * ipa-inline.h (estimate_edge_growth): Likewise.
10319 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
10320 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
10321 * ipa-prop.h (IPA_NODE_REF): Likewise.
10322 (IPA_EDGE_REF): Likewise.
10323 * ipa-pure-const.c (malloc_candidate_p): Likewise.
10324 (propagate_malloc): Likewise.
10325 * ipa-split.c (execute_split_functions): Likewise.
10326 * symbol-summary.h: Rename get to get_create.
10328 (get_create): Likewise.
10329 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
10331 2018-06-08 Martin Liska <mliska@suse.cz>
10333 * symbol-summary.h (release): Move definition out of class
10335 (symtab_removal): Likewise.
10336 (symtab_duplication): Likewise.
10338 2018-06-08 Martin Liska <mliska@suse.cz>
10340 * symbol-summary.h (function_summary): Move constructor
10341 implementation out of class declaration.
10342 (release): Likewise.
10343 (symtab_insertion): Likewise.
10344 (symtab_removal): Likewise.
10345 (symtab_duplication): Likewise.
10348 2018-06-08 Martin Liska <mliska@suse.cz>
10350 * Makefile.in: Remove support for MPX (macros, related functions,
10351 fields in cgraph_node, ...).
10352 * builtin-types.def (BT_BND): Likewise.
10353 (BT_FN_BND_CONST_PTR): Likewise.
10354 (BT_FN_CONST_PTR_BND): Likewise.
10355 (BT_FN_VOID_PTR_BND): Likewise.
10356 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
10357 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
10358 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
10359 (expand_builtin_mempcpy_with_bounds): Likewise.
10360 (expand_builtin_memset_with_bounds): Likewise.
10361 (expand_builtin_memset_args): Likewise.
10362 (std_expand_builtin_va_start): Likewise.
10363 (expand_builtin): Likewise.
10364 (expand_builtin_with_bounds): Likewise.
10365 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
10366 (DEF_LIB_BUILTIN_CHKP): Likewise.
10367 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
10368 (DEF_CHKP_BUILTIN): Likewise.
10369 (BUILT_IN_MEMCPY): Likewise.
10370 (BUILT_IN_MEMMOVE): Likewise.
10371 (BUILT_IN_MEMPCPY): Likewise.
10372 (BUILT_IN_MEMSET): Likewise.
10373 (BUILT_IN_STPCPY): Likewise.
10374 (BUILT_IN_STRCAT): Likewise.
10375 (BUILT_IN_STRCHR): Likewise.
10376 (BUILT_IN_STRCPY): Likewise.
10377 (BUILT_IN_STRLEN): Likewise.
10378 (BUILT_IN_MEMCPY_CHK): Likewise.
10379 (BUILT_IN_MEMMOVE_CHK): Likewise.
10380 (BUILT_IN_MEMPCPY_CHK): Likewise.
10381 (BUILT_IN_MEMSET_CHK): Likewise.
10382 (BUILT_IN_STPCPY_CHK): Likewise.
10383 (BUILT_IN_STRCAT_CHK): Likewise.
10384 (BUILT_IN_STRCPY_CHK): Likewise.
10385 * calls.c (store_bounds): Likewise.
10386 (emit_call_1): Likewise.
10387 (special_function_p): Likewise.
10388 (maybe_warn_nonstring_arg): Likewise.
10389 (initialize_argument_information): Likewise.
10390 (finalize_must_preallocate): Likewise.
10391 (compute_argument_addresses): Likewise.
10392 (expand_call): Likewise.
10393 * cfgexpand.c (expand_call_stmt): Likewise.
10394 (expand_return): Likewise.
10395 (expand_gimple_stmt_1): Likewise.
10396 (pass_expand::execute): Likewise.
10397 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
10398 (cgraph_node::remove): Likewise.
10399 (cgraph_node::dump): Likewise.
10400 (cgraph_node::verify_node): Likewise.
10401 * cgraph.h (chkp_function_instrumented_p): Likewise.
10402 (symtab_node::get_alias_target): Likewise.
10403 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
10404 (cgraph_local_p): Likewise.
10405 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
10406 (cgraph_edge::rebuild_references): Likewise.
10407 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
10408 (walk_polymorphic_call_targets): Likewise.
10409 (cgraph_node::expand_thunk): Likewise.
10410 (symbol_table::output_weakrefs): Likewise.
10411 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
10412 (ix86_handle_option): Likewise.
10413 * config/i386/constraints.md: Likewise.
10414 * config/i386/i386-builtin-types.def (BND): Likewise.
10418 * config/i386/i386-builtin.def (BDESC_END): Likewise.
10419 (BDESC_FIRST): Likewise.
10421 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
10422 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
10423 * config/i386/i386.c (enum reg_class): Likewise.
10424 (ix86_target_string): Likewise.
10425 (ix86_option_override_internal): Likewise.
10426 (ix86_conditional_register_usage): Likewise.
10427 (ix86_valid_target_attribute_inner_p): Likewise.
10428 (ix86_set_indirect_branch_type): Likewise.
10429 (ix86_set_current_function): Likewise.
10430 (ix86_function_arg_regno_p): Likewise.
10431 (init_cumulative_args): Likewise.
10432 (ix86_function_arg_advance): Likewise.
10433 (ix86_function_arg): Likewise.
10434 (ix86_pass_by_reference): Likewise.
10435 (ix86_function_value_regno_p): Likewise.
10436 (ix86_function_value_1): Likewise.
10437 (ix86_function_value_bounds): Likewise.
10438 (ix86_return_in_memory): Likewise.
10439 (ix86_setup_incoming_vararg_bounds): Likewise.
10440 (ix86_va_start): Likewise.
10441 (indirect_thunk_need_prefix): Likewise.
10442 (print_reg): Likewise.
10443 (ix86_print_operand): Likewise.
10444 (ix86_expand_call): Likewise.
10445 (ix86_output_function_return): Likewise.
10446 (reg_encoded_number): Likewise.
10447 (BDESC_VERIFYS): Likewise.
10448 (ix86_init_mpx_builtins): Likewise.
10449 (ix86_init_builtins): Likewise.
10450 (ix86_emit_cmove): Likewise.
10451 (ix86_emit_move_max): Likewise.
10452 (ix86_expand_builtin): Likewise.
10453 (ix86_builtin_mpx_function): Likewise.
10454 (ix86_get_arg_address_for_bt): Likewise.
10455 (ix86_load_bounds): Likewise.
10456 (ix86_store_bounds): Likewise.
10457 (ix86_load_returned_bounds): Likewise.
10458 (ix86_store_returned_bounds): Likewise.
10459 (ix86_class_likely_spilled_p): Likewise.
10460 (ix86_hard_regno_mode_ok): Likewise.
10461 (x86_order_regs_for_local_alloc): Likewise.
10462 (ix86_mitigate_rop): Likewise.
10463 (ix86_bnd_prefixed_insn_p): Likewise.
10464 (ix86_mpx_bound_mode): Likewise.
10465 (ix86_make_bounds_constant): Likewise.
10466 (ix86_initialize_bounds): Likewise.
10467 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
10468 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
10469 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
10470 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
10471 (TARGET_CHKP_BOUND_MODE): Likewise.
10472 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
10473 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
10474 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
10475 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
10476 * config/i386/i386.h (TARGET_MPX): Likewise.
10477 (TARGET_MPX_P): Likewise.
10478 (VALID_BND_REG_MODE): Likewise.
10479 (FIRST_BND_REG): Likewise.
10480 (LAST_BND_REG): Likewise.
10481 (enum reg_class): Likewise.
10482 (BND_REG_P): Likewise.
10483 (BND_REGNO_P): Likewise.
10484 (BNDmode): Likewise.
10485 (ADJUST_INSN_LENGTH): Likewise.
10486 * config/i386/i386.md: Likewise.
10487 * config/i386/i386.opt: Likewise.
10488 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
10489 (defined): Likewise.
10490 (LINK_MPX): Likewise.
10491 (MPX_SPEC): Likewise.
10492 (LIBMPX_SPEC): Likewise.
10493 (LIBMPXWRAPPERS_SPEC): Likewise.
10494 (CHKP_SPEC): Likewise.
10495 * config/i386/predicates.md: Likewise.
10496 * dbxout.c (dbxout_type): Likewise.
10497 * doc/extend.texi: Likewise.
10498 * doc/invoke.texi: Likewise.
10499 * doc/md.texi: Likewise.
10500 * doc/tm.texi: Likewise.
10501 * doc/tm.texi.in: Likewise.
10502 * dwarf2out.c (is_base_type): Likewise.
10503 (gen_formal_types_die): Likewise.
10504 (gen_subprogram_die): Likewise.
10505 (gen_type_die_with_usage): Likewise.
10506 (gen_decl_die): Likewise.
10507 (dwarf2out_late_global_decl): Likewise.
10508 * expr.c (expand_assignment): Likewise.
10509 (emit_storent_insn): Likewise.
10510 (store_expr_with_bounds): Likewise.
10511 (store_expr): Likewise.
10512 (expand_expr_real_1): Likewise.
10513 * expr.h (store_expr_with_bounds): Likewise.
10514 * function.c (use_register_for_decl): Likewise.
10515 (struct bounds_parm_data): Likewise.
10516 (assign_parms_augmented_arg_list): Likewise.
10517 (assign_parm_find_entry_rtl): Likewise.
10518 (assign_parm_is_stack_parm): Likewise.
10519 (assign_parm_load_bounds): Likewise.
10520 (assign_bounds): Likewise.
10521 (assign_parms): Likewise.
10522 (expand_function_start): Likewise.
10523 * gcc.c (CHKP_SPEC): Likewise.
10524 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
10525 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
10526 (wrestrict_dom_walker::check_call): Likewise.
10527 * gimple.c (gimple_build_call_from_tree): Likewise.
10528 * gimple.h (enum gf_mask): Likewise.
10529 (gimple_call_with_bounds_p): Likewise.
10530 (gimple_call_set_with_bounds): Likewise.
10531 * gimplify.c (gimplify_init_constructor): Likewise.
10532 * ipa-cp.c (initialize_node_lattices): Likewise.
10533 (propagate_constants_across_call): Likewise.
10534 (find_more_scalar_values_for_callers_subset): Likewise.
10535 * ipa-hsa.c (process_hsa_functions): Likewise.
10536 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
10537 * ipa-icf.c (sem_function::merge): Likewise.
10538 * ipa-inline.c (early_inliner): Likewise.
10539 * ipa-pure-const.c (warn_function_noreturn): Likewise.
10540 (warn_function_cold): Likewise.
10541 (propagate_pure_const): Likewise.
10542 * ipa-ref.h (enum GTY): Likewise.
10543 * ipa-split.c (find_retbnd): Likewise.
10544 (consider_split): Likewise.
10545 (split_function): Likewise.
10546 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
10547 * ipa.c (walk_polymorphic_call_targets): Likewise.
10548 (symbol_table::remove_unreachable_nodes): Likewise.
10549 (process_references): Likewise.
10550 (cgraph_build_static_cdtor_1): Likewise.
10551 * lto-cgraph.c (lto_output_node): Likewise.
10552 (output_refs): Likewise.
10553 (compute_ltrans_boundary): Likewise.
10554 (input_overwrite_node): Likewise.
10555 (input_node): Likewise.
10556 (input_cgraph_1): Likewise.
10557 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
10558 * passes.c (pass_manager::execute_early_local_passes): Likewise.
10559 (class pass_chkp_instrumentation_passes): Likewise.
10560 (make_pass_chkp_instrumentation_passes): Likewise.
10561 * passes.def: Likewise.
10562 * rtl.h (struct GTY): Likewise.
10563 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
10564 * stor-layout.c (layout_type): Likewise.
10565 * symtab.c: Likewise.
10566 * target.def: Likewise.
10567 * targhooks.c (default_chkp_bound_type): Likewise.
10568 (default_chkp_bound_mode): Likewise.
10569 (default_builtin_chkp_function): Likewise.
10570 (default_chkp_function_value_bounds): Likewise.
10571 (default_chkp_make_bounds_constant): Likewise.
10572 (default_chkp_initialize_bounds): Likewise.
10573 * targhooks.h (default_chkp_bound_type): Likewise.
10574 (default_chkp_bound_mode): Likewise.
10575 (default_builtin_chkp_function): Likewise.
10576 (default_chkp_function_value_bounds): Likewise.
10577 (default_chkp_make_bounds_constant): Likewise.
10578 (default_chkp_initialize_bounds): Likewise.
10579 * toplev.c (compile_file): Likewise.
10580 (process_options): Likewise.
10581 * tree-core.h (DEF_BUILTIN): Likewise.
10582 (DEF_BUILTIN_CHKP): Likewise.
10583 * tree-inline.c (declare_return_variable): Likewise.
10584 (remap_gimple_stmt): Likewise.
10585 (copy_bb): Likewise.
10586 (initialize_inlined_parameters): Likewise.
10587 (expand_call_inline): Likewise.
10588 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
10589 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
10590 (make_pass_ipa_chkp_produce_thunks): Likewise.
10591 (make_pass_chkp): Likewise.
10592 (make_pass_chkp_opt): Likewise.
10593 (make_pass_chkp_instrumentation_passes): Likewise.
10594 * tree-pretty-print.c (dump_generic_node): Likewise.
10595 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
10596 * tree-ssa-dce.c (propagate_necessity): Likewise.
10597 (eliminate_unnecessary_stmts): Likewise.
10598 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
10599 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
10600 * tree-ssa-sccvn.h: Likewise.
10601 * tree-ssa-strlen.c (get_string_length): Likewise.
10602 (valid_builtin_call): Likewise.
10603 (adjust_last_stmt): Likewise.
10604 (handle_builtin_strchr): Likewise.
10605 (handle_builtin_strcpy): Likewise.
10606 (handle_builtin_stxncpy): Likewise.
10607 (handle_builtin_memcpy): Likewise.
10608 (handle_builtin_strcat): Likewise.
10609 (strlen_check_and_optimize_stmt): Likewise.
10610 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
10611 * tree-streamer-in.c: Likewise.
10612 * tree-streamer.c (record_common_node): Likewise.
10613 * tree.c (tree_code_size): Likewise.
10614 (wide_int_to_tree_1): Likewise.
10615 (type_contains_placeholder_1): Likewise.
10616 (build_common_tree_nodes): Likewise.
10617 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
10618 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
10619 (POINTER_BOUNDS_P): Likewise.
10620 (BOUNDED_TYPE_P): Likewise.
10621 (BOUNDED_P): Likewise.
10622 (CALL_WITH_BOUNDS_P): Likewise.
10623 (pointer_bounds_type_node): Likewise.
10624 * value-prof.c (gimple_ic): Likewise.
10625 * var-tracking.c (vt_add_function_parameters): Likewise.
10626 * varasm.c (make_decl_rtl): Likewise.
10627 (assemble_start_function): Likewise.
10628 (output_constant): Likewise.
10629 (maybe_assemble_visibility): Likewise.
10630 * varpool.c (ctor_for_folding): Likewise.
10631 * chkp-builtins.def: Remove.
10632 * ipa-chkp.c: Remove.
10633 * ipa-chkp.h: Remove.
10634 * rtl-chkp.c: Remove.
10635 * rtl-chkp.h: Remove.
10636 * tree-chkp-opt.c: Remove.
10637 * tree-chkp.c: Remove.
10638 * tree-chkp.h: Remove.
10640 2018-06-07 Carl Love <cel@us.ibm.com>
10642 * config/rs6000/vsx.md (vextract_fp_from_shorth,
10643 vextract_fp_from_shortl): Add BE support.
10645 2018-06-07 Paul Koning <ni1d@arrl.net>
10647 * compare-elim.c (try_merge_compare): Don't merge compare if
10648 address contains a side effect.
10649 (try_eliminate_compare): Likewise.
10651 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
10653 * config.gcc: Support "tremont".
10654 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
10655 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10657 * config/i386/i386.c (m_TREMONT): Define.
10658 (processor_target_table): Add "tremont".
10659 (PTA_TREMONT): Define.
10660 (ix86_lea_outperforms): Add TARGET_TREMONT.
10661 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
10662 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
10663 and M_INTEL_GOLDMONT_PLUS.
10664 (fold_builtin_cpu): Add "tremont".
10665 (ix86_add_stmt_cost): Add TARGET_TREMONT.
10666 (ix86_option_override_internal): Add "tremont".
10667 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
10668 (processor_type): Add PROCESSOR_TREMONT.
10669 * config/i386/x86-tune.def: Add m_TREMONT.
10670 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
10672 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10674 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
10675 symbol defined for msp430i* devices to be lower case.
10677 2018-06-07 Richard Biener <rguenther@suse.de>
10679 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
10680 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
10681 Properly wrap signed arithmetic if overflow wraps.
10683 2018-06-07 Jakub Jelinek <jakub@redhat.com>
10685 PR tree-optimization/69615
10686 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
10687 of a cast from a same precision integral SSA_NAME in a bb dominated
10688 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
10689 cast to utype if rhs2 has already a compatible type.
10691 2018-06-07 Richard Biener <rguenther@suse.de>
10693 PR tree-optimization/85935
10694 * graphite-scop-detection.c (find_params_in_bb): Analyze
10695 condition operands with respect to the correct loop. Assert
10696 the analysis doesn't fail.
10698 2018-06-04 Carl Love <cel@us.ibm.com>
10700 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
10701 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
10702 as it is slightly cheaper.
10703 (first_match_or_eos_index_<mode>):
10704 Calculate index using natural element order.
10705 (first_match_index_<mode>):
10706 Calculate index using natural element order.
10707 (first_match_or_eos_index_<mode>):
10708 Calculate index using natural order.
10709 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
10710 for BE and LE modes.
10711 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
10712 P9V_BUILTIN_VCLZLSBB_V16QI.
10713 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
10716 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
10718 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
10719 indentation and line wrap for many prototypes. Add missing
10720 @smallexample directives around block of prototypes for vec_xl and
10723 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
10725 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
10726 track if we pass or return IEEE 128-bit floating point.
10727 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
10728 C++ mangling that is compatible with GCC 8.1.
10729 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
10730 (init_cumulative_args): Note if we pass or return IEEE 128-bit
10731 floating point types.
10732 (rs6000_function_arg_advance_1): Likewise.
10733 (rs6000_mangle_type): Optionally generate mangled names that match
10734 what GCC 8.1 generated for IEEE 128-bit floating point types.
10735 (rs6000_globalize_decl_name): If we have an external function that
10736 passes or returns IEEE 128-bit types, generate a weak reference
10737 from the mangled name used in GCC 8.1 to the current mangled
10739 (rs6000_init_builtins): Make __ibm128 use the long double type if
10740 long double is IBM extended double. Make __float128 use the long
10741 double type if long double is IEEE 128-bit.
10744 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
10745 macro for __ibm128 built-in functions.
10746 (PACK_IF): Add __ibm128 pack/unpack functions.
10747 (UNPACK_IF): Likewise.
10748 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
10749 enable long double built-in functions if long double is IEEE
10750 128-bit floating point.
10751 (rs6000_invalid_builtin): Update long double built-in function
10753 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
10754 functions, adjust the built-in function to use the long double
10755 built-in function if __ibm128 and long double are the same type.
10756 * doc/extend.texi (PowerPC builtins): Update documention for
10757 __builtin_{,un}pack_longdouble. Add documentation for
10758 __builtin_{,un}pack_ibm128.
10760 2018-06-06 Jim Wilson <jimw@sifive.com>
10762 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
10763 (struct machine_function): New field interrupt_mode.
10764 (riscv_handle_type_attribute): New function. Add forward declaration.
10765 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
10766 (riscv_expand_epilogue): Check interrupt_mode field.
10767 (riscv_set_current_function): Check interrupt attribute args and
10768 set interrupt_mode field.
10769 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
10770 (riscv_sret, riscv_uret): New.
10771 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
10772 new arguments to interrupt attribute.
10774 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
10777 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
10778 Don't handle -mcpu=power8 if -mpower9-vector is also used.
10780 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
10782 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
10783 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
10784 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
10785 several redundant entries.
10787 2018-06-06 David Malcolm <dmalcolm@redhat.com>
10789 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
10790 type from "rtx" to "rtx_insn *".
10791 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
10792 for local "call_insn", removing cast.
10793 (ix86_expand_call): Likewise, introducing a "call_insn" local.
10795 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
10797 PR tree-optimization/86066
10798 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
10799 for BIT_INSERT_EXPR stores.
10801 2018-06-06 Richard Biener <rguenther@suse.de>
10803 PR tree-optimization/86062
10804 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
10805 component refs ontop
10806 of to be offsetted base.
10808 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10810 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
10811 to be static and remove check on interrupt attribute name.
10813 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
10815 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
10816 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
10818 2018-06-05 Steve Ellcey <sellcey@cavium.com>
10821 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
10823 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
10824 Remove second argument, change how error is called.
10825 (aarch64_layout_arg): Remove second argument from
10826 aarch64_err_no_fpadvsimd call.
10827 (aarch64_init_cumulative_args): Ditto.
10828 (aarch64_gimplify_va_arg_expr): Ditto.
10829 * config/aarch64/aarch64.md (mov<mode>): Ditto.
10831 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
10833 * config/i386/i386.md (simple_return_indirect_internal): New expander.
10834 (*simple_return_indirect_internal<mode>): Rename from
10835 simple_return_indirect_internal. Use W mode iterator.
10836 (rstorssp): New expander.
10837 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
10838 (clrssbsy): New expander.
10839 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
10841 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
10843 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
10845 (cmse_check_pointed_object): Likewise.
10847 2018-06-05 Martin Liska <mliska@suse.cz>
10849 PR gcov-profile/47618
10850 * doc/invoke.texi: Document how -fprofile-dir format
10853 2018-06-05 Richard Biener <rguenther@suse.de>
10855 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
10856 removal pretend DOM info isn't available so we do not update
10857 it and only remove edges, not dominated blocks. Actually free
10858 DOM info in case we removed something. Remove unreachable blocks.
10859 (mfb_keep_latches): Work with either DOM info or marked backedges.
10860 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
10861 first. Mark backedges if DOM info isn't available.
10862 (Re-)compute DOM info after cleanup_control_flow_pre.
10864 2018-06-05 Richard Biener <rguenther@suse.de>
10866 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
10867 (locus_discrim_hasher::hash): Adjust.
10868 (locus_discrim_hasher::equal): Likewise.
10869 (next_discriminator_for_locus): Work on line directly.
10870 (same_line_p): Pass in expanded locus1 as well.
10871 (assign_discriminators): Avoid redundant location expansions.
10873 2018-06-05 Richard Biener <rguenther@suse.de>
10875 PR tree-optimization/86046
10876 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
10877 if required after clearing TREE_ADDRESSABLE.
10879 2018-06-05 Richard Biener <rguenther@suse.de>
10881 PR tree-optimization/86047
10882 * tree-ssa-loop.c (for_each_index): Glob handling of all
10883 decls and constants and really handle all of them.
10885 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10888 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
10889 qualifier_void_pointer and qualifier_const_void_pointer.
10890 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
10891 (arm_init_builtins): Handle the above.
10892 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
10893 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
10894 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
10897 2018-06-05 Martin Liska <mliska@suse.cz>
10899 * auto-profile.c (read_autofdo_file): Do not use
10900 gcov_ctr_summary struct.
10901 (afdo_callsite_hot_enough_for_early_inline): Likewise.
10902 * coverage.c (struct counts_entry): Likewise.
10903 (read_counts_file): Read just single summary entry.
10904 (get_coverage_counts): Use gcov_summary struct.
10905 * coverage.h (get_coverage_counts): Likewise.
10906 * gcov-dump.c (dump_working_sets): Likewise.
10907 (tag_summary): Dump just single summary.
10908 * gcov-io.c (gcov_write_summary): Write just histogram
10910 (gcov_read_summary): Read just single summary.
10911 (compute_working_sets): Use gcov_summary struct.
10912 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
10913 of GCOV_COUNTERS_SUMMABLE.
10914 (GCOV_COUNTERS_SUMMABLE): Remove.
10915 (GCOV_FIRST_VALUE_COUNTER): Replace with
10916 GCOV_COUNTER_V_INTERVAL.
10917 (struct gcov_ctr_summary): Remove.
10918 (struct gcov_summary): Directly use fields of former
10920 (compute_working_sets): Use gcov_summary struct.
10921 * gcov.c (read_count_file): Do not use ctrs fields.
10922 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
10924 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
10926 * profile.c: Likewise.
10927 * profile.h: Likewise.
10929 2018-06-05 Martin Liska <mliska@suse.cz>
10931 PR gcov-profile/84846
10932 * gcov.c (output_lines): Print working directory only
10933 in intermediate format.
10935 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
10937 * config/s390/s390-builtin-types.def: Add void function type.
10938 * config/s390/s390-builtins.def: Use the function type for the
10941 2018-06-04 Jim Wilson <jimw@sifive.com>
10943 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
10945 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
10946 and maybe_eh_return. Change regno to unsigned int. Use new args to
10947 handle EH_RETURN_DATA_REGNO registers properly.
10948 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
10949 (riscv_expand_epilogue): Update comment. Change argument name and
10950 type. Update code to use new name and type. Pass new args to
10951 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
10953 * config/riscv/riscv.md (NORMAL_RETURN): New.
10954 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
10955 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
10956 (eh_return): Call gen_eh_return_internal and emit barrier.
10957 (eh_return_internal): Call riscv_expand_epilogue.
10959 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
10961 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
10962 bit_insertion field and declare can_be_merged_into method.
10963 (merged_store_group::can_be_merged_into): New method.
10964 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
10965 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
10966 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
10968 2018-06-04 Richard Biener <rguenther@suse.de>
10970 PR tree-optimization/85955
10971 * builtins.c (fold_builtin_sincos): Convert pointers to
10972 destination to appropriate type before dereferencing.
10974 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
10976 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
10978 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
10980 * expr.c (expand_expr_real_1): Force the operand into memory if
10981 its TYPE_MODE is BLKmode and if there is no integer mode for
10982 the number of bits being extracted.
10984 2018-06-04 Jakub Jelinek <jakub@redhat.com>
10988 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
10989 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
10991 2018-06-04 Richard Biener <rguenther@suse.de>
10993 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
10994 (cleanup_tree_cfg_noloop): ... single caller. Do
10995 start_recording_case_labels later.
10997 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
10999 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
11000 to _IMMINTRIN_H_INCLUDED.
11001 * config/i386/pconfigintrin.h: Ditto.
11002 * config/i386/waitpkgintrin.h: Ditto.
11003 * config/i386/immintrin.h: Add includes for sgxintrin.h,
11004 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
11005 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
11006 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
11007 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
11008 waitpkgintrin.h and cldemoteintrin.h.
11010 2018-06-04 Richard Biener <rguenther@suse.de>
11012 PR tree-optimization/86038
11013 * tracer.c (find_best_successor): Check probability for
11014 being initialized, bail out if not.
11016 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
11019 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
11020 of bits to ignore when comparing architectures.
11022 2018-06-04 Jakub Jelinek <jakub@redhat.com>
11024 PR tree-optimization/69615
11025 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
11026 maximum or minimum of the type, try to merge it also as if
11027 range1 is + [-, x - 1] or + [x + 1, -].
11030 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
11032 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
11034 PR tree-optimization/86034
11035 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
11036 the unsigned bitfield type in a bit insertion sequence if it does not
11037 have a larger precision than the bitfield size.
11038 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
11040 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
11042 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
11044 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
11046 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
11047 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
11048 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
11049 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
11051 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
11053 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
11054 Disable -fdelete-null-pointer-checks for ELF toolchain.
11056 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
11057 Kito Cheng <kito.cheng@gmail.com>
11059 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
11060 (nds32le-*-*, nds32be-*-*): Integrate checking process.
11061 (nds32*-*-*): Add glibc and uclibc conditions.
11062 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
11063 (TARGET_EXCEPT_UNWIND_INFO): Define.
11064 * config/nds32/elf.h: New file.
11065 * config/nds32/linux.h: New file.
11066 * config/nds32/nds32-elf.opt: New file.
11067 * config/nds32/nds32-linux.opt: New file.
11068 * config/nds32/nds32-fp-as-gp.c
11069 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
11070 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
11072 (nds32_asm_file_end): Ditto.
11073 (nds32_print_operand): Ditto.
11074 (nds32_insert_attributes): Ditto.
11075 (nds32_init_libfuncs): New function.
11076 (TARGET_HAVE_TLS): Define.
11077 (TARGET_INIT_LIBFUNCS): Define.
11078 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
11080 (TARGET_ELF): Apply different mcmodel setting.
11081 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
11082 been migrated into elf.h and linux.h files.
11083 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
11084 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
11085 (mcmodel): The content has been migrated into nds32-elf.opt and
11086 nds32-linux.opt files.
11087 * config/nds32/t-elf: New file.
11088 * config/nds32/t-linux: New file.
11090 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
11091 Shiva Chen <shiva0217@gmail.com>
11093 * config/nds32/constants.md (unspec_volatile_element): Add
11094 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
11095 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
11097 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
11098 (make_pass_nds32_fp_as_gp): Declare.
11099 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
11101 (nds32_asm_function_end_prologue): Remove unused asm output.
11102 (nds32_asm_function_begin_epilogue): Remove unused asm output.
11103 (nds32_asm_file_start): Output necessary fp_as_gp information.
11104 (nds32_option_override): Adjust register usage.
11105 (nds32_expand_prologue): Consider fp_as_gp situation.
11106 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
11107 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
11110 (simple_return): Ditto.
11111 (omit_fp_begin): Output special directive for fp_as_gp.
11112 (omit_fp_end): Output special directive for fp_as_gp.
11113 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
11114 mforbid-fp-as-gp): New options.
11116 2018-06-01 Mark Wielaard <mark@klomp.org>
11118 * dwarf2out.c (dwarf2out_finish): Remove generation of
11119 DW_AT_loclists_base.
11121 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
11123 * gimple-ssa-store-merging.c: Include gimple-fold.h.
11124 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
11125 (struct merged_store_group): Add bit_insertion field.
11126 (dump_char_array): Use standard hexadecimal format.
11127 (merged_store_group::merged_store_group): Set bit_insertion to false.
11128 (merged_store_group::apply_stores): Use optimal buffer size. Deal
11129 with BIT_INSERT_EXPR stores. Move up code updating the mask and
11130 also print the mask in the dump file.
11131 (pass_store_merging::gate): Minor tweak.
11132 (imm_store_chain_info::coalesce_immediate): Fix wrong association
11133 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
11134 stores with INTEGER_CST stores.
11135 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
11136 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
11137 and use it throughout. Generate bit insertion sequences if need be.
11138 (pass_store_merging::process_store): Remove redundant condition.
11139 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
11141 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
11143 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
11144 the 128-bit floating point types. Fix function comment.
11146 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11148 * config/aarch64/aarch64-simd.md
11149 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
11151 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
11154 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
11156 PR tree-optimization/85989
11157 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
11159 (backprop::intersect_uses): Check it when deciding whether this
11160 is a backedge reference.
11161 (backprop::process_block): Add each phi to m_visited_phis
11162 after visiting it, then clear it at the end.
11164 2018-06-01 Richard Biener <rguenther@suse.de>
11166 * tree-vectorizer.h (vect_dr_stmt): New function.
11167 (vect_get_load_cost): Adjust.
11168 (vect_get_store_cost): Likewise.
11169 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
11170 Use vect_dr_stmt instead of DR_SMTT.
11171 (vect_record_base_alignments): Likewise.
11172 (vect_calculate_target_alignment): Likewise.
11173 (vect_compute_data_ref_alignment): Likewise and make static.
11174 (vect_update_misalignment_for_peel): Likewise.
11175 (vect_verify_datarefs_alignment): Likewise.
11176 (vector_alignment_reachable_p): Likewise.
11177 (vect_get_data_access_cost): Likewise. Pass down
11178 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
11179 (vect_get_peeling_costs_all_drs): Likewise.
11180 (vect_peeling_hash_get_lowest_cost): Likewise.
11181 (vect_enhance_data_refs_alignment): Likewise.
11182 (vect_find_same_alignment_drs): Likewise.
11183 (vect_analyze_data_refs_alignment): Likewise.
11184 (vect_analyze_group_access_1): Likewise.
11185 (vect_analyze_group_access): Likewise.
11186 (vect_analyze_data_ref_access): Likewise.
11187 (vect_analyze_data_ref_accesses): Likewise.
11188 (vect_vfa_segment_size): Likewise.
11189 (vect_small_gap_p): Likewise.
11190 (vectorizable_with_step_bound_p): Likewise.
11191 (vect_prune_runtime_alias_test_list): Likewise.
11192 (vect_analyze_data_refs): Likewise.
11193 (vect_supportable_dr_alignment): Likewise.
11194 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
11195 (vect_gen_prolog_loop_niters): Likewise.
11196 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
11197 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
11199 (vect_recog_mask_conversion_pattern): Likewise.
11200 (vect_try_gather_scatter_pattern): Likewise.
11201 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
11202 to vect_get_store_cost.
11203 (vect_get_store_cost): Get stmt_info instead of DR.
11204 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
11205 (vect_get_load_cost): Get stmt_info instead of DR.
11207 2018-06-01 Richard Biener <rguenther@suse.de>
11209 PR middle-end/86017
11210 * gimple-fold.c (var_decl_component_p): Also allow offsetted
11211 vars wrapped in MEM_REFs.
11213 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
11215 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
11216 Fix subreg tests so that we only return a choice between
11217 GENERAL_REGS and FP_REGS if the original classes included both.
11219 2018-06-01 Richard Biener <rguenther@suse.de>
11222 * tree-ssa-structalias.c (get_function_part_constraint):
11223 Handle NULL fi->decl.
11224 (find_func_aliases_for_call): Properly handle indirect
11225 fi from direct call.
11226 (find_func_clobbers): Likewise.
11227 (ipa_pta_execute): Likewise.
11228 (create_variable_info_for): For functions that are ifunc_resolver
11229 resolve to a varinfo that contains the result of the resolver call.
11230 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
11233 2018-05-31 Michael Collison <michael.collison@arm.com>
11235 * config/aarch64/aarch64.md:
11236 (*fix_to_zero_extenddfdi2): New pattern.
11237 * gcc.target/aarch64/fix_extend1.c: New testcase.
11239 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
11241 PR middle-end/78809
11242 PR middle-end/83026
11243 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
11244 and BUILT_IN_STRNCMP_EQ.
11245 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
11246 BUILT_IN_STRNCMP_EQ.
11247 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
11248 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
11249 (gimple_fold_builtin): Likewise.
11250 * tree-ssa-strlen.c (compute_string_length): New function.
11251 (determine_min_obsize): New function.
11252 (handle_builtin_string_cmp): New function to handle calls to
11253 string compare functions.
11254 (strlen_optimize_stmt): Add handling to builtin string compare
11256 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
11257 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
11258 * tree.c (build_common_builtin_nodes): Add new defines of
11259 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
11261 2018-05-31 Jakub Jelinek <jakub@redhat.com>
11264 * bb-reorder.c (pass_partition_blocks::gate): Return false for
11265 functions with naked attribute.
11267 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
11269 * config/i386/sse.md (avx_vec_concat<mode>):
11270 Substitute concat_tg_mode mode attribute with xtg_mode.
11271 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
11272 (concat_tg_mode): Remove mode attribute.
11274 2018-05-31 Martin Sebor <msebor@redhat.com>
11277 * calls.c (alloc_max_size): Correct a logic error/typo.
11278 Treat excessive arguments as infinite. Warn for invalid arguments.
11279 * doc/invoke.texi (-Walloc-size-larger-than): Update.
11281 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
11284 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
11285 and movx for Haswell.
11287 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
11288 Cesar Philippidis <cesar@codesourcery.com>
11290 PR middle-end/85879
11291 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
11292 when emitting error on private/firstprivate reductions.
11293 * omp-low.c (lower_omp_target): Avoid reference-type processing
11294 on pointers for firstprivate clause.
11296 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
11298 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
11301 * config/aarch64/aarch64-simd.md
11302 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
11303 (aarch64_ld1_x3_<mode>): Likewise
11304 (aarch64_st1x2<VALLDIF:mode>): Likewise
11305 (aarch64_st1_x2_<mode>): Likewise
11306 (aarch64_st1x3<VALLDIF:mode>): Likewise
11307 (aarch64_st1_x3_<mode>): Likewise
11308 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
11309 (vld1_s8_x3): Likewise.
11310 (vld1_u16_x3): Likewise.
11311 (vld1_s16_x3): Likewise.
11312 (vld1_u32_x3): Likewise.
11313 (vld1_s32_x3): Likewise.
11314 (vld1_u64_x3): Likewise.
11315 (vld1_s64_x3): Likewise.
11316 (vld1_f16_x3): Likewise.
11317 (vld1_f32_x3): Likewise.
11318 (vld1_f64_x3): Likewise.
11319 (vld1_p8_x3): Likewise.
11320 (vld1_p16_x3): Likewise.
11321 (vld1_p64_x3): Likewise.
11322 (vld1q_u8_x3): Likewise.
11323 (vld1q_s8_x3): Likewise.
11324 (vld1q_u16_x3): Likewise.
11325 (vld1q_s16_x3): Likewise.
11326 (vld1q_u32_x3): Likewise.
11327 (vld1q_s32_x3): Likewise.
11328 (vld1q_u64_x3): Likewise.
11329 (vld1q_s64_x3): Likewise.
11330 (vld1q_f16_x3): Likewise.
11331 (vld1q_f32_x3): Likewise.
11332 (vld1q_f64_x3): Likewise.
11333 (vld1q_p8_x3): Likewise.
11334 (vld1q_p16_x3): Likewise.
11335 (vld1q_p64_x3): Likewise.
11336 (vst1_s64_x2): Likewise.
11337 (vst1_u64_x2): Likewise.
11338 (vst1_f64_x2): Likewise.
11339 (vst1_s8_x2): Likewise.
11340 (vst1_p8_x2): Likewise.
11341 (vst1_s16_x2): Likewise.
11342 (vst1_p16_x2): Likewise.
11343 (vst1_s32_x2): Likewise.
11344 (vst1_u8_x2): Likewise.
11345 (vst1_u16_x2): Likewise.
11346 (vst1_u32_x2): Likewise.
11347 (vst1_f16_x2): Likewise.
11348 (vst1_f32_x2): Likewise.
11349 (vst1_p64_x2): Likewise.
11350 (vst1q_s8_x2): Likewise.
11351 (vst1q_p8_x2): Likewise.
11352 (vst1q_s16_x2): Likewise.
11353 (vst1q_p16_x2): Likewise.
11354 (vst1q_s32_x2): Likewise.
11355 (vst1q_s64_x2): Likewise.
11356 (vst1q_u8_x2): Likewise.
11357 (vst1q_u16_x2): Likewise.
11358 (vst1q_u32_x2): Likewise.
11359 (vst1q_u64_x2): Likewise.
11360 (vst1q_f16_x2): Likewise.
11361 (vst1q_f32_x2): Likewise.
11362 (vst1q_f64_x2): Likewise.
11363 (vst1q_p64_x2): Likewise.
11364 (vst1_s64_x3): Likewise.
11365 (vst1_u64_x3): Likewise.
11366 (vst1_f64_x3): Likewise.
11367 (vst1_s8_x3): Likewise.
11368 (vst1_p8_x3): Likewise.
11369 (vst1_s16_x3): Likewise.
11370 (vst1_p16_x3): Likewise.
11371 (vst1_s32_x3): Likewise.
11372 (vst1_u8_x3): Likewise.
11373 (vst1_u16_x3): Likewise.
11374 (vst1_u32_x3): Likewise.
11375 (vst1_f16_x3): Likewise.
11376 (vst1_f32_x3): Likewise.
11377 (vst1_p64_x3): Likewise.
11378 (vst1q_s8_x3): Likewise.
11379 (vst1q_p8_x3): Likewise.
11380 (vst1q_s16_x3): Likewise.
11381 (vst1q_p16_x3): Likewise.
11382 (vst1q_s32_x3): Likewise.
11383 (vst1q_s64_x3): Likewise.
11384 (vst1q_u8_x3): Likewise.
11385 (vst1q_u16_x3): Likewise.
11386 (vst1q_u32_x3): Likewise.
11387 (vst1q_u64_x3): Likewise.
11388 (vst1q_f16_x3): Likewise.
11389 (vst1q_f32_x3): Likewise.
11390 (vst1q_f64_x3): Likewise.
11391 (vst1q_p64_x3): Likewise.
11393 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
11395 * config/msp430/msp430.c (msp430_output_labelref): Prepend
11396 user_label_prefix to name.
11398 * tree-core.h: Update comment about the format of NAME string
11399 passed to handler in attribute_spec.
11401 * config/msp430/msp430.md: Remove erroneous subreg expression from
11402 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
11403 zero_extend{q,h}isi2.
11405 2018-05-30 Borislav Petkov <bp@suse.de>
11407 * doc/extend.texi: Document some architecture specific
11408 constraints and sort entries.
11410 2018-05-30 Martin Sebor <msebor@redhat.com>
11412 PR middle-end/85369
11413 * builtins.c (expand_builtin_stpcpy_1): New function.
11414 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
11415 only if the former succeeds.
11417 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
11419 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
11422 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11424 * doc/invoke.texi (-flinker-output): Document
11426 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11428 * passes.c (ipa_write_summaries): Only modify statements if body
11430 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
11431 incrementally linking.
11432 (ipa_passes): Likewise.
11433 * lto-cgraph.c (lto_output_node): When incrementally linking do not
11434 pass down resolution info.
11435 * common.opt (flag_incremental_link): Update info.
11436 * gcc.c (plugin specs): Turn flinker-output=* to
11437 -plugin-opt=-linker-output-known
11438 * toplev.c (compile_file): Also cut compilation when doing incremental
11440 * flag-types. (enum lto_partition_model): Add
11441 LTO_LINKER_OUTPUT_NOLTOREL.
11442 (invoke.texi): Add -flinker-output docs.
11443 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
11444 link same way as WPA; do not stream in dead initializers.
11446 * dwarf2out.c (dwarf2out_die_ref_for_decl,
11447 darf2out_register_external_decl): Support incremental link.
11449 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11451 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
11453 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11455 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
11456 it down to simple_object_copy_lto_debug_sections.
11457 (run_gcc): Determine incremental LTO link time and configure
11458 lto1 into non-wpa mode, disable renaming of debug sections.
11460 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
11462 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
11463 descriptions of various incorrectly documented functions.
11465 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
11468 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
11469 address check not strict.
11471 2018-05-30 Richard Biener <rguenther@suse.de>
11473 PR tree-optimization/85964
11474 * tracer.c (better_p): Drop initialized count check, we only
11475 call the function with initialized counts now.
11476 (find_best_successor): Do find a best edge if one
11477 has uninitialized count.
11478 (find_best_predecessor): Likewise. Do BB frequency check only
11479 if count is initialized.
11481 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
11483 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
11484 (aarch64_ldrstr_offset_compare): New.
11485 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
11486 load/store orderings.
11487 (aarch64_gen_adjusted_ldpstp): Likewise.
11489 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
11491 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
11492 Check for subset of GENERAL_REGS and FP_REGS.
11493 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
11496 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
11498 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
11499 and wi::to_poly_offset. Add the current offset and then check
11500 whether the sum fits, rather than using an unchecked addition of
11501 a checked term. Check for a shwi rather than a uhwi.
11502 * expr.c (get_bit_range): Use tree_to_poly_uint64.
11503 (store_constructor): Use poly_int_tree_p.
11504 (expand_expr_real_1): Likewise.
11505 * function.c (assign_temp): Likewise.
11506 * fold-const.c (const_binop): Use poly_int_tree_p and
11507 wi::to_poly_offset.
11508 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
11510 * ipa-icf-gimple.c (func_checker::compare_operand): Use
11511 to_poly_offset for MEM offsets.
11512 * ipa-icf.c (sem_variable::equals): Likewise.
11513 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
11514 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
11515 wi::to_poly_offset for BIT_FIELD_REF offsets.
11516 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
11517 wi::to_poly_offset.
11518 * var-tracking.c (emit_note_insn_var_location): Use
11519 tree_to_poly_uint64.
11521 2018-05-29 Jim Wilson <jimw@sifive.com>
11523 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
11525 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
11528 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
11529 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
11531 (sse4_1_round<mode>2): Use nonimmediate_operand
11532 for operand 1 predicate.
11534 2018-05-29 Martin Sebor <msebor@redhat.com>
11535 Richard Biener <rguenther@suse.de>
11538 * calls.c (get_size_range): Call determine_value_range instead
11539 of get_value_range..
11540 * tree-vrp.h (determine_value_range): Declared new function.
11541 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
11543 2018-05-29 Richard Biener <rguenther@suse.de>
11545 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
11546 sure to use non-pattern stmts for get_earlier_stmt arguments.
11547 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
11548 called on pattern stmts.
11549 (get_later_stmt): Likewise.
11551 2018-05-29 Martin Liska <mliska@suse.cz>
11553 PR gcov-profile/85759
11554 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
11557 2018-05-29 Jakub Jelinek <jakub@redhat.com>
11559 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
11561 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
11565 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
11566 VEC_PACK_FLOAT_EXPR): New tree codes.
11567 * tree-pretty-print.c (op_code_prio): Handle
11568 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
11569 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
11570 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
11571 * tree-inline.c (estimate_operator_cost): Likewise.
11572 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
11573 * fold-const.c (const_binop): Likewise.
11574 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
11575 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
11576 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
11577 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
11578 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
11579 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
11580 * expr.c (expand_expr_real_2): Likewise.
11581 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
11582 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
11583 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
11585 * optabs.c (expand_widen_pattern_expr): For
11586 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
11587 sign from result type rather than operand's type.
11588 (expand_binop_directly): For vec_packu_float_optab and
11589 vec_packs_float_optab allow result type to be different from operand's
11591 * optabs-tree.c (optab_for_tree_code): Handle
11592 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
11593 VEC_PACK_FLOAT_EXPR. Formatting fixes.
11594 * tree-vect-generic.c (expand_vector_operations_1): Handle
11595 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
11596 VEC_PACK_FLOAT_EXPR.
11597 * tree-vect-stmts.c (supportable_widening_operation): Handle
11599 (supportable_narrowing_operation): Handle FLOAT_EXPR.
11600 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
11601 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
11602 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
11603 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
11605 (vec_pack<floatprefix>_float_<mode>): New expander.
11606 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
11608 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
11609 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
11610 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
11611 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
11612 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
11614 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
11615 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
11616 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
11617 VEC_PACK_FLOAT_EXPR): Document.
11619 2018-05-29 Richard Biener <rguenther@suse.de>
11621 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
11623 (stmt_vec_info_vec): Make pointer.
11624 (init_stmt_vec_info_vec): Remove.
11625 (free_stmt_vec_info_vec): Likewise.
11626 (set_stmt_vec_info_vec): New function.
11627 (free_stmt_vec_infos): Likewise.
11628 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
11629 (set_vinfo_for_stmt): Likewise.
11630 (get_earlier_stmt): Likewise.
11631 (get_later_stmt): Likewise.
11632 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
11633 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
11634 (vec_info::~vec_info): Free stmt_vec_infos.
11635 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
11636 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
11637 (pass_slp_vectorize::execute): Likewise.
11638 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
11639 (free_stmt_vec_info_vec): Likewise.
11640 (set_stmt_vec_info_vec): New function.
11641 (free_stmt_vec_infos): Likewise.
11642 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
11643 the global stmt_vec_info_vec.
11644 * tree-parloops.c (gather_scalar_reductions): Use
11645 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
11648 2018-05-29 Richard Biener <rguenther@suse.de>
11650 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
11652 2018-05-29 Martin Liska <mliska@suse.cz>
11653 David Malcolm <dmalcolm@redhat.com>
11655 * vec.c (test_reverse): New.
11656 (vec_c_tests): Add new test.
11657 * vec.h (vl_ptr>::reverse): New function.
11659 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
11661 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
11663 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
11666 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
11668 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
11670 2018-05-28 Richard Biener <rguenther@suse.de>
11672 PR tree-optimization/85933
11673 * tree-vect-data-refs.c (vect_record_base_alignments): Only
11674 look at stmts marked as vectorizable.
11676 2018-05-28 Richard Biener <rguenther@suse.de>
11678 PR tree-optimization/85934
11679 * tree-vect-generic.c (expand_vector_operations_1): Hoist
11680 vector boolean check before scalar optimization.
11682 2018-05-28 Jakub Jelinek <jakub@redhat.com>
11684 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
11687 2018-05-28 Mark Wielaard <mark@klomp.org>
11689 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
11690 if it is an expression containing a minus sign.
11692 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
11694 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
11696 2018-05-27 Paul Koning <ni1d@arrl.net>
11698 * config/pdp11/pdp11.md (truncsihi2): Remove.
11700 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
11701 Chung-Ju Wu <jasonwucj@gmail.com>
11703 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
11705 (unaligned_store_dw): Ditto.
11706 * config/nds32/nds32-memory-manipulation.c
11707 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
11708 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
11709 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
11710 (emit_setmem_word_loop): Rename to ...
11711 (emit_setmem_doubleword_loop): ... this.
11712 (nds32_gen_dup_4_byte_to_word_value): New function.
11713 (nds32_gen_dup_8_byte_to_double_word_value): New function.
11714 (nds32_expand_setmem_loop): Refine implementation.
11715 (nds32_expand_setmem_loop_v3m): Ditto.
11716 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
11719 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
11721 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
11723 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
11725 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
11726 (nds32_init_machine_status): Initialize machine->attr_naked_p and
11727 machine->attr_no_prologue_p.
11728 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
11729 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
11730 (nds32_expand_epilogue): Consider attr_naked_p.
11731 (nds32_expand_epilogue_v3pop): Likewise.
11732 (nds32_can_use_return_insn): Likewise.
11733 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
11734 attr_no_prologue_p fields.
11735 * config/nds32/nds32.opt (mret-in-naked-func): New option.
11737 2018-05-27 Jakub Jelinek <jakub@redhat.com>
11740 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
11742 * config/i386/sse.md
11743 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
11745 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
11747 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
11749 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
11751 (*<floatsuffix>floatv2div2sf2): Rename to ...
11752 (*float<floatunssuffix>v2div2sf2): ... this.
11753 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
11754 (float<floatunssuffix>v2div2sf2_mask): ... this.
11755 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
11756 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
11757 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
11759 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
11761 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
11763 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
11765 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
11767 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
11769 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
11770 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
11771 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
11772 gen_ufix_truncv8dfv8si2.
11773 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
11774 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
11775 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
11776 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
11777 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
11778 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
11779 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
11780 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
11782 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
11786 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
11788 2018-05-25 Jim Wilson <jimw@sifive.com>
11790 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
11791 * config/riscv/riscv.c (struct machine_function): Add
11792 interrupt_handler_p and attribute_checked_p fields.
11793 (riscv_attribute_table): Add interrupt.
11794 (riscv_interrupt_type_p): New.
11795 (riscv_save_reg_p): Save extra regs for interrupt handler.
11796 (riscv_use_save_libcall): Return false for interrupt handler.
11797 (riscv_first_stack_step): Add forward declaration.
11798 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
11799 for interrupt handler with large frame. Use it for saved reg list.
11800 (riscv_expand_prologue): Move flag_stack_usage_info support to
11801 eliminate duplication.
11802 (riscv_expand_epilogue): Generate mret for interrupt handler.
11803 (riscv_epilogue_uses): New.
11804 (riscv_can_use_return_insn): Return false for interrupt handler.
11805 (riscv_function_ok_for_sibcall): Likewise.
11806 (riscv_set_current_function): Add interrupt handler support.
11807 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
11808 * config/riscv/riscv.md (UNSPECV_MRET): New.
11810 (riscv_frflags, riscv_fsflags): Use tab after opcode.
11812 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
11814 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
11816 PR tree-optimization/85712
11817 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
11818 this candidate has already been replaced in-situ by a copy.
11820 2018-05-25 Jason Merrill <jason@redhat.com>
11822 PR c++/80485 - inline function non-zero address.
11823 * symtab.c (nonzero_address): Check DECL_COMDAT.
11825 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
11828 * config/alpha/alpha.md (ashlsi3): New insn pattern.
11829 (*ashlsi_se): Rename from *ashldi_se. Define as sign
11830 extension of SImode operation. Use const123_operand predicate.
11831 (*saddsi_1): Remove.
11832 (*saddl_se_1): Ditto.
11833 (*ssubsi_1): Ditto.
11834 (*ssubl_se_1): Ditto.
11835 * config/alpha/predicates.md (const123_operand): New predicate.
11836 * config/alpha/constraints.md (P): Use IN_RANGE.
11838 2018-05-25 Richard Biener <rguenther@suse.de>
11840 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
11842 (ref_maybe_used_by_stmt_p): Likewise.
11843 (stmt_may_clobber_ref_p): Likewise.
11844 (stmt_may_clobber_ref_p_1): Likewise.
11845 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
11847 (ref_maybe_used_by_stmt_p): Likewise.
11848 (stmt_may_clobber_ref_p): Likewise.
11849 (stmt_may_clobber_ref_p_1): Likewise.
11850 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
11851 the alias oracle to disambiguate DRs with stmts DR analysis
11853 (vect_analyze_data_refs): Do not give up on not analyzable
11854 DRs for BB vectorization. Remove code truncating the dataref
11857 2018-05-25 Jakub Jelinek <jakub@redhat.com>
11860 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
11861 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
11862 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
11864 2018-05-25 Richard Biener <rguenther@suse.de>
11866 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
11867 function, combining stmt data ref gathering and fatal analysis
11869 (vect_analyze_data_refs): Remove now redudnant code and simplify.
11870 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
11871 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
11872 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
11873 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
11875 2018-05-25 Bin Cheng <bin.cheng@arm.com>
11877 PR tree-optimization/85720
11878 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
11879 SCC if all partitions are builtins.
11880 (version_loop_by_alias_check): New parameter. Generate cancelable
11881 runtime alias check if all partitions are builtins.
11882 (distribute_loop): Update call to above function.
11884 2018-05-25 Bin Cheng <bin.cheng@arm.com>
11886 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
11887 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
11888 (parm_default_def_partition_arg): Ditto.
11889 (set_parm_default_def_partition): Ditto.
11890 (get_parm_default_def_partitions): Ditto and make it static.
11891 (get_undefined_value_partitions): Ditto and make it static.
11892 (remove_ssa_form): Refactor call to init_var_map here.
11893 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
11894 computation for loop region.
11895 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
11896 (register_default_def): Delete.
11897 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
11898 (parm_default_def_partition_arg): Ditto.
11899 (set_parm_default_def_partition): Ditto.
11900 (get_parm_default_def_partitions): Ditto and make it static.
11901 (get_undefined_value_partitions): Ditto and make it static.
11902 (coalesce_with_default, coalesce_with_default): Update comment.
11903 (create_coalesce_list_for_region): New func factored out from
11904 create_outofssa_var_map.
11905 (populate_coalesce_list_for_outofssa): New func factored out from
11906 create_outofssa_var_map and coalesce_ssa_name.
11907 (create_outofssa_var_map): Delete.
11908 (coalesce_ssa_name): Refactor to support live range computation.
11909 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
11910 (get_parm_default_def_partitions): Delete.
11911 (get_undefined_value_partitions): Ditto.
11912 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
11913 computation for loop region.
11914 (new_tree_live_info, loe_visit_block): Ditto.
11915 (live_worklist, set_var_live_on_entry): Ditto.
11916 (calculate_live_on_exit, verify_live_on_entry): Ditto.
11917 * tree-ssa-live.h (struct _var_map): New fields.
11918 (init_var_map): Change decl.
11919 (region_contains_p): New.
11921 2018-05-25 Bin Cheng <bin.cheng@arm.com>
11923 * tree-ssa-live.h (live_merge_and_clear): Delete.
11925 2018-05-25 Richard Biener <rguenther@suse.de>
11928 * tree-dump.c (dequeue_and_dump): Remove access to removed
11929 operand 2 of a SWITCH_EXPR.
11931 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11933 * doc/sourcebuild.texi (vect_double_cond_arith): Include
11934 multiplication and division.
11935 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
11936 (cond_udiv@var{m}, cond_umod@var{m}): Document.
11937 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
11938 (cond_udiv_optab, cond_umod_optab): New optabs.
11939 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
11940 (IFN_COND_RDIV): New internal functions.
11941 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
11942 TRUNC_MOD_EXPR and RDIV_EXPR.
11943 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
11944 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
11946 (SVE_INT_BINARY): Include mult.
11947 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
11948 (optab, sve_int_op): Handle mult.
11949 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
11951 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
11952 for SVE_INT_BINARY_SD.
11954 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11956 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
11957 (optab, sve_int_op): Handle div and udiv.
11958 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
11959 for SVE_INT_BINARY_SD.
11960 (*<optab><mode>3): New insn for the same.
11962 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11964 * tree-vect-patterns.c: Include predict.h.
11965 (vect_recog_divmod_pattern): Restrict check for division support
11966 to when optimizing for size.
11968 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11970 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
11971 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
11972 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
11973 (gimple_match_op::set_op): Likewise.
11974 (gimple_resimplify4): Declare.
11975 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
11976 (expr::gen_transform): Likewise.
11977 (decision_tree::gen): Generate a simplification routine for 4 operands.
11978 * gimple-match-head.c (gimple_simplify): Add an overload for
11979 4 operands. In the top-level function, handle up to 4 call
11980 arguments and call gimple_resimplify4.
11981 (gimple_resimplify4): New function.
11982 (build_call_internal): Pass a fourth operand.
11983 (maybe_push_to_seq): Likewise.
11984 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
11985 Fold VEC_COND_EXPRs of an operation and a default value into
11986 an IFN_COND_* function if possible.
11987 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
11989 (SVE_COND_FP_BINARY): Include them.
11990 (optab, sve_fp_op): Handle them.
11991 (SVE_INT_BINARY_REV): New code iterator.
11992 (SVE_COND_FP_BINARY_REV): New int iterator.
11993 (commutative): New int attribute.
11994 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
11996 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
11998 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
11999 (*cond_<optab><mode>): New patterns for reversed operands.
12001 2018-05-25 Richard Biener <rguenther@suse.de>
12003 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
12004 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
12005 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
12006 (STMT_VINFO_GROUPED_ACCESS): Adjust.
12007 * tree-vect-data-refs.c (everywhere): Adjust users.
12008 * tree-vect-loop.c (everywhere): Likewise.
12009 * tree-vect-slp.c (everywhere): Likewise.
12010 * tree-vect-stmts.c (everywhere): Likewise.
12011 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
12013 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12015 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
12017 (gcc_cv_as_section_exclude): ... this.
12018 Try Solaris as #exclude syntax.
12019 * configure: Regenerate.
12020 * config.in: Regenerate.
12021 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
12023 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
12024 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
12026 * varasm.c (default_elf_asm_named_section): Don't check if
12027 HAVE_GAS_SECTION_EXCLUDE is defined.
12029 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
12031 * doc/md.texi: Update the documentation of the cond_* optabs
12032 to mention the new final operand. Fix GET_MODE_NUNITS call.
12033 Describe the scalar case too.
12034 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
12035 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
12037 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
12038 (get_conditional_internal_fn): Update comment.
12039 * tree-vect-loop.c (vectorizable_reduction): Pass the original
12040 accumulator value as a final argument to conditional functions.
12041 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
12042 a define_expand and add an "else" operand. Assert for now that
12043 the else operand is equal to operand 2. Use SVE_INT_BINARY and
12044 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
12045 (*cond_<optab><mode>): New patterns.
12046 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
12047 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
12048 (UNSPEC_COND_EOR): Delete.
12049 (optab): Remove associated mappings.
12050 (SVE_INT_BINARY): New code iterator.
12051 (sve_int_op): Remove int attribute and add "minus" to the code
12053 (SVE_COND_INT_OP): Delete.
12054 (SVE_COND_FP_OP): Rename to...
12055 (SVE_COND_FP_BINARY): ...this.
12057 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
12059 * optabs.c (can_reuse_operands_p): New function.
12060 (maybe_legitimize_operands): Try to reuse the results for
12063 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
12065 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
12066 Add {q} suffix to insn mnemonic.
12068 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
12070 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
12071 (msp430_warn_func_return): New.
12073 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
12075 * fold-const.c (tree_nonzero_bits): New function.
12076 * fold-const.h (tree_nonzero_bits): Likewise.
12077 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
12078 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
12080 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
12084 * varasm.c (assemble_alias): Check ifunc_resolver only on
12087 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
12090 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
12091 when memory input operand is handled.
12093 2018-05-24 Luis Machado <luis.machado@linaro.org>
12095 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
12097 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
12099 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
12101 * match.pd: Delay FMA folds until after vectorization.
12103 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
12106 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
12107 address check not strict.
12109 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
12111 * gimple-match.h (gimple_match_op): New class.
12112 (mprts_hook): Replace parameters with a gimple_match_op *.
12113 (maybe_build_generic_op): Likewise.
12114 (gimple_simplified_result_is_gimple_val): Replace parameters with
12115 a const gimple_match_op *.
12116 (gimple_simplify): Replace code_helper * and tree * parameters with
12117 a gimple_match_op * parameter.
12118 (gimple_resimplify1): Replace code_helper *, tree and tree *
12119 parameters with a gimple_match_op * parameter.
12120 (gimple_resimplify2): Likewise.
12121 (gimple_resimplify3): Likewise.
12122 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
12123 parameters with a gimple_match_op * parameter.
12124 * gimple-match-head.c (gimple_simplify): Change prototypes of
12125 auto-generated functions to take a gimple_match_op * instead of
12126 separate code_helper * and tree * parameters. Make the same
12127 change in the top-level overload and update calls to the
12128 gimple_resimplify routines. Update calls to the auto-generated
12129 functions and to maybe_push_res_to_seq in the publicly-facing
12130 operation-specific gimple_simplify overloads.
12131 (gimple_match_op::MAX_NUM_OPS): Define.
12132 (gimple_resimplify1): Replace rcode and ops with a single res_op
12133 parameter. Update call to gimple_simplify.
12134 (gimple_resimplify2): Likewise.
12135 (gimple_resimplify3): Likewise.
12136 (mprts_hook): Replace parameters with a gimple_match_op *.
12137 (maybe_build_generic_op): Likewise.
12138 (build_call_internal): Replace type, nargs and ops with
12139 a gimple_match_op *.
12140 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
12141 with a single gimple_match_op *. Update calls to mprts_hook,
12142 build_call_internal and gimple_simplified_result_is_gimple_val.
12143 Factor out code that is common to the tree_code and combined_fn cases.
12144 * genmatch.c (expr::gen_transform): Replace tem_code and
12145 tem_ops with a gimple_match_op called tem_op. Update calls
12146 to the gimple_resimplify functions and maybe_push_res_to_seq.
12147 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
12148 res_ops. Update call to the gimple_resimplify functions.
12149 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
12150 (decision_tree::gen): Make the functions take a gimple_match_op *
12151 called res_op instead of separate res_code and res_ops parameters.
12152 Update call accordingly.
12153 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
12154 and ops with a single res_op parameter. Update calls to
12155 maybe_build_generic_op and maybe_push_res_to_seq.
12156 (fold_stmt_1): Update calls to gimple_simplify and
12157 replace_stmt_with_simplification.
12158 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
12159 and gimple_simplified_result_is_gimple_val.
12160 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
12162 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
12163 with a gimple_match_op *.
12164 (vn_nary_build_or_lookup): Likewise. Update call to
12165 vn_nary_build_or_lookup_1.
12166 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
12167 gimple_match_op *. Update calls to the gimple_resimplify routines
12168 and to gimple_simplified_result_is_gimple_val.
12169 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
12170 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
12171 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
12172 (visit_nary_op): Likewise.
12173 (visit_reference_op_load): Likewise.
12175 2018-05-23 Luis Machado <luis.machado@linaro.org>
12177 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
12178 modifier for printing the step amount.
12180 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
12183 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
12186 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
12188 * doc/sourcebuild.texi (Endianness): New subsubsection.
12190 2018-05-23 Luis Machado <luis.machado@linaro.org>
12192 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12193 <prefetch_dynamic_strides>: New const bool field.
12194 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12195 prefetch_dynamic_strides.
12196 (exynosm1_prefetch_tune): Likewise.
12197 (thunderxt88_prefetch_tune): Likewise.
12198 (thunderx_prefetch_tune): Likewise.
12199 (thunderx2t99_prefetch_tune): Likewise.
12200 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
12202 (aarch64_override_options_internal): Update to set
12203 PARAM_PREFETCH_DYNAMIC_STRIDES.
12204 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
12205 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
12206 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
12207 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
12208 prefetch-dynamic-strides setting.
12210 2018-05-23 Luis Machado <luis.machado@linaro.org>
12212 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12213 <minimum_stride>: New const int field.
12214 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12215 minimum_stride field defaulting to -1.
12216 (exynosm1_prefetch_tune): Likewise.
12217 (thunderxt88_prefetch_tune): Likewise.
12218 (thunderx_prefetch_tune): Likewise.
12219 (thunderx2t99_prefetch_tune): Likewise.
12220 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
12221 <default_opt_level>: Set to 3.
12222 (aarch64_override_options_internal): Update to set
12223 PARAM_PREFETCH_MINIMUM_STRIDE.
12224 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
12225 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
12226 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
12227 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
12228 stride is constant and is below the minimum stride threshold.
12230 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12232 * config/arm/arm-cpus.in (mode26): Delete.
12233 (armv4): Delete mode26 reference.
12234 * config/arm/arm.c (arm_configure_build_target): Delete use of
12237 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
12239 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
12241 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
12242 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
12244 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
12245 (floatunsdidf2): Ditto.
12247 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
12249 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
12250 (fixuns_trunc<mode>si2_avx512f): Ditto.
12251 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
12252 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
12253 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
12255 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
12257 PR rtl-optimization/79985
12258 * df-scan.c (df_insn_refs_collect): Remove special case for
12259 global registers and asm statements.
12261 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
12263 * extend.texi (Global Register Variables): Rewrite the bullet list.
12264 Note that the register is available for allocation. Note that access
12265 via inline asm must use constraints. Add note about async-signal
12266 handlers. Remove paragraph about automagic register selection.
12268 2018-05-23 Richard Biener <rguenther@suse.de>
12270 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
12271 of fixed offset from memset VN.
12273 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
12275 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
12276 first_interp field.
12277 (alloc_cand_and_find_basis): Initialize first_interp field.
12278 (slsr_process_mul): Modify first_interp field.
12279 (slsr_process_add): Likewise.
12280 (slsr_process_cast): Modify first_interp field for each new
12282 (slsr_process_copy): Likewise.
12283 (dump_candidate): Dump first_interp field.
12284 (replace_mult_candidate): Process all interpretations, not just
12286 (replace_rhs_if_not_dup): Likewise.
12287 (replace_one_candidate): Likewise.
12289 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
12291 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
12293 (aarch64_needs_frame_chain): New function.
12294 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
12296 2018-05-23 Sudakshina Das <sudi.das@arm.com>
12299 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
12300 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
12301 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
12302 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
12303 as true for strict-align.
12304 (aarch64_can_inline_p): Perform checks even when callee has no
12305 attributes to check for strict alignment.
12306 * doc/extend.texi (AArch64 Function Attributes): Document
12308 * doc/invoke.texi: (AArch64 Options): Likewise.
12310 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
12312 PR tree-optimization/85853
12313 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
12314 the handling of the root of the node to...
12315 (vect_slp_analyze_node_operations_1): ...this new function,
12316 and run the whole thing with the child nodes' def types
12317 set according to their SLP node's def type.
12319 2018-05-23 Richard Biener <rguenther@suse.de>
12321 PR middle-end/85874
12322 * tree-data-ref.c (create_runtime_alias_checks): Defer
12323 and ignore overflow warnings.
12325 2018-05-23 Yury Gribov <tetra2005@gmail.com>
12327 PR tree-optimization/85822
12328 * tree-vrp.c (is_masked_range_test): Fix handling of negative
12331 2018-05-23 Richard Biener <rguenther@suse.de>
12333 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
12334 memset constants via native_interpret_expr.
12336 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
12339 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
12341 (cgraph_node::create_alias): Likewise.
12342 (cgraph_node::get_availability): Check ifunc_resolver instead
12343 of looking up ifunc attribute.
12344 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
12345 * varasm.c (do_assemble_alias): Likewise.
12346 (assemble_alias): Likewise.
12347 (default_binds_local_p_3): Likewise.
12348 * cgraph.h (cgraph_node): Add ifunc_resolver.
12349 (cgraph_node::only_called_directly_or_aliased_p): Return false
12350 for IFUNC resolver.
12351 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
12353 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
12354 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
12355 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
12356 instead of looking up ifunc attribute.
12358 2018-05-22 Luis Machado <luis.machado@linaro.org>
12360 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
12362 2018-05-22 Martin Sebor <msebor@redhat.com>
12364 PR middle-end/85359
12365 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
12366 only when expasion succeeds.
12367 (expand_builtin_strcmp): Same.
12368 (expand_builtin_strncmp): Same.
12370 2018-05-22 Martin Sebor <msebor@redhat.com>
12372 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
12374 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
12375 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12377 * config/aarch64/aarch64-ldpstp.md: Replace uses of
12378 aarch64_mem_pair_operand with memory_operand and delete operand swapping
12380 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
12381 Add check for legitimate_address.
12382 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
12383 (aarch64_swap_ldrstr_operands): New.
12384 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
12387 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
12388 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12390 * config/aarch64/aarch64.md: New patterns to generate stp
12392 (store_pair_sw, store_pair_dw): New patterns to generate stp for
12393 single words and double words.
12394 (load_pair_sw, load_pair_dw): Likewise.
12395 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
12397 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
12399 * config/aarch64/aarch64-ldpstp.md: Modify peephole
12400 for different mode ldpstp and add peephole for merged zero stores.
12401 Likewise for loads.
12402 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
12404 (aarch64_gen_store_pair): Rename calls to match new patterns.
12405 (aarch64_gen_load_pair): Rename calls to match new patterns.
12406 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
12407 (load_pair<DREG:mode><DREG2:mode>): ... This.
12408 (store_pair<mode>): Rename to...
12409 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
12410 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
12411 New mode iterators.
12412 (V_INT_EQUIV): Handle SImode.
12413 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
12416 2018-05-22 Martin Sebor <msebor@redhat.com>
12419 * calls.c (maybe_warn_nonstring_arg): Use string length to set
12420 or ajust the presumed bound on an operation to avoid unnecessary
12423 2018-05-22 Martin Sebor <msebor@redhat.com>
12425 PR tree-optimization/85826
12426 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
12427 assuming that a DECL necesarily has a constant size.
12429 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
12431 PR middle-end/85862
12432 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
12434 2018-05-22 Richard Biener <rguenther@suse.de>
12436 PR tree-optimization/85834
12437 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
12438 non-constant and non-zero memset arguments.
12440 2018-05-22 Martin Liska <mliska@suse.cz>
12443 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
12445 2018-05-22 Richard Biener <rguenther@suse.de>
12447 PR tree-optimization/85863
12448 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
12449 comparisons when vectype is specified.
12450 (vectorizable_condition): Do not specify vectype for
12451 vect_is_simple_cond when SLP vectorizing.
12453 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
12456 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
12457 define __ibm128 as long double.
12458 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
12459 as a distinct type when IEEE 128-bit support is enabled.
12460 (init_float128_ieee): Fix up conversions between IFmode and IEEE
12461 128-bit types to use the correct functions.
12462 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
12463 convert between 128-bit floating point types that have different
12464 modes but the same representation, instead of using gen_lowpart to
12466 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
12468 (IFKF_reg): New attributes to give the register constraints for
12470 (extend<mode>tf2_internal): New insns to mark an explicit
12471 conversion between 128-bit floating point types that have a
12472 different mode but share the same representation.
12474 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
12476 PR tree-optimization/85814
12477 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
12478 a null return from get_strinfo when unsharing the next
12479 strinfo in the chain.
12481 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
12484 * varasm.c (weak_finish): Clean up weak_decls.
12486 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12488 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
12489 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
12490 UNSPEC_UADALP values.
12491 * config/aarch64/iterators.md (ABAL): New int iterator.
12494 (sur): Add mappings for the above.
12495 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
12497 (aarch64_<sur>abal<mode>_4): Likewise.
12498 (aarch64_<sur>adalp<mode>_3): Likewise.
12499 (<sur>sadv16qi): New define_expand.
12501 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
12503 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
12504 (*movdf_internal): Ditto.
12505 (*rcpsf2_sse): Ditto.
12506 (*rsqrtsf2_sse): Ditto.
12507 (*sqrt<mode>2_sse): Ditto.
12509 2018-05-21 Tamar Christina <tamar.christina@arm.com>
12511 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
12513 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
12514 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
12515 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
12516 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
12518 * config/aarch64/arm_neon.h: Likewise.
12519 * config/aarch64/iterators.md (VQ_I): New.
12521 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
12523 * config.gcc: Add arc/t-multilib-linux to tmake_file for
12525 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
12528 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
12530 * config/nds32/constraints.md (S): New constraint.
12531 * config/nds32/nds32.md (call_internal): Use constraint S.
12532 (call_value_internal): Likewise.
12533 (sibcall_internal): Likewise.
12534 (sibcall_value_internal): Likewise.
12536 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
12537 Chung-Ju Wu <jasonwucj@gmail.com>
12539 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
12540 into consideration.
12542 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
12543 Chung-Ju Wu <jasonwucj@gmail.com>
12545 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
12546 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
12547 (nds32_rtx_costs_impl): Simplify.
12548 (nds32_address_cost_impl): Simplify.
12549 (nds32_init_rtx_costs): New function.
12550 (nds32_rtx_costs_speed_prefer): Likewise.
12551 (nds32_rtx_costs_size_prefer): Likewise.
12552 (nds32_address_cost_speed_prefer): Likewise.
12553 (nds32_address_cost_speed_fwprop): Likewise.
12554 (nds32_address_cost_size_prefer): Likewise.
12555 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
12556 * config/nds32/nds32.c (nds32_option_override): Use
12557 nds32_init_rtx_costs function.
12559 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
12561 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
12562 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
12563 (TARGET_PIPELINE_N8): Likewise.
12564 (TARGET_PIPELINE_N10): Likewise.
12565 (TARGET_PIPELINE_N13): Likewise.
12566 (TARGET_PIPELINE_GRAYWOLF): Likewise.
12568 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
12570 * config/nds32/nds32-fpu.md: Update copyright year.
12572 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12574 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
12576 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12578 * config/nds32/nds32.c
12579 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
12580 * config/nds32/nds32.opt (minline-asm-r15): New option.
12582 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12584 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
12586 * config/nds32/nds32.md (abssi2): New pattern.
12588 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
12590 * config/i386/i386.md (rex64namesuffix): New mode attribute.
12591 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
12592 Merge insn pattern from sse_cvtsi2ss<round_name> and
12593 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
12594 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
12595 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
12596 using SWI48 mode iterator.
12597 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
12598 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
12599 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
12600 pattern from sse_cvttss2si<round_saeonly_name>
12601 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
12602 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
12603 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
12604 using SWI48 mode iterator.
12605 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
12606 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
12607 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
12608 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
12609 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
12610 using SWI48 mode iterator.
12611 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
12612 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
12613 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
12614 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
12615 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
12616 SWI48 mode iterator.
12617 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
12618 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
12619 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
12620 pattern from sse_cvttsd2si<round_saeonly_name>
12621 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
12623 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12625 * config/nds32/nds32-md-auxiliary.c
12626 (nds32_valid_smw_lwm_base_p): Refine.
12627 (nds32_output_smw_single_word): Refine.
12628 (nds32_output_smw_double_word): New.
12629 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
12631 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12633 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
12634 (nds32_output_stack_pop): Refine.
12635 (nds32_expand_unaligned_load): Refine.
12636 (nds32_expand_unaligned_store): Refine.
12638 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
12639 Chung-Ju Wu <jasonwucj@gmail.com>
12641 * config/nds32/constants.md: Add TP_REGNUM constant.
12642 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
12643 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
12645 * config/nds32/nds32-doubleword.md: Consider flag_pic.
12646 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
12647 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
12648 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
12649 and PIC code generation.
12650 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
12652 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
12654 * config/nds32/nds32.md: Support TLS and PIC.
12655 * config/nds32/nds32.c: Support TLS and PIC.
12656 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
12657 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
12660 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12662 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
12663 mode with E_ prefix.
12665 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
12666 Chung-Ju Wu <jasonwucj@gmail.com>
12668 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
12669 * config/nds32/nds32-md-auxiliary.c
12670 (symbolic_reference_mentioned_p): New.
12671 (nds32_legitimize_ict_address): New.
12672 (nds32_expand_ict_move): New.
12673 (nds32_indirect_call_referenced_p): New.
12674 (nds32_symbol_binds_local_p): Delete.
12675 (nds32_long_call_p): Modify.
12676 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
12677 * config/nds32/nds32-protos.h
12678 (symbolic_reference_mentioned_p): Declare.
12679 (nds32_legitimize_ict_address): Declare.
12680 (nds32_expand_ict_move): Declare.
12681 (nds32_indirect_call_referenced_p): Declare.
12682 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
12683 (nds32_relax_group): Use nds32_ict_const_p as condition.
12684 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
12685 (nds32_asm_file_start): Output ict_model directive in asm code.
12686 (nds32_legitimate_address_p): Consider indirect call.
12687 (nds32_print_operand): Consider indirect call.
12688 (nds32_print_operand_address): Consider indirect call.
12689 (nds32_insert_attributes): Handle "indirect_call" attribute.
12690 (TARGET_LEGITIMATE_ADDRESS_P): Define.
12691 (TARGET_LEGITIMATE_CONSTANT_P): Define.
12692 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
12693 (TARGET_DELEGITIMIZE_ADDRESS): Define.
12694 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
12695 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
12696 (TARGET_ICT_MODEL_SMALL): Define.
12697 (TARGET_ICT_MODEL_LARGE): Define.
12698 * config/nds32/nds32.md (movsi): Consider ict model.
12699 (call, call_value): Consider ict model.
12700 (sibcall, sibcall_value): Consider ict model.
12701 * config/nds32/nds32.opt (mict-model): New option.
12702 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
12705 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
12706 Monk Chiang <sh.chiang04@gmail.com>
12707 Jim Wilson <jimw@sifive.com>
12709 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
12710 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
12711 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
12712 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
12713 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
12714 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
12715 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
12716 compute save_libcall_adjustment properly.
12717 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
12718 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
12719 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
12720 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
12721 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
12722 (ABI_SPEC): Handle mabi=ilp32e.
12723 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
12724 (RVE): Add RVE mask.
12725 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
12726 <-march>: Add rv32e as an example.
12728 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
12731 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
12732 (intra_create_variable_infos): Handle C++ constructors.
12734 2018-05-18 Martin Liska <mliska@suse.cz>
12736 * passes.def: Remove a redundant pass.
12738 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
12741 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
12743 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12745 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
12747 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
12749 (armv2, armv2a, armv3, armv3m): Delete architectures.
12750 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
12751 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
12752 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
12754 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
12755 (*mulsidi3adddi): Likewise.
12756 (mulsidi3): Likewise.
12757 (*mulsidi3_nov6): Likewise.
12758 (umulsidi3): Likewise.
12759 (umulsidi3_nov6): Likewise.
12760 (umaddsidi4): Likewise.
12761 (*umulsidi3adddi): Likewise.
12762 (smulsi3_highpart): Likewise.
12763 (*smulsi3_highpart_nov6): Likewise.
12764 (umulsi3_highpart): Likewise.
12765 (*umulsi3_highpart_nov6): Likewise.
12766 * config/arm/arm.h (arm_arch3m): Delete.
12767 * config/arm/arm.c (arm_arch3m): Delete.
12768 (arm_option_override_internal): Update armv3-related comment.
12769 (arm_configure_build_target): Delete use of isa_bit_mode32.
12770 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
12771 (arm_rtx_costs_internal): Delete check of arm_arch3m.
12772 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
12773 (mulsa3): Likewise.
12774 (mulusa3): Likewise.
12775 * config/arm/arm-protos.h (arm_arch3m): Delete.
12776 * config/arm/arm-tables.opt: Regenerate.
12777 * config/arm/arm-tune.md: Likewise.
12778 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
12779 deleted architectures.
12781 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12783 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
12784 (armv5t, armv5te): New features.
12785 (ARMv5, ARMv5e): Delete fgroups.
12786 (ARMv5t, ARMv5te): Adjust for above changes.
12787 (ARMv6m): Likewise.
12788 (armv5, armv5e): Delete arches.
12789 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
12791 (*call_reg_arm): Likewise.
12792 (*call_value_reg_armv5): Likewise.
12793 (*call_value_reg_arm): Likewise.
12794 (*call_symbol): Likewise.
12795 (*call_value_symbol): Likewise.
12796 (*sibcall_insn): Likewise.
12797 (*sibcall_value_insn): Likewise.
12798 (clzsi2): Likewise.
12799 (prefetch): Likewise.
12800 (define_split and define_peephole2 dependent on arm_arch5):
12802 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
12804 (TARGET_ARM_QBIT): Likewise.
12805 (TARGET_DSP_MULTIPLY): Likewise.
12806 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
12807 (arm_arch5, arm_arch5e): Delete.
12808 (arm_arch5t, arm_arch5te): Declare.
12809 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
12810 (arm_arch5t): Declare.
12811 (arm_option_reconfigure_globals): Update for the above.
12812 (arm_options_perform_arch_sanity_checks): Update comment, replace
12813 use of arm_arch5 with arm_arch5t.
12814 (use_return_insn): Likewise.
12815 (arm_emit_call_insn): Likewise.
12816 (output_return_instruction): Likewise.
12817 (arm_final_prescan_insn): Likewise.
12818 (arm_coproc_builtin_available): Likewise.
12819 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
12820 arm_arch5e with arm_arch5t and arm_arch5te.
12821 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
12822 (arm_arch5t, arm_arch5te): Declare.
12823 * config/arm/arm-tables.opt: Regenerate.
12824 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
12825 * config/arm/t-multilib: Likewise.
12826 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
12827 instead of arm_arch5.
12828 (*call_reg_thumb1): Likewise.
12829 (*call_value_reg_thumb1_v5): Likewise.
12830 (*call_value_reg_thumb1): Likewise.
12831 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
12833 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
12835 2018-05-18 Martin Liska <mliska@suse.cz>
12837 PR gcov-profile/84846
12838 * doc/gcov.texi: Document -t option of gcov tool.
12840 2018-05-18 Martin Liska <mliska@suse.cz>
12842 PR gcov-profile/84846
12843 * gcov.c (print_usage): Add new -t option.
12844 (process_args): Handle the option.
12845 (generate_results): Use stdout as output when requested by
12848 2018-05-18 Martin Liska <mliska@suse.cz>
12850 PR gcov-profile/84846
12851 * coverage.c (coverage_init): Write PWD to .gcno file.
12852 * doc/gcov.texi: Document how working directory is printed.
12853 * gcov-dump.c (dump_gcov_file): Print PWD.
12854 * gcov.c (output_intermediate_file): Likewise.
12855 (read_graph_file): Read PWD string.
12856 (output_lines): Print PWD.
12858 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12860 PR middle-end/85817
12861 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
12862 for retval and return false if all args to phi are zero.
12864 2018-05-18 Richard Biener <rguenther@suse.de>
12866 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
12868 (evrp_dom_walker::before_dom_children): Call it.
12870 2018-05-18 Richard Biener <rguenther@suse.de>
12872 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
12873 results when processing array refs with variable index.
12875 2018-05-18 Toon Moene <toon@moene.org>
12877 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
12878 directly after that of -floop-interchange. Indicate that both
12879 options are enabled by default when specifying -O3.
12881 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12883 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
12884 iterator. Delete separate integer-mode vec_set<mode> expander.
12885 (aarch64_simd_vec_setv2di): Delete.
12886 (vec_setv2di): Delete.
12887 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
12888 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
12889 the "w, r" alternative.
12891 2018-05-18 Martin Liska <mliska@suse.cz>
12893 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
12894 * tree-pass.h (make_pass_lower_switch_O0): New function.
12895 * tree-switch-conversion.c (node_has_low_bound): Remove.
12896 (node_has_high_bound): Likewise.
12897 (node_is_bounded): Likewise.
12898 (class pass_lower_switch): Make it a template type and create
12900 (pass_lower_switch::execute): Add template argument.
12901 (make_pass_lower_switch): New function.
12902 (make_pass_lower_switch_O0): New function.
12903 (do_jump_if_equal): Remove.
12904 (emit_case_nodes): Simplify to just handle all 3 cases and leave
12905 all the hard work to tree optimization passes.
12907 2018-05-18 Martin Liska <mliska@suse.cz>
12909 * dbgcnt.c (limit_low): Renamed from limit.
12910 (limit_high): New variable.
12911 (dbg_cnt_is_enabled): Check for upper limit.
12912 (dbg_cnt): Adjust dumping.
12913 (dbg_cnt_set_limit_by_index): Add new argument for high
12915 (dbg_cnt_set_limit_by_name): Likewise.
12916 (dbg_cnt_process_single_pair): Parse new format.
12917 (dbg_cnt_process_opt): Use strtok.
12918 (dbg_cnt_list_all_counters): Remove 'value' and add
12920 * doc/invoke.texi: Document changes.
12922 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
12924 * doc/sourcebuild.texi (scalar_all_fma): Document.
12925 * tree.def (FMA_EXPR): Delete.
12926 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
12927 * internal-fn.c (ternary_direct): New macro.
12928 (expand_ternary_optab_fn): Likewise.
12929 (direct_ternary_optab_supported_p): Likewise.
12930 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
12931 * builtins.c (fold_builtin_fma): Delete.
12932 (fold_builtin_3): Don't call it.
12933 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
12934 * expr.c (expand_expr_real_2): Likewise.
12935 * fold-const.c (operand_equal_p): Likewise.
12936 (fold_ternary_loc): Likewise.
12937 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
12938 * gimple.c (DEFTREECODE): Likewise.
12939 * gimplify.c (gimplify_expr): Likewise.
12940 * optabs-tree.c (optab_for_tree_code): Likewise.
12941 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
12942 * tree-eh.c (operation_could_trap_p): Likewise.
12943 (stmt_could_throw_1_p): Likewise.
12944 * tree-inline.c (estimate_operator_cost): Likewise.
12945 * tree-pretty-print.c (dump_generic_node): Likewise.
12946 (op_code_prio): Likewise.
12947 * tree-ssa-loop-im.c (stmt_cost): Likewise.
12948 * tree-ssa-operands.c (get_expr_operands): Likewise.
12949 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
12950 * fold-const-call.h (fold_fma): Delete.
12951 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
12952 CFN_FNMA and CFN_FNMS.
12953 (fold_fma): Delete.
12954 * genmatch.c (combined_fn): New enum.
12955 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
12956 (commutative_op): New function.
12957 (commutate): Use it. Handle more than 2 operands.
12958 (dt_operand::gen_gimple_expr): Use commutative_op.
12959 (parser::parse_expr): Allow :c to be used with non-binary
12960 operators if the commutative operand is known.
12961 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
12962 CFN_FMS, CFN_FNMA and CFN_FNMS.
12963 (backprop::process_assign_use): Remove FMA_EXPR handling.
12964 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
12965 (gen_hsa_fma): New function.
12966 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
12967 IFN_FNMA and IFN_FNMS.
12968 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
12969 * gimple-fold.h (follow_all_ssa_edges): Declare.
12970 * gimple-fold.c (follow_all_ssa_edges): New function.
12971 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
12972 gimple_build interface and use follow_all_ssa_edges to fold the result.
12973 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
12974 instead of checking for optabs directly.
12975 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
12976 rather than FMA_EXPRs.
12977 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
12978 call to IFN_FMA instead of an FMA_EXPR.
12980 2018-05-17 Jim Wilson <jimw@sifive.com>
12982 * expr.c (do_tablejump): When converting index to Pmode, if we have a
12983 sign extended promoted subreg, and the range does not have the sign bit
12984 set, then do a sign extend.
12986 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
12987 test, check for sign extended subreg and/or constant operands, and
12988 do a sign extend in that case.
12990 2018-05-17 Steve Ellcey <sellcey@cavium.com>
12992 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
12993 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
12995 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
12996 Change logics_shift_reg to logics_shift_imm.
12997 (thunderx2t99_fp_loadpair_basic): Delete.
12998 (thunderx2t99_fp_storepair_basic): Delete.
12999 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
13000 (thunderx2t99_asimd_polynomial): Delete.
13001 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
13002 and neon_fp_mul_d_scalar_q.
13003 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
13004 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
13005 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
13006 (thunderx2t99_asimd_lut): Add missing tbl types.
13007 (thunderx2t99_asimd_ext): Delete.
13008 (thunderx2t99_asimd_load1_1_mult): Delete.
13009 (thunderx2t99_asimd_load1_2_mult): Delete.
13010 (thunderx2t99_asimd_load1_ldp): New.
13011 (thunderx2t99_asimd_load1): New.
13012 (thunderx2t99_asimd_load2): Add missing *load2* types.
13013 (thunderx2t99_asimd_load3): New.
13014 (thunderx2t99_asimd_load4): New.
13015 (thunderx2t99_asimd_store1_1_mult): Delete.
13016 (thunderx2t99_asimd_store1_2_mult): Delete.
13017 (thunderx2t99_asimd_store2_mult): Delete.
13018 (thunderx2t99_asimd_store2_onelane): Delete.
13019 (thunderx2t99_asimd_store_stp): New.
13020 (thunderx2t99_asimd_store1): New.
13021 (thunderx2t99_asimd_store2): New.
13022 (thunderx2t99_asimd_store3): New.
13023 (thunderx2t99_asimd_store4): New.
13025 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
13027 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
13028 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
13030 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
13031 Segher Boessenkool <segher@kernel.crashing.org>
13034 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
13037 2018-05-17 Richard Biener <rguenther@suse.de>
13039 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
13040 for pruning loop and prune defs feeding only already visited PHIs.
13042 2018-05-17 Richard Biener <rguenther@suse.de>
13044 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
13046 2018-05-17 Bin Cheng <bin.cheng@arm.com>
13047 Richard Biener <rguenther@suse.de>
13049 PR tree-optimization/85793
13050 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
13051 for VMAT_ELEMENTWISE.
13053 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
13055 * internal-fn.h (lookup_internal_fn): Declare
13056 * internal-fn.c (lookup_internal_fn): New function.
13057 * gimple.c (gimple_build_call_from_tree): Handle calls to
13058 internal functions.
13059 * gimple-pretty-print.c (dump_gimple_call): Print "." before
13060 internal function names.
13061 * tree-pretty-print.c (dump_generic_node): Likewise.
13062 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
13064 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
13066 * gimple-fold.h (gimple_build): Make the function forms take
13067 combined_fn rather than built_in_function.
13068 (gimple_simplify): Likewise.
13069 * gimple-match-head.c (gimple_simplify): Likewise.
13070 * gimple-fold.c (gimple_build): Likewise.
13071 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
13072 rather than gimple_build_call_internal.
13073 (get_initial_defs_for_reduction): Likewise.
13074 (vect_create_epilog_for_reduction): Likewise.
13075 (vectorizable_live_operation): Likewise.
13077 2018-05-17 Martin Liska <mliska@suse.cz>
13079 * gimple-ssa-sprintf.c (format_directive): Do not use
13080 space in between 'G_' and '('.
13082 2018-05-17 Jakub Jelinek <jakub@redhat.com>
13085 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
13086 even if the mask is not all ones.
13089 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
13091 (ix86_gimple_fold_builtin): Likewise.
13094 * config/i386/i386.c: Include tree-vector-builder.h.
13095 (ix86_vector_shift_count): New function.
13096 (ix86_fold_builtin): Fold shift builtins by scalar count.
13097 (ix86_gimple_fold_builtin): Likewise.
13099 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
13100 _mm512_setzero): New intrinsics.
13102 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
13103 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13105 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
13106 code generation for cases where splatting a value is not useful.
13107 * simplify-rtx.c (simplify_ternary_operation): Simplify
13108 vec_merge across a vec_duplicate and a paradoxical subreg forming
13109 a vector mode to a vec_concat.
13111 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
13113 * config.gcc: Support "goldmont-plus".
13114 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
13116 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13117 PROCESSOR_GOLDMONT_PLUS.
13118 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
13119 (processor_target_table): Add "goldmont-plus".
13120 (PTA_GOLDMONT_PLUS): Define.
13121 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
13122 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
13123 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
13124 (fold_builtin_cpu): Add "goldmont-plus".
13125 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
13126 (ix86_option_override_internal): Add "goldmont-plus".
13127 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
13128 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
13129 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
13130 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
13132 2018-05-17 Richard Biener <rguenther@suse.de>
13134 PR tree-optimization/85757
13135 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
13136 remove defs that only feed that PHI from further processing.
13138 2018-05-16 Jim Wilson <jimw@sifive.com>
13140 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
13142 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
13144 2018-05-16 Mark Wielaard <mark@klomp.org>
13146 * dwarf2out.c (count_index_strings): New function.
13147 (output_indirect_strings): Call count_index_strings and generate
13148 header for dwarf_version >= 5.
13150 2018-05-16 Mark Wielaard <mark@klomp.org>
13152 * dwarf2out.c (dwarf_FORM): New function.
13153 (set_indirect_string): Use dwarf_FORM.
13154 (reset_indirect_string): Likewise.
13155 (size_of_die): Likewise.
13156 (value_format): Likewise.
13157 (output_die): Likewise.
13158 (add_skeleton_AT_string): Likewise.
13159 (output_macinfo_op): Likewise.
13160 (index_string): Likewise.
13161 (output_index_string_offset): Likewise.
13162 (output_index_string): Likewise.
13163 (count_index_strings): Likewise.
13165 2018-05-16 Carl Love <cel@us.ibm.com>
13167 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
13168 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
13170 2018-05-16 Martin Jambor <mjambor@suse.cz>
13172 * ipa-prop.c (ipa_free_all_edge_args): Remove.
13173 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
13175 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
13177 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
13178 (fnma<mode>4): Likewise.
13179 (fms<mode>4): Likewise.
13180 (fnms<mode>4): Likewise.
13181 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
13182 (aarch64_fnma<mode>4): Likewise.
13183 (aarch64_fms<mode>4): Likewise.
13184 (aarch64_fnms<mode>4): Likewise.
13185 (aarch64_fnmadd<mode>4): Likewise.
13187 2018-05-16 Jason Merrill <jason@redhat.com>
13189 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
13191 2018-05-16 Richard Biener <rguenther@suse.de>
13193 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
13194 (dump_stmt_cost): Declare.
13195 (add_stmt_cost): Dump cost we add.
13196 (add_stmt_costs): New function.
13197 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
13198 No longer exported.
13199 (vect_analyze_stmt): Adjust prototype.
13200 (vectorizable_condition): Likewise.
13201 (vectorizable_live_operation): Likewise.
13202 (vectorizable_reduction): Likewise.
13203 (vectorizable_induction): Likewise.
13204 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
13205 cost vector to pass to vectorizable_ and record afterwards.
13206 (vect_model_reduction_cost): Take cost vector argument and adjust.
13207 (vect_model_induction_cost): Likewise.
13208 (vectorizable_reduction): Likewise.
13209 (vectorizable_induction): Likewise.
13210 (vectorizable_live_operation): Likewise.
13211 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
13212 SLP_TREE_NUMBER_OF_VEC_STMTS.
13213 (vect_analyze_slp_cost_1): Remove.
13214 (vect_analyze_slp_cost): Likewise.
13215 (vect_slp_analyze_node_operations): Take visited args and
13216 a target cost vector. Avoid processing already visited stmt sets.
13217 (vect_slp_analyze_operations): Use a local cost vector to gather
13218 costs and register those of non-discarded instances.
13219 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
13220 (vect_schedule_slp_instance): Remove copying of
13221 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
13223 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
13224 adding cost. Record cost entry location.
13225 (vect_prologue_cost_for_slp_op): Function to compute cost of
13226 a constant or invariant generated for SLP vect in the prologue,
13227 split out from vect_analyze_slp_cost_1.
13228 (vect_model_simple_cost): Make static. Adjust for SLP costing.
13229 (vect_model_promotion_demotion_cost): Likewise.
13230 (vect_model_store_cost): Likewise, make static.
13231 (vect_model_load_cost): Likewise.
13232 (vectorizable_bswap): Add cost vector arg and adjust.
13233 (vectorizable_call): Likewise.
13234 (vectorizable_simd_clone_call): Likewise.
13235 (vectorizable_conversion): Likewise.
13236 (vectorizable_assignment): Likewise.
13237 (vectorizable_shift): Likewise.
13238 (vectorizable_operation): Likewise.
13239 (vectorizable_store): Likewise.
13240 (vectorizable_load): Likewise.
13241 (vectorizable_condition): Likewise.
13242 (vectorizable_comparison): Likewise.
13243 (can_vectorize_live_stmts): Likewise.
13244 (vect_analyze_stmt): Likewise.
13245 (vect_transform_stmt): Adjust calls to vectorizable_*.
13246 * tree-vectorizer.c: Include gimple-pretty-print.h.
13247 (dump_stmt_cost): New function.
13249 2018-05-16 Richard Biener <rguenther@suse.de>
13251 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
13252 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
13253 * tree-ssa-dse.c: Include tree-ssa-loop.h.
13254 (check_name): New callback.
13255 (dse_classify_store): Track cycles via a visited bitmap of PHI
13256 defs and simplify handling of in-loop and across loop dead stores
13257 and properly fail for loop-variant refs. Handle byte-tracking with
13258 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
13261 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
13263 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
13264 (vect_get_mask_type_for_stmt): Likewise.
13265 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
13267 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
13268 to determine the statement's vector type and the vector type that
13269 should be used for calculating nunits. Deal with cases in which
13270 the type has to be deferred.
13271 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
13272 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
13273 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
13274 (vect_determine_vf_for_stmt): New functions, split out from...
13275 (vect_determine_vectorization_factor): ...here.
13276 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
13277 (vect_get_mask_type_for_stmt): New functions, split out from
13278 vect_determine_vectorization_factor.
13280 2018-05-16 Richard Biener <rguenther@suse.de>
13282 * tree-cfg.c (verify_gimple_assign_ternary): Properly
13283 verify the [VEC_]COND_EXPR embedded comparison.
13285 2018-05-15 Martin Sebor <msebor@redhat.com>
13287 PR tree-optimization/85753
13288 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
13289 RECORD_TYPE in addition to ARRAY_TYPE.
13291 2018-05-15 Martin Sebor <msebor@redhat.com>
13293 PR middle-end/85643
13294 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
13296 2018-05-15 Richard Biener <rguenther@suse.de>
13298 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
13299 add by_clobber_p one. Change algorithm to collect all defs
13300 representing uses we need to walk and try reducing them to
13301 a single one before failing.
13302 (dse_dom_walker::dse_optimize_stmt): Adjust.
13304 2018-05-13 Mark Wielaard <mark@klomp.org>
13306 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
13307 (size_of_loc_descr): Likewise.
13308 (output_loc_operands): Likewise.
13309 (output_loc_operands_raw): Likewise.
13310 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
13311 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
13312 (hash_loc_operands): Likewise.
13313 (compare_loc_operands): Likewise.
13315 2018-05-14 Mark Wielaard <mark@klomp.org>
13317 * dwarf2out.c (count_index_addrs): New function.
13318 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
13320 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13322 PR tree-optimization/83648
13323 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
13324 return value as malloc candidate.
13326 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13329 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
13330 param as true in call to suggest_attribute.
13332 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
13334 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
13335 -mreadonly-in-sdata.
13337 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13339 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
13341 (aarch64_crypto_aesd_fused): Likewise.
13343 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
13345 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
13346 (movsi_aarch64): Likewise.
13347 (load_pairsi): Likewise.
13348 (load_pairdi): Likewise.
13349 (store_pairsi): Likewise.
13350 (store_pairdi): Likewise.
13351 (load_pairsf): Likewise.
13352 (load_pairdf): Likewise.
13353 (store_pairsf): Likewise.
13354 (store_pairdf): Likewise.
13355 (zero_extend): Likewise.
13356 (trunc): Swap alternatives.
13357 (fcvt_target): Add '?' to prefer w over r.
13359 2018-05-14 Jakub Jelinek <jakub@redhat.com>
13362 * config/i386/i386.md: Disallow non-commutative arithmetics in
13363 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
13364 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
13365 in the peephole2 before it.
13367 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
13369 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
13370 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
13371 (ix86_handle_option): Handle -mcldemote.
13372 * config.gcc: New header.
13373 * config/i386/cldemoteintrin.h: New file.
13374 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
13375 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
13377 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13378 OPTION_MASK_ISA_CLDEMOTE.
13379 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
13380 (ix86_valid_target_attribute_inner_p): Ditto.
13381 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
13382 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
13383 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
13384 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
13385 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
13387 * config/i386/i386.opt: Add -mcldemote.
13388 * config/i386/x86intrin.h: New header.
13389 * doc/invoke.texi: Add -mcldemote.
13391 2018-05-14 Richard Biener <rguenther@suse.de>
13393 * doc/match-and-simplify.texi: Adjust :s documentation.
13395 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
13397 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
13398 intended memcpy size.
13399 (REORDER_45): Likewise.
13401 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
13403 * sort.cc: New file.
13404 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
13405 * vec.c (qsort_chk): Use gcc_qsort.
13406 * Makefile.in (OBJS-libcommon): Add sort.o.
13407 (build/sort.o): New target. Use it...
13408 (BUILD_RTL): ... here, and...
13409 (build/gencfn-macros): ... here, and...
13410 (build/genmatch): ... here.
13412 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
13413 Chung-Ju Wu <jasonwucj@gmail.com>
13415 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
13416 * config/nds32/nds32-graywolf.md: New file.
13417 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
13418 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
13420 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
13421 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
13422 * config/nds32/nds32.md (pipeline_model): Add graywolf.
13423 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
13424 * config/nds32/pipelines.md: Include n15 settings.
13426 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
13427 Chung-Ju Wu <jasonwucj@gmail.com>
13429 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
13430 * config/nds32/nds32-n13.md: New file.
13431 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
13432 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
13434 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
13435 * config/nds32/nds32.md (pipeline_model): Add n13.
13436 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
13437 * config/nds32/pipelines.md: Include n13 settings.
13439 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
13440 Chung-Ju Wu <jasonwucj@gmail.com>
13442 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
13443 * config/nds32/nds32-n10.md: New file.
13444 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
13445 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
13447 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
13448 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
13449 * config/nds32/nds32.md (pipeline_model): Add n10.
13450 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
13451 * config/nds32/pipelines.md: Include n10 settings.
13453 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
13454 Kito Cheng <kito.cheng@gmail.com>
13455 Chung-Ju Wu <jasonwucj@gmail.com>
13457 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
13458 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
13459 Add enum values for DSP extension instructions.
13460 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
13462 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
13463 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
13464 New code iterators.
13465 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
13466 * config/nds32/nds32-dspext.md: New file for DSP implementation.
13467 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
13468 * config/nds32/nds32-intrinsic.md: Likewise.
13469 * config/nds32/nds32_intrinsic.h: Likewise.
13470 * config/nds32/nds32-md-auxiliary.c: Likewise.
13471 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
13472 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
13473 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
13474 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
13475 * config/nds32/nds32-protos.h: New declarations for DSP extension.
13476 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
13477 TYPE_DMAC in switch statement.
13478 * config/nds32/nds32.c: New checking and implementation for DSP
13479 extension instructions.
13480 * config/nds32/nds32.h: Likewise.
13481 * config/nds32/nds32.md: Likewise.
13482 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
13483 * config/nds32/predicates.md: Implement new predicates for DSP
13486 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
13488 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
13489 Reformat alternatives and attributes so it is easier to identify
13490 which constraints/attributes go with which instruction.
13491 (mov<mode>_hardfloat32, FMOVE64): Likewise.
13492 (mov<mode>_softfloat32, FMOVE64): Likewise.
13493 (mov<mode>_hardfloat64, FMOVE64): Likewise.
13494 (mov<mode>_softfloat64, FMOVE64): Likewise.
13496 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
13498 * doc/extend.texi (PowerPC Built-in Functions): Rename this
13500 (Basic PowerPC Built-in Functions): The new name of the
13501 subsection previously known as "PowerPC Built-in Functions".
13502 (Basic PowerPC Built-in Functions Available on all Configurations):
13504 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
13505 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
13506 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
13507 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
13509 2018-05-11 Martin Jambor <mjambor@suse.cz>
13512 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
13515 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
13518 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
13520 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
13522 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
13523 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
13524 (ix86_handle_option): Handle -mwaitpkg.
13525 * config.gcc: New header.
13526 * config/i386/cpuid.h (bit_WAITPKG): New bit.
13527 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
13528 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
13530 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13531 OPTION_MASK_ISA_WAITPKG.
13532 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
13533 (ix86_option_override_internal): Add PTA_WAITPKG.
13534 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
13535 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
13536 IX86_BUILTIN_TPAUSE.
13537 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
13538 __builtin_ia32_umwait and __builtin_ia32_tpause.
13539 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
13540 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
13541 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
13542 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
13543 UNSPECV_TPAUSE): New.
13544 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
13545 * config/i386/i386.opt: Add -mwaitpkg.
13546 * config/i386/waitpkgintrin.h: New file.
13547 * config/i386/x86intrin.h: New header.
13548 * doc/invoke.texi: Add -mwaitpkg.
13550 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
13553 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
13555 (cortex-m0): Use armv6s-m isa.
13556 (cortex-m0plus): Likewise.
13557 (cortex-m1): Likewise.
13558 (cortex-m0.small-multiply): Likewise.
13559 (cortex-m0plus.small-multiply): Likewise.
13560 (cortex-m1.small-multiply): Likewise.
13562 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
13563 Jakub Jelinek <jakub@redhat.com>
13565 PR tree-optimization/85692
13566 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
13567 source permute as well.
13569 2018-05-11 Martin Liska <mliska@suse.cz>
13572 * doc/extend.texi: Document LLVM style format for no_sanitize
13575 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
13577 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
13578 mode_supports_vsx_dform_quad to mode_supports_dq_form.
13579 (mode_supports_vsx_dform_quad): Likewise.
13580 (mode_supports_vmx_dform): Move these functions to be next to the
13581 other mode_supports functions.
13582 (mode_supports_dq_form): Likewise.
13583 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
13584 mode_supports_dq_form.
13585 (reg_offset_addressing_ok_p): Likewise.
13586 (offsettable_ok_by_alignment): Likewise.
13587 (rs6000_legitimate_offset_address_p): Likewise.
13588 (legitimate_lo_sum_address_p): Likewise.
13589 (rs6000_legitimize_address): Likewise.
13590 (rs6000_legitimize_reload_address): Likewise.
13591 (rs6000_secondary_reload_inner): Likewise.
13592 (rs6000_preferred_reload_class): Likewise.
13593 (rs6000_output_move_128bit): Likewise.
13595 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
13597 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
13598 Generate SImode target register for null target.
13599 <case IX86_BUILTIN_XGETBV>: Ditto.
13600 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
13601 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
13603 2018-05-10 Carl Love <cel@us.ibm.com>
13605 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
13606 dcbtt and dcbtstt if operands[2] is 0.
13608 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
13611 * config/i386/sse.md (usadv64qi): New expander.
13613 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
13615 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
13616 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
13617 -maltivec=be support.
13618 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
13619 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
13620 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
13621 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
13622 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
13623 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
13624 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
13625 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
13626 altivec_vsumsws): Adjust.
13627 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
13628 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
13629 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
13631 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
13632 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
13633 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
13634 (altivec_lve<VI_char>x): Delete expand.
13635 (*altivec_lve<VI_char>x_internal): Rename to...
13636 (altivec_lve<VI_char>x): ... this.
13637 (altivec_lvxl_<mode>): Delete expand.
13638 (*altivec_lvxl_<mode>_internal): Rename to ...
13639 (altivec_lvxl_<mode>): ... this.
13640 (altivec_stvxl_<mode>): Delete expand.
13641 (*altivec_stvxl_<mode>_internal): Rename to ...
13642 (altivec_stvxl_<mode>): ... this.
13643 (altivec_stve<VI_char>x): Delete expand.
13644 (*altivec_stve<VI_char>x_internal): Rename to ...
13645 (altivec_stve<VI_char>x): ... this.
13646 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
13647 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
13648 reduc_plus_scal_<mode>): Adjust.
13649 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
13651 (rs6000_cpu_cpp_builtins): Adjust.
13652 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
13653 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
13654 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
13655 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
13656 -maltivec=be support.
13657 (rs6000_split_vec_extract_var): Adjust.
13658 (rs6000_split_v4si_init): Adjust.
13659 (swap_selector_for_mode): Delete.
13660 (altivec_expand_lvx_be, altivec_expand_stvx_be,
13661 altivec_expand_stvex_be): Delete.
13662 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
13663 -maltivec=be support.
13664 (rs6000_gimple_fold_builtin): Ditto.
13665 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
13667 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
13668 (TARGET_DIRECT_MOVE_64BIT): Adjust.
13669 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
13670 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
13671 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
13672 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
13673 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
13674 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
13675 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
13676 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
13677 anonymous split): Adjust.
13678 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
13679 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
13681 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
13683 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
13684 when --with-gxx-include-dir is also specified.
13685 * configure: Regenerate.
13687 2018-05-09 Jim Wilson <jimw@sifive.com>
13690 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
13691 * config/riscv/t-withmultilib: New.
13692 * config/riscv/withmultilib.h: New.
13693 * doc/install.texi: Document RISC-V --with-multilib-list support.
13695 2018-05-09 Richard Biener <rguenther@suse.de>
13697 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
13699 (vect_bb_vectorization_profitable_p): Adjust. Compute
13700 actual scalar cost using the cost vector and the add_stmt_cost
13703 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13705 PR rtl-optimization/85645
13706 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
13707 in the REG_CFA_REGISTER note for LR, don't leave it empty.
13709 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13711 PR rtl-optimization/85645
13712 * shrink-wrap.c (spread_components): Return a boolean saying if
13713 anything was changed.
13714 (try_shrink_wrapping_separate): Iterate spread_components until
13715 nothing changes anymore.
13717 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13719 PR rtl-optimization/85645
13720 * regrename.c (build_def_use): Also kill the chains that include the
13721 destination of a REG_CFA_REGISTER note.
13723 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13725 PR rtl-optimization/85645
13726 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
13727 insn that has a REG_CFA_REGISTER note.
13729 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
13731 * cfgexpand.c (expand_clobber): New function.
13732 (expand_gimple_stmt_1): Use it.
13733 * tree-vect-stmts.c (vect_clobber_variable): New function,
13735 (vectorizable_simd_clone_call): ...here.
13736 (vectorizable_store): Emit a clobber either side of an
13737 IFN_STORE_LANES sequence.
13738 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
13740 2018-05-09 Tom de Vries <tom@codesourcery.com>
13743 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
13744 (define_insn "trap_if_false"): Add exit after trap.
13746 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
13748 PR rtl-optimization/85638
13749 * bb-reorder.c: Include common/common-target.h.
13750 (create_forwarder_block): New function extracted from...
13751 (fix_up_crossing_landing_pad): ...here. Rename into...
13752 (dw2_fix_up_crossing_landing_pad): ...this.
13753 (sjlj_fix_up_crossing_landing_pad): New function.
13754 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
13755 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
13756 from both partitions and exit the loop after one iteration.
13758 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
13761 * doc/extend.texi (PowerPC Built-in Functions): Rename this
13763 (Basic PowerPC Built-in Functions): The new name of the
13764 subsection previously known as "PowerPC Built-in Functions".
13765 (Basic PowerPC Built-in Functions Available on all Configurations):
13767 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
13769 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
13771 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
13773 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
13776 2018-05-08 Jim Wilson <jimw@sifive.com>
13778 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
13779 (LD_EMUL_SUFFIX): New.
13780 (LINK_SPEC): Use it.
13782 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
13784 * doc/extend.texi (PowerPC Built-in Functions): Rename this
13786 (Basic PowerPC Built-in Functions): The new name of the
13787 subsection previously known as "PowerPC Built-in Functions".
13788 (Basic PowerPC Built-in Functions Available on all Configurations):
13790 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
13792 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
13794 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
13796 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
13799 2018-05-08 Jakub Jelinek <jakub@redhat.com>
13802 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
13803 after cmpelim optimization.
13805 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
13807 * config.gcc: Support "goldmont".
13808 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
13809 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13810 PROCESSOR_GOLDMONT.
13811 * config/i386/i386.c (m_GOLDMONT): Define.
13812 (processor_target_table): Add "goldmont".
13813 (PTA_GOLDMONT): Define.
13814 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
13815 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
13816 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
13817 (fold_builtin_cpu): Add "goldmont".
13818 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
13819 (ix86_option_override_internal): Add "goldmont".
13820 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
13821 (processor_type): Add PROCESSOR_GOLDMONT.
13822 * config/i386/i386.md: Add CPU "glm".
13823 * config/i386/glm.md: New file.
13824 * config/i386/x86-tune.def: Add m_GOLDMONT.
13825 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
13827 2018-05-08 Jakub Jelinek <jakub@redhat.com>
13830 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
13832 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
13833 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
13834 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
13837 * config/i386/i386.c (ix86_fold_builtin): Handle
13838 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
13841 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
13842 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
13844 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
13847 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
13848 (check_arch): Likewise.
13849 (check_fpu): Return the result rather than printing it.
13850 (end arch): Fix operator precedence.
13851 (end cpu): Likewise.
13852 (END): Print the result from check_fpu.
13854 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
13855 Alan Hayward <alan.hayward@arm.com>
13856 David Sherwood <david.sherwood@arm.com>
13858 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
13859 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
13860 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
13861 (*fcmuo<mode>_and): New patterns.
13863 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
13865 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
13866 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
13867 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
13868 (cmp_op, sve_imm_con): New code attributes.
13869 (SVE_COND_INT_CMP, imm_con): Delete.
13870 (cmp_op): Remove above unspecs from int attribute.
13871 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
13873 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
13874 comparison-specific unspecs.
13875 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
13876 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
13877 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
13878 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
13879 (*vec_fcm<cmp_op><mode>): Rename to...
13880 (*fcm<cmp_op><mode>): ...this and adjust likewise.
13881 (*vec_fcmuo<mode>): Rename to...
13882 (*fcmuo<mode>): ...this and adjust likewise.
13883 (*pred_fcm<cmp_op><mode>): New pattern.
13884 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
13885 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
13887 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
13889 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
13890 (aarch64_emit_sve_predicated_cond): New function.
13891 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
13892 (aarch64_emit_unspec_cond_or): Replace with...
13893 (aarch64_emit_sve_or_conds): ...this new function. Use
13894 aarch64_emit_sve_ptrue_op for the individual comparisons and
13895 aarch64_emit_binop to OR them together.
13896 (aarch64_emit_inverted_unspec_cond): Replace with...
13897 (aarch64_emit_sve_inverted_cond): ...this new function. Use
13898 aarch64_emit_sve_ptrue_op for the comparison and
13899 aarch64_emit_unop to invert the result.
13900 (aarch64_expand_sve_vec_cmp_float): Update after the above
13901 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
13903 2018-05-07 Nathan Sidwell <nathan@acm.org>
13905 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
13906 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
13907 (Backwards Compatibility): Likewise.
13909 2018-05-07 Luis Machado <luis.machado@linaro.org>
13913 2018-05-07 Luis Machado <luis.machado@linaro.org>
13915 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13916 <prefetch_dynamic_strides>: New const bool field.
13917 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13918 prefetch_dynamic_strides.
13919 (exynosm1_prefetch_tune): Likewise.
13920 (thunderxt88_prefetch_tune): Likewise.
13921 (thunderx_prefetch_tune): Likewise.
13922 (thunderx2t99_prefetch_tune): Likewise.
13923 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
13925 (aarch64_override_options_internal): Update to set
13926 PARAM_PREFETCH_DYNAMIC_STRIDES.
13927 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
13928 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
13929 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
13930 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
13931 prefetch-dynamic-strides setting.
13933 2018-05-07 Luis Machado <luis.machado@linaro.org>
13935 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13936 <minimum_stride>: New const int field.
13937 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13938 minimum_stride field.
13939 (exynosm1_prefetch_tune): Likewise.
13940 (thunderxt88_prefetch_tune): Likewise.
13941 (thunderx_prefetch_tune): Likewise.
13942 (thunderx2t99_prefetch_tune): Likewise.
13943 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
13944 (aarch64_override_options_internal): Update to set
13945 PARAM_PREFETCH_MINIMUM_STRIDE.
13946 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
13947 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
13948 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
13949 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
13950 stride is constant and is below the minimum stride threshold.
13952 2018-05-07 Luis Machado <luis.machado@linaro.org>
13954 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
13957 2018-05-07 Luis Machado <luis.machado@linaro.org>
13959 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13960 <prefetch_dynamic_strides>: New const bool field.
13961 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13962 prefetch_dynamic_strides.
13963 (exynosm1_prefetch_tune): Likewise.
13964 (thunderxt88_prefetch_tune): Likewise.
13965 (thunderx_prefetch_tune): Likewise.
13966 (thunderx2t99_prefetch_tune): Likewise.
13967 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
13969 (aarch64_override_options_internal): Update to set
13970 PARAM_PREFETCH_DYNAMIC_STRIDES.
13971 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
13972 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
13973 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
13974 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
13975 prefetch-dynamic-strides setting.
13977 2018-05-07 Luis Machado <luis.machado@linaro.org>
13979 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13980 <minimum_stride>: New const int field.
13981 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13982 minimum_stride field.
13983 (exynosm1_prefetch_tune): Likewise.
13984 (thunderxt88_prefetch_tune): Likewise.
13985 (thunderx_prefetch_tune): Likewise.
13986 (thunderx2t99_prefetch_tune): Likewise.
13987 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
13988 (aarch64_override_options_internal): Update to set
13989 PARAM_PREFETCH_MINIMUM_STRIDE.
13990 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
13991 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
13992 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
13993 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
13994 stride is constant and is below the minimum stride threshold.
13996 2018-05-06 Jakub Jelinek <jakub@redhat.com>
13999 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
14000 the type is addressable. Don't force op into register if it has
14003 2018-05-05 Roland McGrath <mcgrathr@google.com>
14006 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
14007 any section for which we don't know a specific type it should have,
14008 regardless of name. Previously this was done only for the exact
14009 names ".init_array", ".fini_array", and ".preinit_array".
14010 (default_elf_asm_named_section): Add comment about
14011 relationship with default_section_type_flags and SECTION_NOTYPE.
14012 (get_section): Don't consider it a type conflict if one side has
14013 SECTION_NOTYPE and the other doesn't, as long as neither has the
14014 SECTION_BSS et al used in the default_section_type_flags logic.
14016 2018-05-05 Tom de Vries <tom@codesourcery.com>
14019 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
14020 (workaround_barsyncs): New function.
14021 (nvptx_reorg): Use workaround_barsyncs.
14022 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
14023 (define_expand "nvptx_membar_cta"): New define_expand.
14024 (define_insn "*nvptx_membar_cta"): New insn.
14026 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
14028 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
14029 To improve optimization opportunities.
14030 * builtin-types.def: The new needed builtin types for the above.
14032 2018-05-04 Richard Biener <rguenther@suse.de>
14034 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
14035 * gimple-ssa-store-merging.c
14036 (imm_store_chain_info::output_merged_store): Remove redundant create,
14037 release split_store vector contents on failure.
14038 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
14039 scalar stmt vector on cache hit.
14041 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
14043 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
14045 * config.gcc (powerpc-xilinx-eabi*): Remove.
14046 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
14048 (fusion_addis_mem_combo_load): Ditto.
14049 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
14051 (rs6000_cpu_cpp_builtins): Ditto.
14052 * config/rs6000/rs6000-linux.c
14053 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
14054 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
14055 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
14057 (rs6000_setup_reg_addr_masks): Ditto.
14058 (rs6000_init_hard_regno_mode_ok): Ditto.
14059 (rs6000_option_override_internal): Ditto.
14060 (legitimate_lo_sum_address_p): Ditto.
14061 (rs6000_legitimize_address): Ditto.
14062 (rs6000_legitimize_reload_address): Ditto.
14063 (rs6000_legitimate_address_p): Ditto.
14064 (abi_v4_pass_in_fpr): Ditto.
14065 (setup_incoming_varargs): Ditto.
14066 (rs6000_gimplify_va_arg): Ditto.
14067 (rs6000_split_multireg_move): Ditto.
14068 (rs6000_savres_strategy): Ditto.
14069 (rs6000_emit_prologue_components): Ditto.
14070 (rs6000_emit_epilogue_components): Ditto.
14071 (rs6000_emit_prologue): Ditto.
14072 (rs6000_emit_epilogue): Ditto.
14073 (rs6000_elf_file_end): Ditto.
14074 (rs6000_function_value): Ditto.
14075 (rs6000_libcall_value): Ditto.
14076 * config/rs6000/rs6000.h: Ditto.
14077 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
14078 (TARGET_MINMAX): ... this. New.
14079 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
14080 * config/rs6000/rs6000.md: Remove Xilinx FP support.
14081 (*movsi_internal1_single): Delete.
14082 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
14083 mfpu=, mxilinx-fpu): Delete.
14084 * config/rs6000/singlefp.h: Delete.
14085 * config/rs6000/sysv4.h: Remove Xilinx FP support.
14086 * config/rs6000/t-rs6000: Ditto.
14087 * config/rs6000/t-xilinx: Delete.
14088 * config/rs6000/titan.md: Adjust for fp_type removal.
14089 * config/rs6000/vsx.md: Remove Xilinx FP support.
14090 (VStype_simple): Delete.
14091 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
14092 * config/rs6000/xfpu.h: Delete.
14093 * config/rs6000/xfpu.md: Delete.
14094 * config/rs6000/xilinx.h: Delete.
14095 * config/rs6000/xilinx.opt: Delete.
14096 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
14097 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
14099 2018-05-04 Tom de Vries <tom@codesourcery.com>
14102 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
14105 2018-05-04 Richard Biener <rguenther@suse.de>
14107 PR middle-end/85627
14108 * tree-complex.c (update_complex_assignment): We are always in SSA form.
14109 (expand_complex_div_wide): Likewise.
14110 (expand_complex_operations_1): Likewise.
14111 (expand_complex_libcall): Preserve EH info of the original stmt.
14112 (tree_lower_complex): Handle removed blocks.
14113 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
14114 on complex multiplication and division libcall builtins.
14116 2018-05-04 Richard Biener <rguenther@suse.de>
14118 PR middle-end/85574
14119 * fold-const.c (negate_expr_p): Restrict negation of operand
14120 zero of a division to when we know that can happen without
14122 (fold_negate_expr_1): Likewise.
14124 2018-05-04 Jakub Jelinek <jakub@redhat.com>
14127 * real.h (real_nextafter): Declare.
14128 * real.c (real_nextafter): New function.
14129 * fold-const-call.c (fold_const_nextafter): New function.
14130 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
14131 CASE_CFN_NEXTTOWARD.
14132 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
14133 even when arg1_mode is different from arg0_mode.
14135 2018-05-03 Nathan Sidwell <nathan@acm.org>
14137 * doc/extend.texi (Deprecated Features): Remove
14138 -ffriend-injection.
14139 (Backwards Compatibility): Likewise.
14140 * doc/invoke.texi (C++ Language Options): Likewise.
14141 (C++ Dialect Options): Likewise.
14143 2018-05-03 Jakub Jelinek <jakub@redhat.com>
14146 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
14147 _mm512_mask_mullox_epi64): New intrinsics.
14149 2018-05-03 Tom de Vries <tom@codesourcery.com>
14152 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
14153 dump files): Add offload-tree.
14155 2018-05-03 Richard Biener <rguenther@suse.de>
14157 PR tree-optimization/85615
14158 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
14159 to loops not nested in BBs loop father to avoid creating multi-entry
14162 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14164 PR tree-optimization/70291
14165 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
14166 arguments. Change return type to tree. Emit libcall as a new
14167 statement rather than replacing existing one when inplace_p is true.
14168 (expand_complex_multiplication_components): New function.
14169 (expand_complex_multiplication): Expand floating-point complex
14170 multiplication using the above.
14171 (expand_complex_division): Rename inner_type parameter to type.
14172 Update expand_complex_libcall call-site.
14173 (expand_complex_operations_1): Update expand_complex_multiplication
14174 and expand_complex_division call-sites.
14176 2018-05-02 Jakub Jelinek <jakub@redhat.com>
14179 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
14180 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
14181 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
14182 the highest significant bit of the shift count mask is clear. In
14183 check whether and[sq]i3 is needed verify that all significant bits
14184 of the shift count other than the highest are set.
14186 2018-05-02 Tom de Vries <tom@codesourcery.com>
14189 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
14190 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
14191 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
14192 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
14193 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
14194 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
14195 __builtin_goacc_parlevel_size.
14197 2018-05-02 Richard Biener <rguenther@suse.de>
14199 PR tree-optimization/85597
14200 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
14201 do not use split vect_get_vec_defs call but call vect_get_slp_defs
14204 2018-05-02 Tom de Vries <tom@codesourcery.com>
14207 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
14208 dump files): Add ltrans-tree.
14210 2018-05-02 Tom de Vries <tom@codesourcery.com>
14213 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
14214 dump files): Add wpa-ipa.
14216 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
14218 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
14219 powerpc*-*-linux*paired* target.
14220 * config/rs6000/750cl.h: Delete.
14221 * config/rs6000/paired.h: Delete.
14222 * config/rs6000/paired.md: Delete.
14223 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
14225 * config/rs6000/rs6000-builtin.def: Remove paired float support.
14226 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
14227 comment. Remove paired float support.
14228 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
14229 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
14231 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
14232 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
14234 * config/rs6000/rs6000.c: Remove paired float support.
14235 (paired_expand_vector_init, paired_expand_vector_move,
14236 paired_emit_vector_compare, paired_emit_vector_cond_expr,
14237 (paired_expand_lv_builtin, paired_expand_stv_builtin,
14238 paired_expand_builtin, paired_expand_predicate_builtin,
14239 paired_init_builtins): Delete.
14240 * config/rs6000/rs6000.h: Remove paired float support.
14241 * config/rs6000/rs6000.md: Remove paired float support.
14242 (move_from_CR_ov_bit): Delete.
14243 * config/rs6000/rs6000.opt (mpaired): Delete.
14244 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
14245 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
14247 2018-05-02 Richard Biener <rguenther@suse.de>
14249 PR middle-end/85567
14250 * gimplify.c (gimplify_save_expr): When in SSA form allow
14251 SAVE_EXPRs to compute to SSA vars.
14253 2018-05-02 Jakub Jelinek <jakub@redhat.com>
14256 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
14257 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
14258 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
14259 clobber operands[2], instead use a new pseudo. Formatting fixes.
14261 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
14263 PR tree-optimization/85586
14264 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
14265 exit early for statements in the same group if the accesses are
14268 2018-05-02 Tom de Vries <tom@codesourcery.com>
14271 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
14274 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
14276 PR tree-optimization/85143
14277 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
14279 2018-05-01 Tom de Vries <tom@codesourcery.com>
14282 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
14283 not found" error message.
14285 2018-05-01 Tom de Vries <tom@codesourcery.com>
14288 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
14289 * vec.c (test_ordered_remove_if): New function.
14290 (vec_c_tests): Call test_ordered_remove_if.
14291 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
14292 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
14293 * tree-vect-patterns.c (vect_pattern_recog_1): Use
14294 VEC_ORDERED_REMOVE_IF.
14296 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14298 PR tree-optimization/82665
14299 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
14300 pointer subtraction where arguments come from a memchr call.
14302 2018-05-01 Jakub Jelinek <jakub@redhat.com>
14304 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
14305 --push-state --as-needed and --pop-state instead of --as-needed and
14306 --no-as-needed if ld supports it.
14307 * configure: Regenerated.
14310 * doc/install.texi2html: Replace _002d with - and _002a with * in
14311 generated html files using sed.
14313 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14316 * gcc-rich-location.c (blank_line_before_p): New function.
14317 (use_new_line): New function.
14318 (gcc_rich_location::add_fixit_insert_formatted): New function.
14319 * gcc-rich-location.h
14320 (gcc_rich_location::add_fixit_insert_formatted): New function.
14322 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14324 * selftest.c (assert_streq): Rename "expected" and "actual" to
14325 "val1" and "val2". Extend NULL-handling to cover both inputs
14326 symmetrically, while still requiring both to be non-NULL for a pass.
14327 * selftest.h (assert_streq): Rename "expected" and "actual" to
14329 (ASSERT_EQ): Likewise.
14330 (ASSERT_EQ_AT): Likewise.
14331 (ASSERT_KNOWN_EQ): Likewise.
14332 (ASSERT_KNOWN_EQ_AT): Likewise.
14333 (ASSERT_NE): Likewise.
14334 (ASSERT_MAYBE_NE): Likewise.
14335 (ASSERT_MAYBE_NE_AT): Likewise.
14336 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
14337 the assertion to pass.
14338 (ASSERT_STREQ_AT): Likewise.
14340 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
14342 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
14343 interaction with -pie.
14345 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14347 * selftest.h: Fix alphabetization of per-source-file selftest
14350 2018-04-30 Jason Merrill <jason@redhat.com>
14352 PR c++/61982 - dead stores to destroyed objects.
14353 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
14356 2018-04-30 Jason Merrill <jason@redhat.com>
14358 * tree.c (build_clobber): New.
14359 * tree.h: Declare it.
14360 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
14362 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14364 * diagnostic-show-locus.c (layout::layout): Update for
14365 location_get_source_line returning a char_span.
14366 (struct char_span): Move to input.h.
14367 (struct correction): Update for fields in char_span becoming
14369 (struct source_line): Update for location_get_source_line
14370 returning a char_span.
14371 (layout::print_line): Likewise.
14372 * edit-context.c (edited_file::print_content): Likewise.
14373 (edited_file::print_diff_hunk): Likewise.
14374 (edited_file::print_run_of_changed_lines): Likewise.
14375 (edited_file::get_num_lines): Likewise.
14376 (edited_line::edited_line): Likewise.
14377 * final.c (asm_show_source): Likewise.
14378 * input.c (location_get_source_line): Convert return type
14379 from const char * to char_span, losing the final "line_len"
14381 (dump_location_info): Update for the above.
14382 (get_substring_ranges_for_loc): Likewise. Use a char_span
14383 when handling the literal within the line.
14384 (test_reading_source_line): Update for location_get_source_line
14385 returning a char_span.
14386 * input.h (class char_span): Move here from
14387 diagnostic-show-locus.c, converting from a struct to a class.
14388 Make data members private.
14389 (char_span::operator bool): New.
14390 (char_span::length): New.
14391 (char_span::get_buffer): New.
14392 (char_span::operator[]): New.
14393 (char_span::subspan): Make const.
14394 (char_span::xstrdup): New.
14395 (location_get_source_line): Convert return type from const char *
14396 to char_span, losing the final "line_size" param.
14398 2018-04-30 Jan Hubicka <jh@suse.cz>
14400 * lto-wrapper.c (ltrans_priorities): New static var.
14401 (cmp_priority): New.
14402 (run_gcc): Read priorities and if doing parallel build order
14403 the Makefile by them.
14405 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14407 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
14409 2018-04-30 Richard Biener <rguenther@suse.de>
14411 * tree-cfg.c (verify_address): Remove base argument, add
14412 flag whether to check TREE_ADDRESSABLE and do that.
14413 (verify_expr): Remove.
14414 (verify_types_in_gimple_reference): Add pieces from verify_expr.
14415 (verify_gimple_assign_single): Likewise.
14416 (verify_gimple_switch): Likewise.
14417 (verify_expr_location_1): Dereference tp once. Add (disabled)
14418 piece from verify_expr.
14419 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
14421 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
14423 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
14425 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
14427 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
14428 (small_data_pattern): Likewise.
14429 (arc_rewrite_small_data): Likewise.
14430 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
14431 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
14432 (get_symbol_alignment): New function.
14433 (legitimate_small_data_address_p): Likewise.
14434 (legitimate_scaled_address): Update, call
14435 legitimate_small_data_address_p.
14436 (output_sdata): New static variable.
14437 (arc_print_operand): Update how we handle small data operands.
14438 (arc_print_operand_address): Likewise.
14439 (arc_legitimate_address_p): Update, use
14440 legitimate_small_data_address_p.
14441 (arc_rewrite_small_data_p): Remove.
14442 (arc_rewrite_small_data_1): Likewise.
14443 (arc_rewrite_small_data): Likewise.
14444 (small_data_pattern): Likewise.
14445 (compact_sda_memory_operand): Update to use
14446 legitimate_small_data_address_p and get_symbol_alignment.
14447 (prepare_move_operands): Don't rewite sdata pattern.
14448 (prepare_extend_operands): Remove.
14449 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
14451 (zero_extendqisi2): Likewise.
14452 (zero_extendhisi2): Likewise.
14453 (extendqihi2): Likewise.
14454 (extendqisi2): Likewise.
14455 (extendhisi2): Likewise.
14456 (addsi3): Likewise.
14457 (subsi3): Likewise.
14458 (andsi3): Likewise.
14459 * config/arc/constraints.md (Usd): Change it to memory constraint.
14461 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
14463 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
14464 as source of std instructions.
14465 * config/arc/arc.md (movsi_insn): Update pattern predicate to
14466 allow 6-bit constants as source for store instructions.
14467 (movdi_insn): Update instruction pattern to allow 6-bit constants
14468 as source for store instructions.
14470 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
14472 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
14474 2018-04-30 Nathan Sidwell <nathan@acm.org>
14475 Sandra Loosemore <sandra@codesourcery.com>
14477 * dumpfile.c (dump_open): Allow '-' for stdout.
14478 * doc/invoke.texi (Developer Options): Document dump filename
14479 determination early. Document stdin/stdout selection.
14481 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
14483 Microblaze Target: PIC data text relative
14485 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
14486 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
14488 * config/microblaze/microblaze.h (microblaze_constant_address_p):
14489 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
14490 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
14491 New addressing mode for data-text relative position indepenedent code.
14492 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
14493 'ADDRESS_SYMBOLIC_TXT_REL'.
14494 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
14495 (microblaze_legitimate_pic_operand): Exclude function calls from
14496 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
14497 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
14499 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
14500 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
14501 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
14502 for 'address + offset'.
14503 (microblaze_expand_prologue): Add new function prologue call for
14505 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
14506 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
14507 table in case of TARGET_PIC_DATA_TEXT_REL.
14508 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
14509 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
14510 Add new macros 'UNSPEC_TEXT',
14511 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
14512 + exclude function calls from 'UNSPEC_PLT' in case of data text
14514 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
14515 new target hook for generating address diff vector tables in case of
14517 * doc/tm.texi : Regenerate.
14518 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
14519 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
14520 of addr diff vector generation.
14521 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
14522 target hook definition.
14523 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
14524 Add default function for generate_pic_addr_diff_vec -> flag_pic.
14525 * doc/invoke.texi (Add new pic option): Add new microblaze pic
14526 option for data text relative.
14528 2018-04-30 Richard Biener <rguenther@suse.de>
14530 * tree-chrec.h (evolution_function_is_constant_p): Remove
14532 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
14534 2018-04-30 Richard Biener <rguenther@suse.de>
14537 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
14539 2018-04-30 Richard Biener <rguenther@suse.de>
14541 PR tree-optimization/28364
14542 PR tree-optimization/85275
14543 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
14544 copying first exit test.
14546 2018-04-28 Mark Wielaard <mark@klomp.org>
14548 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
14549 dwarf_version >= 5.
14550 (dwarf_AT): Handle DW_AT_addr_base.
14551 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
14553 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
14556 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
14557 (*ashl<dwi>3_doubleword_mask_1): Ditto.
14558 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
14559 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
14561 2018-04-28 Richard Biener <rguenther@suse.de>
14563 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
14564 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
14565 to reflect use. Only add interesting stmts.
14567 2018-04-27 Martin Jambor <mjambor@suse.cz>
14570 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
14571 the jump function allows for passing through aggregate values.
14573 2018-04-27 David Malcolm <dmalcolm@redhat.com>
14575 * input.h (in_system_header_at): Convert from macro to inline
14577 (from_macro_expansion_at): Likewise.
14578 (from_macro_definition_at): Likewise.
14580 2018-04-27 Jeff Law <law@redhat.com>
14582 * config.gcc: Mark tile* targets as deprecated/obsolete.
14584 2018-04-27 Richard Biener <rguenther@suse.de>
14586 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
14589 2018-04-27 Richard Biener <rguenther@suse.de>
14591 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
14593 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
14595 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
14596 with Yd constraint. Set "preferred_for_speed" attribute from
14597 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
14598 with Yd constraint.
14599 (*movdi_internal): Ditto.
14600 (movti_interunit splitters): Remove
14601 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
14602 (movdi_interunit splitters): Ditto.
14603 * config/i386/constraints.md (Ye): Remove.
14604 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
14606 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14609 * config/aarch64/constraints.md (Usg): Limit to 31.
14610 (Usj): Limit to 63.
14612 2018-04-27 Jakub Jelinek <jakub@redhat.com>
14614 PR tree-optimization/85529
14615 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
14616 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
14617 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
14618 zero extension or masking of the MSB bit.
14619 (optimize_range_tests): Add FIRST_BB argument, pass it through
14620 to optimize_range_tests_var_bound.
14621 (maybe_optimize_range_tests, reassociate_bb): Adjust
14622 optimize_range_tests callers.
14624 2018-04-26 Richard Biener <rguenther@suse.de>
14625 Jakub Jelinek <jakub@redhat.com>
14627 * cgraph.h (symbol_table): Just declare debug method here.
14628 * symtab.c (symbol_table::debug): Define.
14630 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
14632 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
14634 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
14636 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
14637 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
14638 (*movdi_internal): Substitute Yi and Yj constraint with x
14639 and Ym and Yn constraint with y constraint. Update "isa"
14640 attribute and set "preferred_for_speed" attribute from
14641 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
14642 (*movsi_internal): Ditto.
14643 (*movdf_internal): Ditto.
14644 (*movsf_internal): Ditto.
14645 (*zero_extendsidi2): Ditto.
14646 * config/i386/sse.md (vec_set<mode>_0): Ditto.
14647 (sse2_loadld): Ditto.
14648 (*vec_extract<ssevecmodelower>_0): Ditto.
14649 (*vec_extractv4si_0_zext_sse4): Ditto.
14650 (vec_concatv2di): Ditto.
14651 (*vec_dup<mode>): Ditto.
14652 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
14653 * config/i386/constraints.md (Yi): Remove.
14658 2018-04-26 Nathan Sidwell <nathan@acm.org>
14660 * dumpfile.c (dump_open): New.
14661 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
14662 (dump_finish): Detect stdio/stderr by value not name.
14664 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
14666 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
14668 2018-04-26 Tom de Vries <tom@codesourcery.com>
14671 * config/nvptx/nvptx.c (verify_neutering_jumps)
14672 (verify_neutering_labels): New function
14673 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
14675 2018-04-26 Tom de Vries <tom@codesourcery.com>
14678 * config/nvptx/nvptx.c (needs_neutering_p): New function.
14679 (nvptx_single): Use needs_neutering_p to skip over insns that do not
14682 2018-04-26 Richard Biener <rguenther@suse.de>
14683 Tom de Vries <tom@codesourcery.com>
14686 * lto-streamer-out.c (output_function): Fixup loops if required to match
14687 discovery done in the reader.
14689 2018-04-26 Richard Biener <rguenther@suse.de>
14691 PR tree-optimization/85116
14692 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
14693 have a loop exit from the single latch predecessor. Remove
14694 case of header with just condition.
14695 (ch_base::copy_headers): Exclude infinite loops from any
14697 (pass_ch::execute): Record exits.
14699 2018-04-26 Richard Biener <rguenther@suse.de>
14701 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
14702 prologue cost vector and pass it to vect_get_load_cost.
14703 (vect_get_peeling_costs_all_drs): Likewise.
14704 (vect_peeling_hash_get_lowest_cost): Likewise.
14705 (vect_enhance_data_refs_alignment): Likewise.
14707 2018-04-26 Richard Biener <rguenther@suse.de>
14709 PR middle-end/85450
14710 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
14711 checking of integer<->pointer conversions.
14712 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
14713 sign-/zero-extending pointer types.
14714 (expand_omp_for_static_chunk): Likewise.
14716 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
14717 Jean Lee <xiaoyur347@gmail.com>
14719 * config/mips/mips.c (mips_asan_shadow_offset): New function.
14720 (TARGET_ASAN_SHADOW_OFFSET): Define.
14721 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
14722 true for -fsanitize=address.
14724 2018-04-25 Mark Wielaard <mark@klomp.org>
14726 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
14729 2018-04-25 Jakub Jelinek <jakub@redhat.com>
14731 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
14732 than "alu", remove explicit "memory" and "imm_disp" attributes.
14733 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
14735 PR middle-end/85414
14736 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
14737 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
14738 gen_lowpart_no_emit.
14740 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
14743 * config/i386/i386.c (ix86_expand_builtin): Change memory
14744 operand to XI, extend p0 to Pmode.
14745 * config/i386/i386.md: Change unspec volatile and operand
14746 1 mode to XI, change operand 0 mode to P.
14748 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14750 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
14751 GET_MODE_MASK before any checking.
14752 (nds32_can_use_bset_p): Likewise.
14753 (nds32_can_use_btgl_p): Likewise.
14755 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14757 * config/nds32/nds32-doubleword.md: New define_split pattern for
14758 illegal register number.
14760 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14762 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
14764 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14766 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
14768 2018-04-25 Richard Biener <rguenther@suse.de>
14770 * lto-streamer.h (LTO_major_version): Bump to 8.
14772 2018-04-25 Jakub Jelinek <jakub@redhat.com>
14774 * BASE-VER: Set to 9.0.0.
14776 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
14778 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
14779 in __abskf2 and __powikf2.
14781 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14784 * config/aarch64/constraints.md (Usg, Usj): New constraints.
14785 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
14786 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
14787 Use the above on operand 2. Reindent.
14788 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
14790 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
14793 * common/config/i386/i386-common.c (ix86_handle_option): Don't
14795 * config/i386/i386.opt (mcet): Removed.
14796 * doc/install.texi: Remove -mcet documentation.
14797 * doc/invoke.texi: Likewise.
14799 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
14802 * doc/install.texi: Remove -mcet from bootstrap-cet.
14804 2018-04-24 Jakub Jelinek <jakub@redhat.com>
14807 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
14808 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
14812 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
14813 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
14814 containing a CONST_VECTOR.
14816 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
14818 * doc/install.texi: Update newlib dependency for nvptx.
14820 2018-04-24 Jakub Jelinek <jakub@redhat.com>
14823 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
14824 instead of INTVAL when shifting x left.
14826 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
14828 PR tree-optimization/85478
14829 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
14830 vect_grouped_store_supported for single element vectors.
14832 2018-04-24 Richard Biener <rguenther@suse.de>
14835 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
14836 load cost increase to the case of non-constant step.
14838 2018-04-24 Jakub Jelinek <jakub@redhat.com>
14841 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
14842 destination if any_malformed_asm.
14844 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
14846 PR middle-end/85496
14847 * expr.c (store_field): In the bitfield case, if the value comes from
14848 a function call and is returned in registers by means of a PARALLEL,
14849 do not change the mode of the temporary unless BLKmode and VOIDmode.
14851 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
14853 PR rtl-optimization/85423
14854 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
14855 dependencies to debug insns when the previous insn is non-debug.
14857 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
14859 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
14860 enums into a single definition.
14861 (fls): Fix predicates and printing.
14864 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
14866 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
14867 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
14868 and short u6 immediate.
14869 (check_if_valid_sleep_operand): Remove.
14870 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
14872 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14874 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
14875 flag_always_save_lp condition.
14876 * config/nds32/nds32.opt (malways-save-lp): New option.
14878 2018-04-22 Shiva Chen <shiva0217@gmail.com>
14880 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
14881 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
14882 * config/nds32/nds32.h
14883 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
14884 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
14886 2018-04-22 Shiva Chen <shiva0217@gmail.com>
14888 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
14889 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
14891 2018-04-22 Shiva Chen <shiva0217@gmail.com>
14892 Chung-Ju Wu <jasonwucj@gmail.com>
14894 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
14896 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
14897 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
14899 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14901 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
14903 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14905 * config/nds32/nds32-protos.h (nds32_data_alignment,
14906 nds32_local_alignment): Declare.
14907 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
14908 nds32_local_alignment): New functions.
14909 (TARGET_CONSTANT_ALIGNMENT): Define.
14910 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
14912 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14914 * config/nds32/nds32.c
14915 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
14916 (TARGET_MODES_TIEABLE_P): Likewise.
14918 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14920 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
14921 level Ofast and Og.
14923 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
14924 Chung-Ju Wu <jasonwucj@gmail.com>
14926 * config/nds32/constants.md (unspec_volatile_element): Add enum values
14927 for unaligned access.
14928 * config/nds32/nds32-intrinsic.c: Implementation of expanding
14930 * config/nds32/nds32-intrinsic.md: Likewise.
14931 * config/nds32/nds32_intrinsic.h: Likewise.
14932 * config/nds32/nds32.h (nds32_builtins): Likewise.
14933 * config/nds32/nds32.opt (munaligned-access): New option.
14934 * config/nds32/nds32.c (nds32_asm_file_start): Display
14935 flag_unaligned_access status.
14937 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
14939 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
14940 -mno-relax is present.
14941 * config/riscv/linux.h (LINK_SPEC): Ditto.
14943 2018-04-20 Martin Sebor <msebor@redhat.com>
14946 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
14948 (gimple_fold_builtin_stxcpy_chk): Same.
14949 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
14951 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
14954 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
14955 __powikf2 when long double is IEEE 128-bit.
14957 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
14959 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
14960 step to make sure stack always aligned.
14962 2018-04-20 Carl Love <cel@us.ibm.com>
14965 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
14966 size check for arg0.
14968 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
14969 Tom de Vries <tom@codesourcery.com>
14972 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
14973 Emit insns for calls too.
14974 (nvptx_find_par): Always look for worker-level predecessor insn.
14975 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
14977 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
14978 (nvptx_process_pars): Propagate frames for calls.
14980 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
14983 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
14985 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
14986 (ix86_handle_option): Don't handle OPT_mibt.
14987 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
14989 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
14991 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
14992 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
14993 (ix86_target_macros): Define __CET__ with flag_cf_protection
14994 for -fcf-protection.
14995 * config/i386/i386.c (isa2_opts): Remove -mibt.
14996 * config/i386/i386.h (TARGET_IBT): Removed.
14997 (TARGET_IBT_P): Likewise.
14998 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
14999 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
15000 * config/i386/i386.opt (mcet): Update help message.
15001 (mshstk): Likewise.
15003 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
15004 -mcet as an alias for -mshstk.
15006 2018-04-20 Richard Biener <rguenther@suse.de>
15008 PR middle-end/85475
15009 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
15010 complexity by forcing a single use of the multiply operand.
15012 2018-04-20 Martin Jambor <mjambor@suse.cz>
15015 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
15016 recursion dependency to only apply to non-clones.
15018 2018-04-20 Martin Jambor <mjambor@suse.cz>
15021 * ipa-cp.c (create_specialized_node): Check that clones of
15022 self-recursive edges exist during IPA-CP.
15024 2018-04-19 Toon Moene <toon@moene.org>
15026 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
15029 2018-04-19 Jakub Jelinek <jakub@redhat.com>
15031 PR tree-optimization/85467
15032 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
15033 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
15034 VECTOR_CST element to type.
15036 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
15039 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
15040 * config/i386/i386.md (builtin_setjmp_setup): Removed.
15041 (builtin_longjmp): Likewise.
15042 (save_stack_nonlocal): New pattern.
15043 (restore_stack_nonlocal): Likewise.
15045 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
15048 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
15049 Replace ASM_OUTPUT_LABEL with fprintf.
15051 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
15054 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
15055 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
15056 * config/i386/i386-c.c (ix86_target_macros_internal): Also
15057 define __IBT__ and __SHSTK__ for -fcf-protection.
15058 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
15060 (ix86_trampoline_init): Likewise.
15061 (x86_output_mi_thunk): Likewise.
15062 (ix86_notrack_prefixed_insn_p): Likewise.
15063 (ix86_option_override_internal): Don't disallow -fcf-protection.
15064 * config/i386/i386.md (rdssp<mode>): Also enable for
15066 (incssp<mode>): Likewise.
15067 (nop_endbr): Likewise.
15068 * config/i386/i386.opt (mcet): Change help message to built-in
15071 (mshstk): Likewise.
15072 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
15073 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
15074 enable CET built-in functions.
15076 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
15078 * common/config/i386/i386-common.c
15079 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
15080 OPTION_MASK_ISA_MOVDIRI_UNSET,
15081 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
15082 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
15083 * config.gcc (movdirintrin.h): New header.
15084 * config/i386/cpuid.h (bit_MOVDIRI,
15085 bit_MOVDIR64B): New bits.
15086 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
15088 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
15089 (VOID, PVOID, PCVOID)): New function types.
15090 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
15091 __builtin_ia32_directstoreu_u64,
15092 __builtin_ia32_movdir64b): New builtins.
15093 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
15094 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
15096 (ix86_valid_target_attribute_inner_p): Ditto.
15097 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
15098 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
15099 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
15100 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
15101 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
15102 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
15103 (movdiri<mode>, movdir64b_<mode>): New.
15104 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
15105 * config/i386/immintrin.h: Include movdirintrin.h.
15106 * config/i386/movdirintrin.h: New file.
15107 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
15109 2018-04-19 Richard Biener <rguenther@suse.de>
15111 PR middle-end/85455
15112 * cfg.c (clear_bb_flags): When loop state says we have
15113 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
15115 2018-04-19 Richard Biener <rguenther@suse.de>
15117 PR tree-optimization/84737
15118 * tree-vect-data-refs.c (vect_copy_ref_info): New function
15119 copying restrict info.
15120 (vect_setup_realignment): Use it.
15121 * tree-vectorizer.h (vect_copy_ref_info): Declare.
15122 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
15123 the first DR to all generated stores.
15124 (vectorizable_load): Likewise for loads.
15126 2018-04-19 Jakub Jelinek <jakub@redhat.com>
15128 PR tree-optimization/85446
15129 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
15130 the integral and pointer types to have the same precision.
15132 * doc/install.texi: Document --disable-cet being the default and
15135 2018-04-18 Martin Liska <mliska@suse.cz>
15137 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
15140 2018-04-18 Martin Liska <mliska@suse.cz>
15143 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
15146 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
15147 arguments if they are comparable.
15149 2018-04-18 Martin Liska <mliska@suse.cz>
15152 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
15155 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
15158 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
15161 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
15162 ENDBR after calling __morestack.
15164 2018-04-18 David Malcolm <dmalcolm@redhat.com>
15167 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
15168 by using gcc_base_ver to generate a gcc_driver_version, and use
15169 it when generating GCC_DRIVER_NAME.
15170 * configure: Regenerate.
15172 2018-04-18 Jakub Jelinek <jakub@redhat.com>
15175 * config.gcc: Obsolete powerpc*-*-*spe*.
15177 2018-04-17 Jakub Jelinek <jakub@redhat.com>
15180 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
15181 (stabstr_D): Change type of unum from unsigned int to
15182 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
15185 2018-04-17 Jim Wilson <jimw@sifive.com>
15188 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
15189 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
15190 Set arg_pointer_offset after using pretend_args_size.
15192 2018-04-17 Jakub Jelinek <jakub@redhat.com>
15194 PR rtl-optimization/85431
15195 * dse.c (record_store): Ignore zero width stores.
15198 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
15199 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
15200 __builtin_stack_restore rather than after it.
15201 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
15202 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
15203 argument instead of virtual_dynamic_stack_rtx.
15205 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
15207 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
15209 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
15210 Add note to error message to explain internal mapping of overloaded
15211 built-in function name to non-overloaded built-in function name.
15212 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
15215 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
15218 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
15219 where the inputs overlap with the output.
15221 2018-04-17 Jakub Jelinek <jakub@redhat.com>
15223 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
15224 (=v, v) alternative and explicit "memory" attribute.
15225 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
15226 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
15228 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
15229 "sselog1" type instead of "sselog".
15230 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
15231 "sselog". Remove explicit "memory" attribute.
15232 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
15233 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
15235 (vec_extract_hi_v32hi): Merge all alternatives into one, use
15236 "sselog1" type instead of "sselog". Remove explicit "memory"
15238 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
15239 use "sselog1" type instead of "sselog". Remove explicit "memory"
15241 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
15242 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
15244 (vec_extract_hi_v64qi): Merge all alternatives into one, use
15245 "sselog1" type instead of "sselog". Remove explicit "memory"
15247 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
15248 use "sselog1" type instead of "sselog". Remove explicit "memory"
15252 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
15254 PR middle-end/85414
15255 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
15258 2018-04-17 Martin Jambor <mjambor@suse.cz>
15261 * ipa-cp.c (create_specialized_node): Call
15262 expand_all_artificial_thunks if necessary.
15264 2018-04-17 Martin Liska <mliska@suse.cz>
15267 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
15268 in message, remote space in between '_G' and '('.
15270 2018-04-17 Jakub Jelinek <jakub@redhat.com>
15273 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
15274 avx512f_vmcmp<mode>3<round_saeonly_name>,
15275 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
15276 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
15277 avx512f_rndscale<mode><round_saeonly_name>,
15278 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
15279 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
15280 Use %<iptr>2 instead of %2 for -masm=intel.
15281 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
15282 avx512f_vcvttss2usi<round_saeonly_name>,
15283 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
15285 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
15286 avx512f_vcvttsd2usi<round_saeonly_name>,
15287 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
15288 Use %q1 instead of %1 for -masm=intel.
15289 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
15290 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
15291 of %3 for -masm=intel.
15292 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
15294 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
15296 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
15298 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
15299 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
15301 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
15303 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
15304 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
15305 %g1 and one with %0 and %1.
15306 (avx512er_vmrcp28<mode><round_saeonly_name>,
15307 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
15308 %1 for -masm=intel.
15309 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
15310 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
15311 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
15312 of %0 and %{%4%} for -masm=intel.
15313 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
15314 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
15315 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
15316 order of %0 and %{%5%}%{z%} for -masm=intel.
15318 2018-04-17 Jan Hubicka <jh@suse.cz>
15321 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
15323 2018-04-17 Martin Liska <mliska@suse.cz>
15326 * multiple_target.c (create_dispatcher_calls): Set apostrophes
15327 for target_clone error message. Make default implementation
15328 clone to be a local declaration.
15329 (separate_attrs): Add new argument and check for an empty
15331 (expand_target_clones): Handle it.
15332 (ipa_target_clone): Make redirection just for target_clones
15335 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
15336 Tom de Vries <tom@codesourcery.com>
15338 PR middle-end/84955
15339 * omp-expand.c (expand_oacc_for): Add dummy false branch for
15340 tiled basic blocks without omp continue statements.
15342 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
15345 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
15346 vec_extract expression as having side effects to make sure it gets
15349 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
15352 * config/i386/i386.c (get_builtin_code_for_version): Check
15355 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
15358 * config.gcc: Support "skylake".
15359 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15361 * config/i386/i386.c (m_SKYLAKE): Define.
15362 (processor_target_table): Add "skylake".
15363 (ix86_option_override_internal): Add "skylake".
15364 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
15365 PROCESSOR_CANNONLAKE.
15366 (get_builtin_code_for_version): Fix priority for
15367 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
15368 PROCESSOR_SKYLAKE-AVX512.
15369 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
15370 (processor_type): Add PROCESSOR_SKYLAKE.
15372 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
15373 Jason Merrill <jason@redhat.com>
15376 * convert.c (convert_to_integer_1): Use direct recursion for
15377 enumeral types and types with a precision less than the number
15378 of bits in their mode.
15380 2018-04-16 Julia Koval <julia.koval@intel.com>
15383 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
15384 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
15386 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
15389 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
15390 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
15391 and -mno-direct-move.
15393 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
15396 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
15397 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
15398 Ensure negative shifts result in {0}.
15400 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
15402 PR rtl-optimization/79916
15403 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
15404 regs (if any) to define how to gnerate SD moves when LRA is in
15407 2018-04-13 Jakub Jelinek <jakub@redhat.com>
15409 PR rtl-optimization/85393
15410 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
15411 * except.c (expand_dw2_landing_pad_for_region): Make static.
15412 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
15413 a label and unconditional jump to old_bb, rather than
15414 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
15417 PR rtl-optimization/85376
15418 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
15419 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
15420 instead of a specific value.
15422 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
15423 Bin Cheng <bin.cheng@arm.com>
15425 PR tree-optimization/82965
15426 PR tree-optimization/83991
15427 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
15428 by_profile_only parameter.
15429 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
15430 information if the loop was predicted to iterate too many times.
15431 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
15433 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
15436 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
15439 2018-04-13 Martin Liska <mliska@suse.cz>
15440 Jakub Jelinek <jakub@redhat.com>
15442 PR middle-end/81657
15443 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
15444 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
15445 * builtins.c (expand_builtin_memory_copy_args): Use
15446 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
15447 handle dest_addr == pc_rtx.
15449 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
15452 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
15453 asked to not generate direct moves.
15454 (fix_trunc<mode>si2_stfiwx): Similar.
15455 (fix_trunc<mode>si2_internal): Similar.
15457 2018-04-12 Jakub Jelinek <jakub@redhat.com>
15460 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
15461 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
15462 lookup if dest in some wider mode is known to be const0_rtx and
15463 if so, record permanent equivalence for it to be ZERO_EXTEND of
15464 the narrower mode destination.
15466 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
15468 * lto-streamer-out.c (output_function): Revert 259346.
15469 * omp-expand.c (expand_oacc_for): Likewise.
15471 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
15473 PR rtl-optimization/85354
15474 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
15475 * sel-sched.c (sel_global_init): ... here.
15477 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
15480 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
15481 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
15482 mode for PE-COFF targets.
15483 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
15484 (i386_pe_asm_lto_end): Likewise.
15485 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
15486 (TARGET_ASM_LTO_END): Likewise.
15487 * config/i386/winnt.c (saved_debug_info_level): New static variable.
15488 (i386_pe_asm_lto_start): New function.
15489 (i386_pe_asm_lto_end): Likewise.
15491 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
15492 Richard Biener <rguenther@suse.de>
15494 PR middle-end/84955
15495 * lto-streamer-out.c (output_function): Fix CFG loop state before
15497 * omp-expand.c (expand_oacc_for): Handle calls to internal
15498 functions like regular functions.
15500 2018-04-12 Richard Biener <rguenther@suse.de>
15503 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
15504 for the early LTO debug to properly generate references to it
15505 during DIE emission. Do not re-use that for the skeleton for
15507 (dwarf2out_early_finish): Likewise.
15509 2018-04-12 Jakub Jelinek <jakub@redhat.com>
15512 * config/i386/sse.md
15513 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
15514 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
15515 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
15516 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
15517 and output is a reg, avoid creating invalid lowpart subreg, but
15518 instead split into a 512-bit move. Don't split if not AVX512VL,
15519 input is xmm16+ reg and output is a mem.
15520 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
15521 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
15522 xmm16+ reg and output is a mem.
15524 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15526 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
15527 also for flag_dwarf2_cfi_asm.
15529 2018-04-12 Jakub Jelinek <jakub@redhat.com>
15531 PR rtl-optimization/85342
15532 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
15533 a bool scalar var inside of the loop instead. Don't try to update
15534 recog_data.operand after failed apply_change_group.
15536 2018-04-12 Tom de Vries <tom@codesourcery.com>
15539 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
15540 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
15541 array with flexible array member as array without given dimension.
15542 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
15543 argument for undefined param to true.
15545 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
15548 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
15549 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
15550 from PowerPC section.
15551 * config/rs6000/sysv4.opt (mcall-): Improve help text.
15552 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
15553 help text that is too long.
15554 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
15555 help text that is too long.
15556 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
15557 help text that is too long.
15559 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
15561 * config/alpha/alpha.md (stack_probe_internal): Rename
15562 from "probe_stack". Update all callers.
15564 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
15566 PR rtl-optimization/84566
15567 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
15568 sched_macro_fuse_insns.
15570 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
15573 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
15574 (compute_block_dependences): ... from here.
15576 2018-04-11 Jakub Jelinek <jakub@redhat.com>
15578 PR tree-optimization/85331
15579 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
15580 from int to HOST_WIDE_INT.
15582 2018-04-11 Martin Jambor <mjambor@suse.cz>
15585 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
15586 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
15587 not the same as the source val.
15588 (cgraph_edge_brings_value_p): New parameter.
15589 (gather_edges_for_value): Pass destination value to
15590 cgraph_edge_brings_value_p.
15591 (perhaps_add_new_callers): Likewise.
15592 (get_info_about_necessary_edges): Likewise and exclude values brought
15593 only by self-recursive edges.
15594 (create_specialized_node): Redirect only clones of self-calling edges.
15595 (+self_recursive_pass_through_p): New function.
15596 (find_more_scalar_values_for_callers_subset): Use it.
15597 (find_aggregate_values_for_callers_subset): Likewise.
15598 (known_aggs_to_agg_replacement_list): Removed.
15599 (decide_whether_version_node): Re-calculate known constants for all
15600 remaining context clones.
15602 2018-04-11 Richard Biener <rguenther@suse.de>
15605 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
15606 from early DWARF output.
15607 (dwarf2out_early_finish): Output line info unconditionally into
15608 early DWARF and add reference to it.
15610 2018-04-11 Jakub Jelinek <jakub@redhat.com>
15613 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
15614 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
15615 other than V2DFmode using iptr mode attribute.
15616 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
15618 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
15620 PR rtl-optimization/84659
15621 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
15623 2018-04-11 Jakub Jelinek <jakub@redhat.com>
15626 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
15628 (output_loc_list): Pass address of a dummy size variable even in the
15629 locview handling loop.
15630 (index_location_lists): Add comment on why skip_loc_list_entry can't
15633 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
15636 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
15639 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
15642 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
15643 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
15644 and -mstring-compare-inline-limit.
15646 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
15649 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
15650 for stack clash protection in a register whenever we need it to be in
15653 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
15655 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
15656 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
15658 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
15661 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
15663 (mlong-double-): Ditto.
15664 * config/rs6000/sysv4.opt (msdata=): Ditto.
15665 (mtls-size=): Ditto.
15667 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
15669 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
15670 erroneous entries for
15671 "vector int vec_ldl (int, long int *)", and
15672 "vector unsigned int vec_ldl (int, unsigned long int *)".
15673 Add comments and entries for
15674 "vector bool char vec_ldl (int, bool char *)",
15675 "vector bool short vec_ldl (int, bool short *)",
15676 "vector bool int vec_ldl (int, bool int *)",
15677 "vector bool long long vec_ldl (int, bool long long *)",
15678 "vector pixel vec_ldl (int, pixel *)",
15679 "vector long long vec_ldl (int, long long *)",
15680 "vector unsigned long long vec_ldl (int, unsigned long long *)".
15681 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
15682 type tree bool_long_long_type_node and correct definition of
15683 bool_V2DI_type_node to make reference to this new type tree.
15684 (rs6000_mangle_type): Replace erroneous reference to
15685 bool_long_type_node with bool_long_long_type_node.
15686 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
15687 comments to emphasize sign distinctions for char and int types and
15688 replace RS6000_BTI_bool_long constant with
15689 RS6000_BTI_bool_long_long constant. Also add comment to restrict
15690 use of RS6000_BTI_pixel.
15691 (bool_long_type_node): Remove this macro definition.
15692 (bool_long_long_type_node): New macro definition
15694 2018-04-10 Jakub Jelinek <jakub@redhat.com>
15696 PR rtl-optimization/85300
15697 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
15698 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
15699 simplify_unary_operation fails.
15701 2018-04-10 Martin Liska <mliska@suse.cz>
15703 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
15704 cgraph_edge and ipa_ref.
15706 2018-04-10 Jakub Jelinek <jakub@redhat.com>
15710 * config/i386/sse.md
15711 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
15712 computation of the VEC_MERGE selector from mask.
15713 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
15714 Fix decoding of the VEC_MERGE selector into mask.
15716 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
15718 PR tree-optimization/85286
15719 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
15721 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
15723 * final.c (final_1): Set insn_last_address as well as
15724 insn_current_address.
15726 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15729 * explow.c (emit_stack_probe): Call validize_mem on memory location
15730 before passing it to gen_probe_stack. Create address operand and
15731 legitimize it for the probe_stack_address case.
15733 2018-04-09 Jan Hubicka <jh@suse.cz>
15736 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
15737 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
15738 * tree.c (free_lang_data_in_type): Fix handling of binfos;
15740 (free_lang_data): Rebuild type inheritance graph.
15742 2018-04-09 Martin Sebor <msebor@redhat.com>
15744 * invoke.texi (-finline-small-functions): Mention other optimization
15746 (-findirect-inlining, -fpartial-inlining): Same.
15747 (-finline-functions-called-once): Same.
15748 (-freorder-blocks-and-partition): Same.
15750 2018-04-09 Jan Hubicka <jh@suse.cz>
15753 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
15754 jumps; choose last target that matches the criteria (i.e.
15755 no partition changes for non-crossing jumps).
15756 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
15757 support for redirecting crossing jumps to non-crossing.
15759 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
15761 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
15762 also for naked functions.
15764 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
15766 * config/arc/arc.md (add_shift): New pattern.
15767 (add_shift2): Likewise.
15768 (sub_shift): Likewise.
15769 (sub_shift_cmp0_noout): Likewise.
15770 (compare_si_ashiftsi): Likewise.
15771 (xbfu_cmp0_noout): New combine pattern.
15772 (xbfu_cmp0"): Likewise.
15773 (movsi_set_cc_insn): Place the predicable variant first.
15774 (commutative_binary_cmp0_noout): Remove clobber.
15775 (commutative_binary_cmp0): New pattern.
15776 (noncommutative_binary_cmp0): Likewise.
15777 (noncommutative_binary_cmp0_noout): Likewise.
15778 (noncommutative_binary_comparison_result_used): Removed.
15779 (rsub_cmp0): New pattern.
15780 (rsub_cmp0_noout): Likewise.
15781 (extzvsi): Changed, keep only meaningful variants.
15782 (SQH, SEZ): New iterators.
15783 (SQH_postfix): New mode attribute.
15784 (SEZ_prefix): New code attribute.
15785 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
15786 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
15787 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
15788 of numerical value.
15789 (noncommutative_operator): Check the availability of barrel
15792 2018-04-09 Richard Biener <rguenther@suse.de>
15794 PR tree-optimization/85284
15795 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
15796 Only use the niter constraining form of simple_iv when the exit
15797 is always executed.
15799 2018-04-09 Tom de Vries <tom@codesourcery.com>
15802 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
15803 (define_expand "*memory_barrier"): New define_expand.
15804 (define_insn "memory_barrier"): New insn.
15806 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15808 PR rtl-optimization/80463
15809 PR rtl-optimization/83972
15810 PR rtl-optimization/83480
15812 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
15813 correct producer for the insn.
15814 (tidy_control_flow): Fixup seqnos in case of debug insns.
15816 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15818 PR rtl-optimization/83913
15820 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
15821 different sched-times when merging exprs.
15823 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15825 PR rtl-optimization/83962
15827 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
15828 tidy_fallthru_edge and tidy_control_flow.
15830 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15832 PR rtl-optimization/83530
15834 * sel-sched.c (force_next_insn): New global variable.
15835 (remove_insn_for_debug): When force_next_insn is true, also leave only
15836 next insn in the ready list.
15837 (sel_sched_region): When the region wasn't scheduled, make another pass
15838 over it with force_next_insn set to 1.
15840 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
15842 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
15844 * config/nds32/constants.md (unspec_volatile_element): Add enum values
15845 for interrupt control.
15846 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
15847 functions for interrupt control.
15848 * config/nds32/nds32-intrinsic.md: Likewise.
15849 * config/nds32/nds32_intrinsic.h: Likewise.
15850 * config/nds32/nds32.h (nds32_builtins): Likewise.
15852 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
15854 * config/nds32/nds32.c (nds32_init_machine_status,
15855 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
15856 strict_aligned_p field.
15857 (nds32_expand_to_rtl_hook): New function.
15858 (TARGET_EXPAND_TO_RTL_HOOK): Define.
15859 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
15861 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15862 Chung-Ju Wu <jasonwucj@gmail.com>
15864 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
15865 * config/nds32/nds32-n7.md: New file.
15866 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
15867 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
15869 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
15870 * config/nds32/nds32.md (pipeline_model): Add n7.
15871 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
15872 * config/nds32/pipelines.md: Include n7 settings.
15874 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15875 Chung-Ju Wu <jasonwucj@gmail.com>
15877 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
15878 * config/nds32/nds32-e8.md: New file.
15879 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
15880 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
15882 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
15883 * config/nds32/nds32.md (pipeline_model): Add e8.
15884 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
15885 * config/nds32/pipelines.md: Include e8 settings.
15887 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15888 Chung-Ju Wu <jasonwucj@gmail.com>
15890 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
15891 * config/nds32/nds32-n8.md: New file.
15892 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
15893 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
15895 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
15896 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
15897 * config/nds32/nds32.md (pipeline_model): Add n8.
15898 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
15899 * config/nds32/pipelines.md: Include n8 settings.
15901 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15902 Chung-Ju Wu <jasonwucj@gmail.com>
15904 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
15905 * config/nds32/nds32-n9-2r1w.md: New file.
15906 * config/nds32/nds32-n9-3r2w.md: New file.
15907 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
15908 nds32_register_ports): New or modify for cpu n9.
15909 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
15911 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
15912 * config/nds32/nds32-utils.c: New file.
15913 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
15914 TARGET_MUL_SLOW): Define.
15915 * config/nds32/nds32.md (pipeline_model): New attribute.
15916 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
15917 New options that support cpu n9.
15918 * config/nds32/pipelines.md: Include n9 settings.
15919 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
15921 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
15923 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
15924 information if necessary.
15925 (output_cond_branch_compare_zero): Likewise.
15926 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
15927 (nds32_target_alignment): Refine for alignment.
15928 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
15929 (FUNCTION_BOUNDARY): Modify.
15930 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
15932 * config/nds32/nds32.opt (malways-align, malign-functions): New.
15934 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
15936 * config/nds32/constants.md (unspec_volatile_element): Add values for
15937 TLB operation and data prefetch.
15938 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
15939 functions for TLB operation and data prefetch.
15940 * config/nds32/nds32-intrinsic.md: Likewise.
15941 * config/nds32/nds32_intrinsic.h: Likewise.
15942 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
15943 (nds32_print_operand): Likewise.
15944 * config/nds32/nds32.h (nds32_builtins): Likewise.
15946 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
15947 Andrew Pinski <pinsika@gcc.gnu.org>
15949 PR middle-end/82976
15950 * match.pd: Use constant_boolean_node of correct type instead of
15951 boolean_true_node or boolean_false_node for simplifying
15952 pointer comparisons to zero.
15954 2018-04-07 Jakub Jelinek <jakub@redhat.com>
15956 PR tree-optimization/80021
15957 * tree.c (verify_type_variant): Make error call in verify_variant_match
15958 translatable and remove final full stop.
15960 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
15962 * config/nds32/constants.md (unspec_volatile_element): Add
15963 UNSPEC_VOLATILE_EH_RETURN.
15964 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
15965 nds32_output_stack_pop): Support dwarf exception handling process.
15966 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
15967 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
15968 exception handling process.
15969 (nds32_compute_stack_frame): Likewise.
15970 (nds32_return_addr_rtx): Likewise.
15971 (nds32_initial_elimination_offset): Likewise.
15972 (nds32_expand_prologue): Likewise.
15973 (nds32_expand_epilogue): Likewise.
15974 (nds32_dynamic_chain_address): New function.
15975 * config/nds32/nds32.h (machine_function): Add fields for dwarf
15976 exception handling.
15977 (DYNAMIC_CHAIN_ADDRESS): Define.
15978 (EH_RETURN_DATA_REGNO): Define.
15979 (EH_RETURN_STACKADJ_RTX): Define.
15980 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
15981 patterns for dwarf exception handling.
15983 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
15985 * config/nds32/nds32.h: Clean up obsolete macros.
15987 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15989 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
15990 Add enum values for particular instructions.
15991 * config/nds32/nds32-intrinsic.c: Implementation of expanding
15992 particular intrinsic functions.
15993 * config/nds32/nds32-intrinsic.md: Likewise.
15994 * config/nds32/nds32_intrinsic.h: Likewise.
15995 * config/nds32/nds32.h (nds32_builtins): Likewise.
15996 * config/nds32/nds32.md (type): Add pbsad and pbsada.
15997 (btst, ave): New patterns for particular instructions.
15999 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
16001 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
16002 Add enum values for atomic load/store and memory sync.
16003 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
16005 * config/nds32/nds32-intrinsic.md: Likewise.
16006 * config/nds32/nds32_intrinsic.h: Likewise.
16007 * config/nds32/nds32.h (nds32_builtins): Likewise.
16009 2018-04-07 Jakub Jelinek <jakub@redhat.com>
16011 PR tree-optimization/85257
16012 * fold-const.c (native_encode_vector): If not all elts could fit
16013 and off is -1, return 0 rather than offset.
16014 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
16015 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
16016 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
16017 adjust buffer in native_interpret_expr call.
16019 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
16021 * config/nds32/constants.md (unspec_volatile_element): Add cache
16022 control enum values.
16023 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
16024 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
16025 * config/nds32/nds32.c (nds32_cctl_names): New.
16026 (nds32_print_operand): Handle cache control register names.
16027 * config/nds32/nds32.h (nds32_builtins): New enum values.
16028 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
16030 * config/nds32/nds32.md (type): Add mmu.
16031 * config/nds32/pipelines.md (simple_insn): Add mmu.
16033 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
16035 * config/nds32/nds32.md (type): Remove call.
16036 * config/nds32/pipelines.md (simple_insn): Likewise.
16038 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
16040 * config/nds32/constants.md (unspec_volatile_element): Add
16041 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
16042 UNSPEC_VOLATILE_FMFCFG.
16043 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
16044 description for fmfcfg and fmfcsr.
16045 (bdesc_1arg): Add fmtcsr.
16046 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
16047 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
16048 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
16049 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
16050 unspec_fmfcfg): New patterns.
16051 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
16052 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
16053 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
16054 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
16055 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
16056 __nds32__fmfcfg): Define.
16058 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
16060 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
16061 intrinsic register names.
16062 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
16063 intrinsic register enum values and macros.
16065 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
16067 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
16068 for load/store addressing form.
16069 (nds32_print_operand_address): Likewise.
16071 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
16074 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
16075 based on LABEL_REF. Remove useless assertion.
16076 (pic_address_needs_scratch): Fix formatting.
16077 (sparc_legitimize_pic_address): Minor tweaks.
16078 (sparc_delegitimize_address): Adjust assertion accordingly.
16079 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
16080 into symbolic_operand.
16081 (movsi_high_pic_label_ref): Likewise.
16082 (movsi_lo_sum_pic_label_ref): Likewise.
16083 (movdi_pic_label_ref): Likewise.
16084 (movdi_high_pic_label_ref): Likewise.
16085 (movdi_lo_sum_pic_label_ref): Likewise.
16087 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
16089 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
16090 custom LIB_SPEC setup.
16092 2018-04-06 Ruslan Bukin <br@bsdpad.com>
16093 Kito Cheng <kito.cheng@gmail.com>
16095 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
16096 * config/riscv/freebsd.h: New.
16098 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
16100 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
16101 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
16104 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
16105 Kito Cheng <kito.cheng@gmail.com>
16107 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
16108 nds32_output_call, nds32_symbol_binds_local_p): New functions.
16109 * config/nds32/nds32-protos.h (nds32_output_call,
16110 nds32_output_return): Declare.
16111 * config/nds32/nds32.md: Refine all the call and return patterns.
16113 2018-04-06 Jakub Jelinek <jakub@redhat.com>
16116 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
16117 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
16119 PR rtl-optimization/84872
16120 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
16121 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
16122 EDGE_CROSSING edge.
16124 2018-04-06 Tamar Christina <tamar.christina@arm.com>
16126 * expr.c (copy_blkmode_to_reg): Revert 254862.
16127 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
16129 2018-04-06 Richard Biener <rguenther@suse.de>
16131 PR middle-end/85244
16132 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
16133 after seeing a component reference with an adjacent field. Treat
16134 refs to arrays at struct end of external decls similar to
16135 refs to unconstrained commons.
16137 2018-04-06 Jakub Jelinek <jakub@redhat.com>
16140 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
16141 look through SAVE_EXPRs with non-side-effects argument. Adjust
16143 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
16146 2018-04-06 Richard Biener <rguenther@suse.de>
16148 PR middle-end/85180
16149 * alias.c (find_base_term): New wrapper around find_base_term
16150 unwinding CSELIB_VAL_PTR changes.
16151 (find_base_term): Do not restore CSELIB_VAL_PTR during the
16154 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
16156 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
16158 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
16159 constant definitions.
16160 ("nop"): lr 0,0 -> nopr r0
16161 ("nop_lr0", "nop_lr1"): New insn definitions.
16163 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
16165 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
16166 NDS32_V3PUSH_AVAILABLE_P macro.
16168 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
16169 Chung-Ju Wu <jasonwucj@gmail.com>
16171 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
16172 (nds32*-*-*): Add float and fpu_config into supported_defaults.
16173 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
16174 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
16175 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
16176 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
16177 * config/nds32/constraints.md: New constraints and checking for hard
16178 float configuration.
16179 * config/nds32/iterators.md: New mode iterator and attribute for hard
16180 float configuration.
16181 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
16183 * config/nds32/nds32-fpu.md: New file.
16184 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
16185 deal with hard float code generation.
16186 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
16188 (abi_type, float_reg_number): New enum type.
16189 * config/nds32/nds32-predicates.c: New predicates for hard float.
16190 * config/nds32/nds32-protos.h: Declare functions for hard float.
16191 * config/nds32/nds32.c: Implementation for hard float configuration.
16192 * config/nds32/nds32.h: Definitions for hard float configuration.
16193 * config/nds32/nds32.md: Include hard float machine description and
16194 modify patterns for hard float configuration.
16195 * config/nds32/nds32.opt: New options for hard float configuration.
16196 * config/nds32/predicates.md: New predicates for hard float
16199 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
16201 * common/config/nds32/nds32-common.c
16202 (nds32_option_optimization_table): Enable -mreleax-hint by default.
16204 2018-04-05 Jakub Jelinek <jakub@redhat.com>
16206 PR middle-end/85195
16207 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
16208 CONSTRUCTOR_ELT (ctor, ...)->value.
16210 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
16213 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
16215 2018-04-05 Tom de Vries <tom@codesourcery.com>
16218 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
16221 2018-04-05 Shiva Chen <shiva0217@gmail.com>
16222 Kito Cheng <kito.cheng@gmail.com>
16224 * config/nds32/constraints.md (U33): Fine-tune checking condition.
16225 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
16226 * config/nds32/nds32.h (nds32_16bit_address_type): Add
16227 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
16229 2018-04-05 Shiva Chen <shiva0217@gmail.com>
16230 Kito Cheng <kito.cheng@gmail.com>
16232 * config/nds32/constraints.md (Ufe): New memory constraint.
16233 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
16234 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
16235 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
16237 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
16238 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
16240 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16242 * config/nds32/nds32.md: Use optimize_size in the condition for
16243 alu-shift instructions.
16245 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16247 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
16249 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16251 * config/nds32/nds32.md (negsi2): Refine pattern.
16253 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
16254 Chung-Ju Wu <jasonwucj@gmail.com>
16256 * config/nds32/iterators.md (shift_rotate): New code iterator.
16257 (shift): New code attribute.
16258 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
16259 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
16260 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
16261 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
16262 bit-wise operations.
16263 (andsi3, *andsi3): Ditto.
16264 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
16265 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
16266 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
16267 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
16268 nds32_ior_operand, nds32_xor_operand): New predicates.
16270 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16272 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
16273 (addsi3, subsi3): ... this.
16275 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16277 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
16279 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16281 * config/nds32/nds32.md: Adjust indention.
16283 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
16285 * config/nds32/nds32.md (feature): New attribute.
16287 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16289 * config/nds32/nds32.md (subtype): New attribute.
16291 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
16294 * config/arm/arm-builtins.c (arm_expand_builtin): Change
16295 expansion to perform a bitwise AND of the argument followed by a
16296 boolean negation of the result.
16298 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
16300 PR rtl-optimization/84878
16301 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
16302 the basic block. Assert the use reference is not artificial and that
16303 it has an associated insn.
16305 2018-04-04 Michael Matz <matz@suse.de>
16307 * builtins.c (compute_objsize): Pass correct operand
16308 to array_at_struct_end_p.
16310 2018-04-04 Richard Biener <rguenther@suse.de>
16313 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
16314 from contexts for DINFO_LEVEL_TERSE and below.
16316 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
16318 * config/nds32/nds32-doubleword.md (move_<mode>): Require
16319 resiter_operand condition.
16320 * config/nds32/nds32.md (*move<mode>): Ditto.
16322 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
16323 Monk Chiang <sh.chiang04@gmail.com>
16325 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
16327 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
16329 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
16331 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
16332 Kito Cheng <kito.cheng@gmail.com>
16334 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
16335 nds32_cond_code_str, output_cond_branch,
16336 output_cond_branch_compare_zero, nds32_expand_cbranch,
16337 nds32_expand_cstore, nds32_expand_movcc,
16338 nds32_output_cbranchsi4_equality_zero,
16339 nds32_output_cbranchsi4_equality_reg,
16340 nds32_output_cbranchsi4_equality_reg_or_const_int,
16341 nds32_output_cbranchsi4_greater_less_zero: New functions.
16342 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
16343 nds32_expand_cstore, nds32_expand_movcc,
16344 nds32_output_cbranchsi4_equality_zero,
16345 nds32_output_cbranchsi4_equality_reg,
16346 nds32_output_cbranchsi4_equality_reg_or_const_int,
16347 nds32_output_cbranchsi4_greater_less_zero): Declare.
16348 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
16349 nds32_rimm11s_operand): New predicates.
16350 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
16351 * config/nds32/nds32.md: Rewrite all the branch and conditional move
16354 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
16356 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
16357 * config/nds32/nds32.md: Ditto.
16358 * config/nds32/pipelines.md: Ditto.
16360 2018-04-04 Richard Biener <rguenther@suse.de>
16362 PR tree-optimization/85168
16363 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
16364 propagating abnormals.
16366 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
16368 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
16370 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
16371 Kito Cheng <kito.cheng@gmail.com>
16373 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
16374 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
16375 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
16376 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
16377 * config/nds32/nds32.md (sibcall_internal): New.
16378 (sibcall_register): Remove.
16379 (sibcall_immediate): Remove.
16380 (sibcall_value_internal): New.
16381 (sibcall_value_register): Remove.
16382 (sibcall_value_immediate): Remove.
16383 * config/nds32/predicates.md (nds32_general_register_operand): New.
16384 (nds32_call_address_operand): New.
16386 2018-04-03 Jakub Jelinek <jakub@redhat.com>
16388 PR rtl-optimization/85167
16389 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
16390 bb_defs if *split_p, instead preinitialize it to NULL.
16392 PR tree-optimization/85156
16393 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
16394 evaluating the argument multiple times.
16396 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
16398 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
16400 (_mm_cvtpd_ps): Likewise.
16401 (_mm_cvttpd_epi32): Likewise.
16402 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
16403 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
16404 vector, pixel, and bool following altivec.h include.
16406 2018-04-03 Martin Sebor <msebor@redhat.com>
16408 * doc/extend.texi (Common Function Attributes): Clarify.
16409 (const attribute): Likewise.
16410 (pure attribute): Likewise.
16412 2018-04-03 Jakub Jelinek <jakub@redhat.com>
16415 * config/i386/i386.c (ix86_expand_vector_set): Use
16416 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
16418 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
16420 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
16421 instructions when changing rounding bits to preserve precision bits
16422 in the x87 control word.
16424 2018-04-03 Martin Liska <mliska@suse.cz>
16426 PR tree-optimization/82491
16427 * rtl.h (strip_offset_and_add): Replace += suboffset with
16428 poly_uint64 () + suboffset.
16430 2018-03-29 Martin Liska <mliska@suse.cz>
16431 Martin Jambor <mjambor@suse.cz>
16434 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
16435 param_type is not an integral or pointer type.
16437 2018-04-03 Richard Biener <rguenther@suse.de>
16439 * sese.h (recompute_all_dominators): Remove.
16441 2018-04-02 Martin Sebor <msebor@redhat.com>
16443 * doc/invoke.texi (-Wrestrict): Fix typos.
16445 2018-04-02 Jim Wilson <jimw@sifive.com>
16447 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
16448 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
16449 (<optab>di3, <optab>si3_extend): Likewise.
16450 (<optab>si3_mask, <optab>si3_mask_1): New.
16451 (<optab>di3_mask, <optab>di3_mask_1): New.
16452 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
16453 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
16454 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
16456 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
16458 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
16461 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
16463 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
16464 (nds32_canonicalize_comparison): New function.
16466 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
16467 Kito Cheng <kito.cheng@gmail.com>
16468 Kuan-Lin Chen <kuanlinchentw@gmail.com>
16470 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
16471 * config/nds32/constants.md (unspec_volatile_element): Add
16472 UNSPEC_VOLATILE_RELAX_GROUP.
16473 * config/nds32/nds32-relax-opt.c: New file.
16474 * config/nds32/nds32-predicates.c
16475 (nds32_symbol_load_store_p): New function.
16476 * config/nds32/nds32-protos.h
16477 (nds32_symbol_load_store_p): Declare function.
16478 (make_pass_nds32_relax_opt): Declare new rtl pass function.
16479 * config/nds32/nds32.c
16480 (nds32_register_pass): New function to register pass.
16481 (nds32_register_passes): New function to register passes.
16482 * config/nds32/nds32.md (relax_group): New pattern.
16483 * config/nds32/nds32.opt (mrelax-hint): New option.
16484 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
16486 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
16488 * config/nds32/t-nds32: Modify files dependency.
16490 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
16492 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
16493 (PROFILE_HOOK): Define its implementation.
16495 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
16497 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
16498 type and 32-bit size.
16500 2018-04-01 Jakub Jelinek <jakub@redhat.com>
16502 PR middle-end/85090
16503 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
16504 (V_128_256): New mode iterator.
16505 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
16506 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
16507 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
16509 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
16510 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
16512 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
16515 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
16516 NaN inputs correctly.
16518 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
16521 * config/rs6000/vsx.md (??r): New mode attribute.
16522 (*vsx_mov<mode>_64bit): Use it.
16523 (*vsx_mov<mode>_32bit): Likewise.
16525 2018-03-30 Martin Sebor <msebor@redhat.com>
16527 PR tree-optimization/84818
16528 * builtins.c (check_access): Use warning_n.
16530 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16533 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
16535 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
16538 2018-03-30 Julia Koval <julia.koval@intel.com>
16541 * x86-tune.def (movx, partial_reg_dependency): Enable for
16544 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
16546 PR inline-asm/84985
16547 * lra-constraints.c (process_alt_operands): Move setting
16548 this_alternative_matches below.
16550 2018-03-29 Martin Liska <mliska@suse.cz>
16553 * doc/invoke.texi: Document how LTO works with debug info.
16554 Describe auto-load support of binutils. Mention 'x86-64'
16555 as valid option value of -march option.
16557 2018-03-29 Jakub Jelinek <jakub@redhat.com>
16559 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
16562 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
16563 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
16564 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
16567 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
16570 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
16571 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
16572 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
16573 for RS6000_BTM_POWERPC64.
16574 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
16575 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
16576 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
16579 (DIVDEU): Likewise.
16581 2018-03-28 Carl Love <cel@us.ibm.com>
16584 2017-09-27 Carl Love <cel@us.ibm.com>
16586 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
16587 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
16588 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
16591 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16593 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
16594 instead of __vector bool.
16595 (_mm_max_pu8): Likewise.
16596 (_mm_min_pi16): Likewise.
16598 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
16601 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
16602 (DIVWEUO): Likewise.
16603 (DIVDEO): Likewise.
16604 (DIVDEUO): Likewise.
16605 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
16606 DIVWEUO and DIVDEUO.
16607 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
16608 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
16609 (div_extend): Likewise.
16610 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
16612 (__builtin_divweuo): Likewise.
16613 (__builtin_divdeo): Likewise.
16614 (__builtin_divdeuo): Likewise.
16616 2018-03-28 Jakub Jelinek <jakub@redhat.com>
16619 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
16620 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
16622 PR tree-optimization/82004
16623 * gimple-match-head.c (optimize_pow_to_exp): New function.
16624 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
16625 Don't fold to exp if optimize_pow_to_exp is false.
16627 2018-03-28 Martin Liska <mliska@suse.cz>
16630 * calls.c (initialize_argument_information): Fix trailing space.
16631 * common.opt: Fix typo and provide better explanation for
16632 -fsanitize-coverage option.
16633 * config/i386/i386.opt: Fix typo.
16635 2018-03-28 Jakub Jelinek <jakub@redhat.com>
16636 Martin Liska <mliska@suse.cz>
16639 * gimplify.c (asan_poison_variable): Don't do the check for
16640 gimplify_omp_ctxp here.
16641 (gimplify_decl_expr): Do it here.
16642 (gimplify_target_expr): Likewise.
16644 2018-03-28 Martin Liska <mliska@suse.cz>
16647 * config/i386/i386.c (ix86_function_arg_advance): Do not call
16648 chkp_type_bounds_count if MPX is not enabled.
16650 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
16652 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
16654 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
16657 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
16658 function to create the function decl for complex long double
16659 multiply and divide for -mabi=ieeelongdouble.
16660 (init_float128_ieee): Call it.
16662 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
16665 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
16666 -fcf-protection=branch -mibt.
16667 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
16669 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
16672 * config/arm/arm.c (arm_valid_symbolic_address): Handle
16673 arm_word_relocations.
16675 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
16678 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
16679 extern array declarations.
16681 2018-03-27 Richard Biener <rguenther@suse.de>
16683 PR middle-end/84067
16684 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
16685 explicit single_use checks.
16687 2018-03-27 Richard Biener <rguenther@suse.de>
16689 PR tree-optimization/85082
16690 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
16693 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
16695 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
16696 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
16697 Turn on fasynchronous-unwind-tables and funwind-tables.
16699 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
16702 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
16703 (*bmi_blsr_<mode>_ccz): Ditto.
16705 2018-03-26 Tom de Vries <tom@codesourcery.com>
16707 PR tree-optimization/85063
16708 * omp-general.c (offloading_function_p): New function. Factor out
16710 * omp-offload.c (pass_omp_target_link::gate): ... here.
16711 * omp-general.h (offloading_function_p): Declare.
16712 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
16713 with attribute omp declare target for offloading functions.
16715 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
16717 PR tree-optimization/84005
16718 * tree-data-ref.h (get_base_for_alignment): Declare.
16719 * tree-data-ref.c (get_base_for_alignment_1): New function.
16720 (get_base_for_alignment): Likewise.
16721 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
16722 get_base_for_alignment to find a suitable base object, instead
16723 of always using drb->base_address.
16725 2018-03-23 Jakub Jelinek <jakub@redhat.com>
16727 PR inline-asm/85022
16728 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
16729 known size by default.
16731 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
16733 PR inline-asm/85030
16734 * lra-constraints.c (process_alt_operands): Don't match BLKmode
16735 and non BLKmode operands.
16737 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16740 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
16741 Clean up attributes.
16743 2018-03-23 Richard Biener <rguenther@suse.de>
16746 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
16747 we are going to emit early debug for LTO.
16749 2018-03-23 Jakub Jelinek <jakub@redhat.com>
16751 PR inline-asm/85034
16752 * function.c (match_asm_constraints_1): Don't optimize if input
16753 doesn't satisfy general_operand predicate for output's mode.
16755 PR inline-asm/85022
16756 * alias.c (write_dependence_p): Don't require for x_canonicalized
16757 non-VOIDmode if x has VOIDmode.
16760 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
16761 just don't try to optimize it rather than assert it never happens.
16763 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
16765 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
16766 macro expansions for definition of ST_INTERNAL_<mode> and
16767 LD_INTERNAL_<mode> builtins.
16768 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
16770 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
16772 (altivec_expand_st_builtin): Likewise.
16773 (altivec_expand_builtin): Remove calls to deleted functions.
16774 (rs6000_address_for_altivec): Delete this function.
16775 * config/rs6000/vector.md: Remove expands for
16776 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
16778 2018-03-22 Sudakshina Das <sudi.das@arm.com>
16781 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
16782 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
16783 re-computing once computed.
16784 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
16785 (arm_init_machine_status): Initialize
16786 machine->static_chain_stack_bytes.
16788 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
16791 * doc/extend.texi: Add four new prototypes for vec_ld.
16792 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
16793 definitions for more logical presentation.
16794 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
16795 entries for V1TI variants of __builtin_altivec_ld builtin.
16796 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
16797 handling of V1TI variant of LVX icode pattern.
16798 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
16799 (rs6000_gimple_fold_builtin): Likewise.
16800 (altivec_init_builtins): Add code to define
16801 __builtin_altivec_lvx_v1ti function.
16803 2018-03-22 Jakub Jelinek <jakub@redhat.com>
16805 PR inline-asm/84941
16806 * function.c (match_asm_constraints_1): Don't do the optimization
16807 if input isn't a REG, SUBREG, MEM or constant.
16809 2018-03-22 Tom de Vries <tom@codesourcery.com>
16811 PR tree-optimization/84956
16812 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
16813 bb_has_abnormal_pred.
16815 2018-03-22 Jakub Jelinek <jakub@redhat.com>
16818 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
16819 DECL_INITIAL (decl) to decl at the end.
16820 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
16821 adjust the comment.
16823 2018-03-21 Joseph Myers <joseph@codesourcery.com>
16825 * doc/extend.texi (__builtin_tgmath): Document when complex
16826 integer types are treated as _Complex _Float64.
16828 2018-03-21 Tom de Vries <tom@codesourcery.com>
16830 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
16832 2018-03-21 Jakub Jelinek <jakub@redhat.com>
16834 PR tree-optimization/84960
16835 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
16836 if it is ENTRY block, move them into single succ of ENTRY in that case.
16838 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
16840 PR tree-optimization/84811
16841 * poly-int.h (poly_span_traits): Remove the T3 parameter and
16842 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
16843 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
16844 (known_subrange_p): Update accordingly. Cast each value involved
16845 in the size comparison, rather than casting the result of the
16848 2018-03-21 Jakub Jelinek <jakub@redhat.com>
16850 PR tree-optimization/84982
16851 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
16852 by flipping the least significant bit rather than all bits from
16853 bitpos to bitpos + bitsize - 1.
16855 2018-03-21 Nathan Sidwell <nathan@acm.org>
16857 * doc/extend.texi (Deprecated Features): Remove mention of
16858 long-deleted deprecations.
16860 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16863 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
16864 * configure: Regenerate.
16866 2018-03-21 Tom de Vries <tom@codesourcery.com>
16868 PR tree-optimization/83126
16869 * tree-parloops.c (num_phis): New function.
16870 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
16872 2018-03-21 Nathan Sidwell <nathan@acm.org>
16874 * doc/extend.texi (Deprecated Features): Update deprecated flags,
16875 mention anon-struct/union members and trailing attributes.
16877 2018-03-21 Bin Cheng <bin.cheng@arm.com>
16879 PR tree-optimization/84969
16880 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
16881 builtin memset partitions if they set different rhs values.
16883 2018-03-21 Jakub Jelinek <jakub@redhat.com>
16885 PR rtl-optimization/84989
16886 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
16887 VEC_DUPLICATE with scalar result mode.
16889 2018-03-21 Martin Liska <mliska@suse.cz>
16892 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
16893 not intended return statement.
16895 2018-03-21 Martin Liska <mliska@suse.cz>
16898 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
16899 (chkp_find_bound_slots_1): Limit number of iterations.
16901 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
16904 * Minor grammar fixes for x86 options.
16906 2018-03-20 Jakub Jelinek <jakub@redhat.com>
16909 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
16910 holding REG_CFA_RESTORE notes, instead turn them into a USE.
16912 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
16915 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
16916 (altivec_lvx_<mode>_1op): Likewise.
16917 (altivec_stvx_<mode>_2op): Likewise.
16918 (altivec_stvx_<mode>_1op): Likewise.
16919 (altivec_lvx_<VM2:mode>): New define_expand.
16920 (altivec_stvx_<VM2:mode>): Likewise.
16921 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
16922 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
16923 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
16924 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
16925 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
16926 (rs6000_gen_lvx): Likewise.
16927 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
16928 (altivec_expand_stv_builtin): Likewise.
16929 (altivec_expand_builtin): Likewise.
16930 * config/rs6000/vector.md: Likewise.
16932 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16935 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
16938 2018-03-20 Richard Biener <rguenther@suse.de>
16941 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
16942 sign-conversions as zero, fall back to standard scalar_stmt
16945 2018-03-20 Martin Liska <mliska@suse.cz>
16948 * predict.c (rebuild_frequencies): Handle case when we have
16949 PROFILE_ABSENT, but flag_guess_branch_prob is false.
16951 2018-03-20 Jakub Jelinek <jakub@redhat.com>
16954 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
16955 flag_section_anchors.
16956 * varasm.c (use_blocks_for_decl_p): Remove hack for
16957 dw2_force_const_mem.
16960 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
16962 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
16963 be created, use lowpart_subreg of operands[0] rather than operands[0]
16965 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
16966 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
16967 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
16968 and n constraint instead of aarch64_shift_imm_di and Usd.
16969 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
16970 (*aarch64_<optab>_reg_minus<mode>3): ... this.
16972 2018-03-20 Sudakshina Das <sudi.das@arm.com>
16975 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
16976 to favor GPR over NEON registers.
16977 (<shift>di3_neon): Likewise.
16979 2018-03-20 Tom de Vries <tom@codesourcery.com>
16982 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
16983 (nvptx_process_pars): Emit bar.sync asap and alap.
16985 2018-03-20 Tom de Vries <tom@codesourcery.com>
16988 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
16989 seen_label if seen_label is already set.
16991 2018-03-20 Jakub Jelinek <jakub@redhat.com>
16994 * config/i386/i386.c (fold_builtin_cpu): For features above 31
16995 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
16996 Use 1U instead of 1. Formatting fixes.
16999 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
17000 instead of TREE_TYPE (s1) for the return value.
17002 2018-03-19 Jakub Jelinek <jakub@redhat.com>
17004 PR tree-optimization/84946
17005 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
17006 bitsize + bitsize in poly_uint64 rather than poly_int64.
17009 * dwarf2asm.c: Include fold-const.c.
17010 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
17011 of decl rather than decl itself.
17013 PR rtl-optimization/84643
17014 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
17016 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
17019 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
17020 calling assemble_variable.
17022 2018-03-19 Sudakshina Das <sudi.das@arm.com>
17025 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
17026 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
17028 2018-03-19 Jim Wilson <jimw@sifive.com>
17031 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
17032 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
17033 (riscv_first_stack_step): Likewise.
17034 (riscv_option_override): Use STACK_BOUNDARY instead of
17035 MIN_STACK_BOUNDARY.
17036 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
17037 MIN_STACK_BOUNDARY.
17038 (BIGGEST_ALIGNMENT): Set to 128.
17039 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
17040 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
17043 2018-03-19 Richard Biener <rguenther@suse.de>
17045 PR tree-optimization/84933
17046 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
17047 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
17049 2018-03-19 Richard Biener <rguenther@suse.de>
17051 PR tree-optimization/84859
17052 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
17053 (cond_if_else_store_replacement): Perform sinking operation on
17054 single-store BBs regardless of MAX_STORES_TO_SINK setting.
17055 Generalize what a BB with a single eligible store is.
17057 2018-03-19 Richard Biener <rguenther@suse.de>
17059 PR tree-optimization/84929
17060 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
17061 chrec_is_positive against non-chrec arg.
17063 2018-03-19 Tamar Christina <tamar.christina@arm.com>
17066 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
17068 2018-03-18 Martin Liska <mliska@suse.cz>
17070 PR rtl-optimization/84635
17071 * regrename.c (build_def_use): Use matches_mode only when
17074 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
17076 PR tree-optimization/84913
17077 * tree-vect-loop.c (vectorizable_reduction): Don't try to
17078 vectorize chains of COND_EXPRs.
17080 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
17082 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
17084 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
17086 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
17088 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
17090 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
17092 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
17093 Kito Cheng <kito.cheng@gmail.com>
17095 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
17096 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
17097 (nds32_adjust_reg_alloc_order): New function.
17098 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
17100 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
17102 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
17103 nds32_print_operand, nds32_print_operand_address): Use
17104 HOST_WIDE_INT_PRINT_DEC instead.
17106 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
17108 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
17110 2018-03-17 Jakub Jelinek <jakub@redhat.com>
17113 * config/i386/i386.c (initial_ix86_tune_features,
17114 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
17115 unsigned long long.
17116 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
17117 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
17118 rather than 1u << ix86_tune. Formatting fix.
17119 (ix86_option_override_internal): Change ix86_arch_mask from
17120 unsigned int to unsigned HOST_WIDE_INT, initialize to
17121 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
17122 (ix86_function_specific_restore): Likewise.
17124 2018-03-16 Jakub Jelinek <jakub@redhat.com>
17127 * postreload.c (reload_combine_recognize_pattern): Perform
17128 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
17129 truncate_int_for_mode the result for the destination's mode.
17132 * hsa-gen.c (mem_type_for_type): Fix comment typo.
17133 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
17135 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
17138 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
17141 * lra-assigns.c (lra_split_hard_reg_for): Don't use
17142 regno_allocno_class_array and sorted_pseudos.
17143 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
17144 insns where regno is used.
17146 2018-03-16 Martin Liska <mliska@suse.cz>
17149 * multiple_target.c (create_dispatcher_calls): Redirect
17150 reference in the symbol table.
17152 2018-03-16 Martin Liska <mliska@suse.cz>
17155 * multiple_target.c (create_dispatcher_calls): Redirect also
17158 2018-03-16 Jakub Jelinek <jakub@redhat.com>
17162 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
17163 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
17164 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
17166 2018-03-16 Julia Koval <julia.koval@intel.com>
17168 * doc/invoke.texi (Skylake Server): Add CLWB.
17169 Cannonlake): Remove CLWB.
17171 2018-03-16 Jakub Jelinek <jakub@redhat.com>
17173 PR tree-optimization/84841
17174 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
17176 (FLOAT_ONE_CONST_TYPE): Define.
17177 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
17178 (sort_by_operand_rank): Put entries with higher constant_type last
17179 rather than first to match comments.
17181 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
17183 * config/nios2/nios2.md (movsi_internal): Fix thinko in
17186 2018-03-15 Jakub Jelinek <jakub@redhat.com>
17189 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
17190 check and use address of target always.
17192 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
17195 * config/i386/i386.c (indirect_thunk_needed): Update comments.
17196 (indirect_thunk_bnd_needed): Likewise.
17197 (indirect_thunks_used): Likewise.
17198 (indirect_thunks_bnd_used): Likewise.
17199 (indirect_return_needed): New.
17200 (indirect_return_bnd_needed): Likewise.
17201 (output_indirect_thunk_function): Add a bool argument for
17203 (output_indirect_thunk_function): Don't generate alias for
17204 function return thunk.
17205 (ix86_code_end): Call output_indirect_thunk_function to generate
17206 function return thunks.
17207 (ix86_output_function_return): Set indirect_return_bnd_needed
17208 and indirect_return_needed instead of indirect_thunk_bnd_needed
17209 and indirect_thunk_needed.
17211 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
17213 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
17214 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
17215 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
17217 2018-03-15 David Malcolm <dmalcolm@redhat.com>
17218 Paul Hua <paul.hua.gm@gmail.com>
17221 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
17223 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
17225 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
17226 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
17227 resp. SFmode cases.
17229 2018-03-15 Tamar Christina <tamar.christina@arm.com>
17232 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
17233 instead of GET_MODE_SIZE when comparing Units.
17235 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
17238 * varasm.c (hash_section): Return an unchangeble hash value
17239 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
17240 Return !aarch64_can_use_per_function_literal_pools_p ().
17242 2018-03-15 Jakub Jelinek <jakub@redhat.com>
17245 * optabs.c (emit_conditional_move): Pass address of cmode's copy
17246 rather than address of cmode as last argument to prepare_cmp_insn.
17248 2018-03-15 Julia Koval <julia.koval@intel.com>
17250 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
17251 F_AVX512VNNI, F_AVX512BITALG): New.
17253 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
17256 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
17257 insn for floating-point loads and stores.
17259 2018-03-14 Carl Love <cel@us.ibm.com>
17261 * config/rs6000/rs6000-c.c: Add macro definitions for
17262 ALTIVEC_BUILTIN_VEC_PERMXOR.
17263 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
17264 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
17265 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
17266 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
17268 * config/doc/extend.texi: Add prototypes for vec_permxor.
17270 2018-03-14 David Malcolm <dmalcolm@redhat.com>
17273 * diagnostic-show-locus.c (class layout_point): Convert m_line
17274 from int to linenum_type.
17275 (line_span::comparator): Use linenum "compare" function when
17276 comparing line numbers.
17277 (test_line_span): New function.
17278 (layout_range::contains_point): Convert param "row" from int to
17280 (layout_range::intersects_line_p): Likewise.
17281 (layout::will_show_line_p): Likewise.
17282 (layout::print_source_line): Likewise.
17283 (layout::should_print_annotation_line_p): Likewise.
17284 (layout::print_annotation_line): Likewise.
17285 (layout::print_leading_fixits): Likewise.
17286 (layout::annotation_line_showed_range_p): Likewise.
17287 (struct line_corrections): Likewise for field m_row.
17288 (line_corrections::line_corrections): Likewise for param "row".
17289 (layout::print_trailing_fixits): Likewise.
17290 (layout::get_state_at_point): Likewise.
17291 (layout::get_x_bound_for_row): Likewise.
17292 (layout::print_line): Likewise.
17293 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
17294 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
17295 * input.c (selftest::test_linenum_comparisons): New function.
17296 (selftest::input_c_tests): Call it.
17297 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
17298 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
17299 * selftest.h (ASSERT_GT): New macro.
17300 (ASSERT_GT_AT): New macro.
17301 (ASSERT_LT): New macro.
17302 (ASSERT_LT_AT): New macro.
17304 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
17306 PR rtl-optimization/84780
17307 * combine.c (distribute_links): Don't make a link based on pc_rtx.
17309 2018-03-14 Martin Liska <mliska@suse.cz>
17311 * tree.c (record_node_allocation_statistics): Use
17312 get_stats_node_kind.
17313 (get_stats_node_kind): New function extracted from
17314 record_node_allocation_statistics.
17315 (free_node): Use get_stats_node_kind.
17317 2018-03-14 Richard Biener <rguenther@suse.de>
17319 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
17320 that the value-set of ANTIC_IN doesn't grow.
17323 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
17325 (BB_VISITED_WITH_VISITED_SUCCS): New define.
17326 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
17328 2018-03-14 Julia Koval <julia.koval@intel.com>
17330 * config.gcc (icelake-client, icelake-server): New.
17332 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
17333 (initial_ix86_arch_features): Ditto.
17334 (PTA_SKYLAKE): Add SGX.
17335 (PTA_ICELAKE): Remove.
17336 (PTA_ICELAKE_CLIENT): New.
17337 (PTA_ICELAKE_SERVER): New.
17338 (ix86_option_override_internal): Split up icelake on icelake client and
17340 (get_builtin_code_for_version): Ditto.
17341 (fold_builtin_cpu): Ditto.
17342 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
17343 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
17344 * config/i386/i386.h (processor_type): Ditto.
17345 * doc/invoke.texi: Ditto.
17347 2018-03-14 Jakub Jelinek <jakub@redhat.com>
17350 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
17351 INTEGER_CST offset, add it together with bitpos / 8 and
17352 sign extend based on POINTER_SIZE.
17356 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
17359 * config/i386/constraints.md (Yc): New register constraint.
17360 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
17361 Use Yc constraint for alternative 2 of operand 0. Remove
17362 preferred_for_speed attribute.
17364 2018-03-14 Richard Biener <rguenther@suse.de>
17366 PR tree-optimization/84830
17367 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
17368 with the old one to avoid oscillations.
17370 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
17373 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
17375 (assign_by_spills): Return a flag of reload assignment failure.
17376 Do not process the reload assignment failures. Do not spill other
17377 reload pseudos if they has the same reg class. Update n if
17379 (lra_assign): Add a return arg. Set up from the result of
17380 assign_by_spills call.
17381 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17382 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
17383 usage_insns if it is not NULL.
17384 (spill_hard_reg_in_range): New function.
17385 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17386 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17387 function prototypes.
17388 (lra_assign): Change prototype.
17389 * lra.c (lra): Add code to deal with fails by splitting hard reg
17392 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
17394 * config/riscv/riscv.opt (mrelax): New option.
17395 * config/riscv/riscv.c (riscv_file_start): Emit ".option
17396 "norelax" when riscv_mrelax is disabled.
17397 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
17399 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17402 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
17403 reassociation for int modes.
17405 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
17407 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
17408 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
17410 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
17411 * config/aarch64/aarch64-sve.md
17412 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
17413 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
17414 (*extend<mode><Vwide>2): Rename to...
17415 (aarch64_sve_extend<mode><Vwide>2): ...this.
17416 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
17417 renaming the old pattern to...
17418 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
17420 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
17421 define_expand, renaming the old pattern to...
17422 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
17423 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
17424 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
17425 account when deciding which SVE instruction the optab should use.
17426 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
17428 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
17430 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
17431 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
17432 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
17433 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
17434 (tlsdesc_small_<mode>): Turn a define_expand and use
17435 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
17436 (tlsdesc_small_advsimd_<mode>): ...this.
17437 (tlsdesc_small_sve_<mode>): New pattern.
17439 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
17441 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
17442 (UNSPEC_UMUL_HIGHPART): New constants.
17443 (MUL_HIGHPART): New int iteraor.
17444 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
17445 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
17447 (*<su>mul<mode>3_highpart): New define_insn.
17449 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
17452 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
17455 2018-03-13 Martin Liska <mliska@suse.cz>
17458 * (sem_item_optimizer::sem_item_optimizer): Initialize new
17460 (sem_item_optimizer::~sem_item_optimizer): Release it.
17461 (sem_item_optimizer::merge_classes): Register variable aliases.
17462 (sem_item_optimizer::fixup_pt_set): New function.
17463 (sem_item_optimizer::fixup_points_to_sets): Likewise.
17464 * ipa-icf.h: Declare new variables and functions.
17466 2018-03-13 Jakub Jelinek <jakub@redhat.com>
17468 PR middle-end/84834
17469 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
17470 integer_pow2p@2 and test integer_pow2p in condition.
17471 (A < 0 ? C : 0): Similarly for @1.
17473 PR middle-end/84831
17474 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
17475 characters starting at p contain '\0' character, don't look beyond
17479 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
17480 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
17483 * reg-stack.c (change_stack): Change update_end var from int to
17484 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
17485 also call set_block_for_insn on the newly added insns and rescan.
17488 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
17489 on the last operand.
17492 * tree.c (stabilize_reference_1): Return save_expr (e) for
17493 STATEMENT_LIST even if it doesn't have side-effects.
17495 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
17497 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
17499 2018-03-12 Renlin Li <renlin.li@arm.com>
17501 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
17502 aarch64_output_scalar_simd_mov_immediate.
17504 2018-03-12 Martin Sebor <msebor@redhat.com>
17506 PR tree-optimization/83456
17507 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
17508 for perfectly overlapping calls to memcpy.
17509 (gimple_fold_builtin_memory_chk): Same.
17510 (gimple_fold_builtin_strcpy): Handle no-warning.
17511 (gimple_fold_builtin_stxcpy_chk): Same.
17512 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
17514 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
17516 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
17517 parameter. Use it for SFmode.
17518 (rs6000_function_arg_advance_1): Adjust.
17519 (rs6000_function_arg): Adjust.
17520 (rs6000_gimplify_va_arg): Pass false for that new parameter.
17522 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
17524 PR rtl-optimization/84169
17525 PR rtl-optimization/84780
17526 * combine.c (can_combine_p): Check for a 2-insn combination whether
17527 the destination register is used between the two insns, too.
17529 2018-03-12 Richard Biener <rguenther@suse.de>
17531 PR tree-optimization/84803
17532 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
17533 for refs DR analysis didn't process.
17535 2018-03-12 Richard Biener <rguenther@suse.de>
17537 PR tree-optimization/84777
17538 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
17539 force-vectorize loops ignore whether we are optimizing for size.
17541 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
17543 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
17544 (TARGET_MD_ASM_ADJUST): Define.
17546 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
17547 Kito Cheng <kito.cheng@gmail.com>
17548 Chung-Ju Wu <jasonwucj@gmail.com>
17550 * config/nds32/nds32.c (nds32_compute_stack_frame,
17551 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
17552 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
17553 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
17554 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
17555 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
17556 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
17557 * config/nds32/nds32.md (prologue, epilogue): Use macro
17558 NDS32_V3PUSH_AVAILABLE_P to do checking.
17560 2018-03-11 Jakub Jelinek <jakub@redhat.com>
17563 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
17564 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
17565 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
17566 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
17567 addition of most attributes on !orig_type_die or the attribute not
17568 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
17570 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
17571 Chung-Ju Wu <jasonwucj@gmail.com>
17573 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
17574 __NDS32_VH__ macro.
17575 * config/nds32/nds32.opt (mvh): New option.
17577 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
17578 Chung-Ju Wu <jasonwucj@gmail.com>
17580 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
17582 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
17583 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
17586 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
17587 Chung-Ju Wu <jasonwucj@gmail.com>
17589 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
17591 * config/nds32/nds32-multiple.md (strlensi): New pattern.
17592 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
17594 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
17595 Kito Cheng <kito.cheng@gmail.com>
17596 Chung-Ju Wu <jasonwucj@gmail.com>
17598 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
17599 UNSPEC_FFMISM and UNSPEC_FLMISM.
17600 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
17601 for ffb, ffmism and flmism.
17602 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
17603 (unspec_ffmism): Ditto.
17604 (unspec_flmism): Ditto.
17605 (nds32_expand_builtin_impl): Check if string extension is available.
17606 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
17607 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
17609 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
17612 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
17615 * lra-assigns.c (assign_by_spills): Return a flag of reload
17616 assignment failure. Do not process the reload assignment
17617 failures. Do not spill other reload pseudos if they has the same
17619 (lra_assign): Add a return arg. Set up from the result of
17620 assign_by_spills call.
17621 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17622 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
17623 usage_insns if it is not NULL.
17624 (spill_hard_reg_in_range): New function.
17625 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17626 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17627 function prototypes.
17628 (lra_assign): Change prototype.
17629 * lra.c (lra): Add code to deal with fails by splitting hard reg
17632 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
17635 * config/i386/i386.opt: Replace Enforcment with Enforcement.
17637 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
17640 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
17641 (dw_val_node): Add val_symbolic_view.
17642 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
17643 (symview_upper_bound): New.
17644 (new_line_info_table): Initialize symviews_since_reset.
17645 (dwarf2out_source_line): Count symviews_since_reset and set
17646 symview_upper_bound.
17647 (dw_val_equal_p): Handle symview.
17648 (add_AT_symview): New.
17649 (print_dw_val): Handle symview.
17650 (attr_checksum, attr_checksum_ordered): Likewise.
17651 (same_dw_val_p, size_of_die): Likewise.
17652 (value_format, output_die): Likewise.
17653 (add_high_low_attributes): Use add_AT_symview for entry_view.
17654 (dwarf2out_finish): Reset symview_upper_bound, clear
17657 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
17660 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
17661 Add strict argument and use it.
17662 (rs6000_split_multireg_move): Update for new strict argument.
17663 (mem_operand_gpr): Disallow all non-offsettable addresses.
17664 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
17666 2018-03-09 Jakub Jelinek <jakub@redhat.com>
17669 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
17670 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
17671 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
17674 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
17675 decl, use remap_type if we want to use the type.
17677 2018-03-09 Martin Sebor <msebor@redhat.com>
17679 PR tree-optimization/84526
17680 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
17682 (builtin_access::generic_overlap): Be prepared to handle non-array
17685 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
17687 PR rtl-optimization/84682
17688 * lra-constraints.c (process_address_1): Check is_address flag
17689 for address constraints.
17690 (process_alt_operands): Likewise.
17691 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
17692 preprocess_constraints.
17693 * recog.h (preprocess_constraints): Add oploc parameter.
17695 * recog.c (preprocess_constraints): Test address_operand for
17696 CT_ADDRESS constraints.
17698 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
17701 * lra-assigns.c (assign_by_spills): Return a flag of reload
17702 assignment failure. Do not process the reload assignment
17703 failures. Do not spill other reload pseudos if they has the same
17705 (lra_assign): Add a return arg. Set up from the result of
17706 assign_by_spills call.
17707 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17708 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
17709 usage_insns if it is not NULL.
17710 (spill_hard_reg_in_range): New function.
17711 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17712 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17713 function prototypes.
17714 (lra_assign): Change prototype.
17715 * lra.c (lra): Add code to deal with fails by splitting hard reg
17718 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17721 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
17722 Accept complain bool parameter. Only emit errors if it is true.
17723 (arm_parse_cpu_option_name): Likewise.
17724 (arm_target_thumb_only): Adjust callers of the above.
17725 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
17726 prototype to take a default true bool parameter.
17727 (arm_parse_arch_option_name): Likewise.
17729 2018-03-09 David Malcolm <dmalcolm@redhat.com>
17730 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
17734 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
17735 * configure: Regenerate.
17736 * configure.ac ("linker --version-script option"): New.
17737 ("linker soname option"): New.
17739 2018-03-09 Richard Biener <rguenther@suse.de>
17741 PR tree-optimization/84775
17742 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
17743 immediate uses of predicate stmts and mark them modified.
17746 PR tree-optimization/84178
17747 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
17749 (version_loop_for_if_conversion): Delay update_ssa call.
17750 (tree_if_conversion): Delay update_ssa until after predicate
17753 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
17756 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
17757 when the function accesses prior frames.
17759 2018-03-08 Jakub Jelinek <jakub@redhat.com>
17762 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
17763 gen_llsym, otherwise call maybe_gen_llsym.
17765 PR inline-asm/84742
17766 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
17767 has ',' character inside of it.
17769 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17772 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
17773 as clobbering CC_REGNUM.
17775 2018-03-08 Richard Biener <rguenther@suse.de>
17777 PR middle-end/84552
17778 * tree-scalar-evolution.c: Include tree-into-ssa.h.
17779 (follow_copies_to_constant): Do not follow SSA names registered
17782 2018-03-08 Richard Biener <rguenther@suse.de>
17784 PR tree-optimization/84178
17785 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
17787 (version_loop_for_if_conversion): Delay update_ssa call.
17788 (tree_if_conversion): Delay update_ssa until after predicate
17791 2018-03-08 David Malcolm <dmalcolm@redhat.com>
17793 PR tree-optimization/84178
17794 * tree-if-conv.c (release_bb_predicate): Remove the
17795 the assertion that the stmts have NULL use_ops.
17796 Discard the statements, asserting that they haven't
17797 yet been added to a BB.
17799 2018-03-08 Richard Biener <rguenther@suse.de>
17801 PR tree-optimization/84746
17802 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
17803 (phi_translate): Pass in destination ANTIC_OUT set.
17804 (phi_translate_1): Likewise. For a simplified result lookup
17805 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
17806 (phi_translate_set): Adjust.
17807 (do_pre_regular_insertion): Likewise.
17808 (do_pre_partial_partial_insertion): Likewise.
17810 2018-03-08 Martin Liska <mliska@suse.cz>
17812 PR gcov-profile/84735
17813 * doc/gcov.texi: Document usage of profile files.
17814 * gcov-io.h: Document changes in the format.
17816 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
17820 * dwarf2out.c (struct dw_line_info_table): Update comments for
17822 (FORCE_RESET_NEXT_VIEW): New.
17823 (FORCE_RESETTING_VIEW_P): New.
17824 (RESETTING_VIEW_P): Check for -1 too.
17825 (ZERO_VIEW_P): Likewise.
17826 (new_line_info_table): Force-reset next view.
17827 (dwarf2out_begin_function): Likewise.
17828 (dwarf2out_source_line): Simplify zero_view_p initialization.
17829 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
17830 view directly. Omit view when omitting .loc at line 0.
17832 2018-03-08 Jakub Jelinek <jakub@redhat.com>
17834 PR tree-optimization/84740
17835 * tree-switch-conversion.c (process_switch): Call build_constructors
17836 only if info.phi_count is non-zero.
17838 PR tree-optimization/84739
17839 * tree-tailcall.c (find_tail_calls): Check call arguments against
17840 DECL_ARGUMENTS (current_function_decl) rather than
17841 DECL_ARGUMENTS (func) when checking for tail recursion.
17843 2018-03-07 Jakub Jelinek <jakub@redhat.com>
17845 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
17846 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
17847 Volker Reichelt's entry and add entries for people that perform
17848 GCC fuzzy testing and report numerous bugs.
17850 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
17853 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
17854 readonly data in sdata, if that is disabled.
17855 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
17856 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
17857 -mreadonly-in-sdata option.
17859 2018-03-07 Martin Sebor <msebor@redhat.com>
17861 PR tree-optimization/84468
17862 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
17863 basic block when looking for nul assignment.
17865 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
17868 * except.h (output_function_exception_table): Adjust prototype.
17869 * except.c (output_function_exception_table): Remove FNNAME parameter
17870 and add SECTION parameter. Ouput one part of the table at a time.
17871 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
17872 the first part of the exception table and emit unwind directives.
17873 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
17874 (i386_pe_seh_cold_init): Likewise.
17875 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
17876 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
17877 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
17878 (ix86_output_call_insn): Emit a nop in one more case for SEH.
17879 * config/i386/winnt.c: Include except.h.
17880 (struct seh_frame_state): Add reg_offset, after_prologue and
17881 in_cold_section fields.
17882 (i386_pe_seh_end_prologue): Set seh->after_prologue.
17883 (i386_pe_seh_cold_init): New function.
17884 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
17885 to seh->in_cold_section.
17886 (seh_emit_push): Record the offset of the push.
17887 (seh_emit_save): Record the offet of the save.
17888 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
17889 Test seh->after_prologue to disregard the epilogue.
17890 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
17891 (i386_pe_end_cold_function): New function.
17893 2018-03-07 Jakub Jelinek <jakub@redhat.com>
17896 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
17897 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
17900 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
17902 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
17903 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
17905 PR middle-end/84723
17906 * multiple_target.c: Include tree-inline.h and intl.h.
17907 (expand_target_clones): Diagnose and fail if node->definition and
17908 !tree_versionable_function_p (node->decl).
17910 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
17912 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
17914 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
17915 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
17916 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
17918 2018-03-06 Jakub Jelinek <jakub@redhat.com>
17921 * combine.c (try_combine): Use reg_or_subregno instead of handling
17922 just paradoxical SUBREGs and REGs.
17924 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
17926 * config/arc/arc.c (arc_finalize_pic): Remove function.
17927 (arc_must_save_register): We use single base PIC register, remove
17928 checks to save/restore the PIC register.
17929 (arc_expand_prologue): Likewise.
17930 * config/arc/arc-protos.h (arc_set_default_type_attributes):
17932 (arc_verify_short): Likewise.
17933 (arc_attr_type): Likewise.
17934 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
17935 (walk_stores): Likewise.
17936 (arc_address_cost): Make it static.
17937 (arc_verify_short): Likewise.
17938 (branch_dest): Likewise.
17939 (arc_attr_type): Likewise.
17940 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
17941 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
17942 (arc_final_prescan_insn): Remove inserting the nops due to
17943 hardware hazards. It is done in reorg step.
17944 (insn_length_variant_t): Remove.
17945 (insn_length_parameters_t): Likewise.
17946 (arc_insn_length_parameters): Likewise.
17947 (arc_get_insn_variants): Likewise.
17948 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
17950 2018-03-06 Jakub Jelinek <jakub@redhat.com>
17952 PR inline-asm/84683
17953 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
17956 PR tree-optimization/84687
17957 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
17959 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
17961 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17963 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
17964 Rename to ppc_speculation_barrier.
17965 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
17966 __builtin_ppc_speculation_barrier.
17968 2018-03-05 Jakub Jelinek <jakub@redhat.com>
17971 * combine.c (combine_simplify_rtx): Don't try to simplify if
17972 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
17975 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
17977 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
17978 to 32 bytes when compiling for POWER9.
17980 2018-03-05 Jakub Jelinek <jakub@redhat.com>
17983 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
17984 regparm >= 3 with no arg reg available also for calls with
17985 flag_force_indirect_call. Pass decl to ix86_function_regparm.
17988 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
17990 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
17992 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
17995 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
17997 2018-03-05 Richard Biener <rguenther@suse.de>
17999 PR tree-optimization/84486
18000 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
18001 When inserting a __builtin_assume_aligned call set the LHS
18002 SSA name alignment info accordingly.
18004 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
18006 PR tree-optimization/84114
18007 * config/aarch64/aarch64.c (aarch64_reassociation_width)
18008 Avoid reassociation of FLOAT_MODE addition.
18010 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
18012 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
18013 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
18014 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
18015 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
18016 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
18017 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
18018 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
18020 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
18021 __builtin_ia32_wbinvd): New builtins.
18022 (SPECIAL_ARGS2): New.
18023 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
18024 (SPECIAL_ARGS2): New.
18025 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
18026 (ix86_valid_target_attribute_inner_p): Ditto.
18027 (ix86_init_mmx_sse_builtins): Add special_args2.
18028 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
18029 TARGET_WBNOINVD_P): New.
18030 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
18031 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
18032 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
18033 * config/i386/immintrin.h (_wbinvd): New intrinsic.
18034 * config/i386/pconfigintrin.h: New file.
18035 * config/i386/wbnoinvdintrin.h: Ditto.
18036 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
18038 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
18040 2018-03-05 Richard Biener <rguenther@suse.de>
18042 PR tree-optimization/84670
18043 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
18045 (BB_VISITED_WITH_VISITED_SUCCS): New define.
18046 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
18047 (compute_antic_aux): Only assert the number of values in ANTIC_IN
18048 doesn't grow if all successors (recursively) were visited at least
18051 2018-03-05 Richard Biener <rguenther@suse.de>
18053 PR tree-optimization/84650
18054 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
18055 if executed in the loop pipeline.
18057 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
18059 * doc/configfiles.texi (Configuration Files): Move info about
18060 conditionalizing $target-protos.h to...
18061 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
18062 differs from $target-protos.h.
18064 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
18065 Chung-Ju Wu <jasonwucj@gmail.com>
18067 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
18068 * config/nds32/nds32-multiple.md (setmemsi): Define.
18069 * config/nds32/nds32-memory-manipulation.c
18070 (nds32_gen_dup_4_byte_to_word_value): New.
18071 (emit_setmem_word_loop): New.
18072 (emit_setmem_byte_loop): New.
18073 (nds32_expand_setmem_loop): New.
18074 (nds32_expand_setmem_loop_v3m): New.
18075 (nds32_expand_setmem_unroll): New.
18076 (nds32_expand_setmem): New.
18078 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
18079 Chung-Ju Wu <jasonwucj@gmail.com>
18081 * config/nds32/nds32-memory-manipulation.c
18082 (nds32_emit_load_store): New.
18083 (nds32_emit_post_inc_load_store): New.
18084 (nds32_emit_mem_move): New.
18085 (nds32_emit_mem_move_block): New.
18086 (nds32_expand_movmemsi_loop_unknown_size): New.
18087 (nds32_expand_movmemsi_loop_known_size): New.
18088 (nds32_expand_movmemsi_loop): New.
18089 (nds32_expand_movmemsi_unroll): New.
18090 (nds32_expand_movmemqi): Rename ...
18091 (nds32_expand_movmemsi): ... to this.
18092 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
18093 (movmemsi): ... to this.
18094 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
18095 (nds32_expand_movmemsi): ... to this.
18097 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
18098 Monk Chiang <sh.chiang04@gmail.com>
18099 Chung-Ju Wu <jasonwucj@gmail.com>
18101 * config/nds32/nds32-protos.h
18102 (nds32_expand_load_multiple): New arguments.
18103 (nds32_expand_store_multiple): Ditto.
18104 (nds32_valid_multiple_load_store): Rename ...
18105 (nds32_valid_multiple_load_store_p): ... to this.
18106 * config/nds32/nds32-memory-manipulation.c
18107 (nds32_expand_load_multiple): Refine implementation.
18108 (nds32_expand_store_multiple): Ditto.
18109 * config/nds32/nds32-multiple.md
18110 (load_multiple): Update nds32_expand_load_multiple interface.
18111 (store_multiple): Update nds32_expand_store_multiple interface.
18112 * config/nds32/nds32-predicates.c
18113 (nds32_valid_multiple_load_store): Rename ...
18114 (nds32_valid_multiple_load_store_p): ... to this and refine
18116 * config/nds32/predicates.md
18117 (nds32_load_multiple_and_update_address_operation): New predicate.
18118 (nds32_store_multiple_and_update_address_operation): New predicate.
18120 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
18121 Chung-Ju Wu <jasonwucj@gmail.com>
18123 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
18124 (combo): New attribute.
18125 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
18127 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
18129 * config/nds32/nds32.opt: Change -mcmodel= default value.
18131 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
18132 Monk Chiang <sh.chiang04@gmail.com>
18133 Chung-Ju Wu <jasonwucj@gmail.com>
18135 * config/nds32/constants.md (unspec_element): New enum.
18136 * config/nds32/constraints.md (Umw): New constraint.
18137 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
18138 * config/nds32/nds32-intrinsic.md: Likewise.
18139 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
18140 (nds32_valid_smw_lwm_base_p): New.
18141 (nds32_output_smw_single_word): New.
18142 (nds32_output_lmw_single_word): New.
18143 (nds32_expand_unaligned_load): New.
18144 (nds32_expand_unaligned_store): New.
18145 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
18146 (nds32_output_smw_single_word): Declare.
18147 (nds32_output_lmw_single_word): Declare.
18148 (nds32_expand_unaligned_load): Declare.
18149 (nds32_expand_unaligned_store): Declare.
18150 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
18151 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
18152 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
18153 NDS32_BUILTIN_UASTORE_DW.
18154 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
18157 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
18158 Kito Cheng <kito.cheng@gmail.com>
18159 Chung-Ju Wu <jasonwucj@gmail.com>
18161 * config/nds32/nds32-intrinsic.c
18162 (nds32_expand_builtin_null_ftype_reg): Delete.
18163 (nds32_expand_builtin_reg_ftype_imm): Ditto.
18164 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
18165 (nds32_read_argument): New.
18166 (nds32_legitimize_target): Ditto.
18167 (nds32_legitimize_argument): Ditto.
18168 (nds32_check_constant_argument): Ditto.
18169 (nds32_expand_unop_builtin): Ditto.
18170 (nds32_expand_unopimm_builtin): Ditto.
18171 (nds32_expand_binop_builtin): Ditto.
18172 (nds32_builtin_decl_impl): Ditto.
18173 (builtin_description): Ditto.
18174 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
18175 (nds32_init_builtins_impl): Ditto.
18176 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
18177 (nds32_builtin_decl): New.
18178 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
18179 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
18181 2018-03-02 Jeff Law <law@redhat.com>
18183 * reorg.c (stop_search_p): Handle DEBUG_INSN.
18184 (redundant_insn, fill_simple_delay_slots): Likewise.
18185 (fill_slots_from_thread): Likewise.
18186 * resource.c (mark_referenced_resources): Likewise.
18187 (mark_set_resources, find_dead_or_set_registers): Likewise.
18189 2018-03-02 Jakub Jelinek <jakub@redhat.com>
18191 * substring-locations.h (format_warning_va): Formatting fix for
18192 ATTRIBUTE_GCC_DIAG.
18193 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
18195 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
18196 * substring-locations.c: Include intl.h.
18197 (format_warning_va): Turned into small wrapper around
18198 format_warning_n_va, renamed to ...
18199 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
18200 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
18202 (format_warning_at_substring_n): New function.
18203 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
18204 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
18205 format_warning_at_substring with just a shorter name instead of
18206 const function pointer.
18207 (fmtwarn_n): New function.
18208 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
18209 appropriate, get rid of all the fmtstr temporaries, move conditionals
18210 with G_() wrapped string literals directly into fmtwarn arguments,
18211 cast dir.len to (int), formatting fixes.
18213 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
18215 * doc/invoke.texi: Remove "Cilk Plus" references.
18217 2018-03-02 Jakub Jelinek <jakub@redhat.com>
18218 Richard Biener <rguenther@suse.de>
18221 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
18222 for error or warning attributes if CALL_FROM_THUNK_P is set.
18225 2018-03-02 Jakub Jelinek <jakub@redhat.com>
18228 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
18229 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
18232 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
18233 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
18235 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
18236 instead of -1U in last predictors element's probability member.
18238 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
18241 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
18242 arguments if they are comparable.
18244 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
18246 PR tree-optimization/84634
18247 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
18248 masks and masked_loop_p with a single loop_masks, making sure it's
18249 null for bb vectorization.
18251 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
18253 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
18254 (vect_analyze_data_ref_access): Use loop->safe_len rather than
18255 loop->force_vectorize to check whether there is no alias.
18257 2018-03-02 Jakub Jelinek <jakub@redhat.com>
18260 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
18262 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
18264 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
18265 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
18266 instead of a loop around prev_real_insn.
18267 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
18270 PR inline-asm/84625
18271 * config/i386/i386.c (ix86_print_operand): Use conditional
18272 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
18275 2018-03-02 Richard Biener <rguenther@suse.de>
18277 PR tree-optimization/84427
18278 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
18279 (bitmap_set_subtract_values): Rewrite to handle multiple
18282 (prune_clobbered_mems): Likewise.
18283 (phi_translate): Take edge instead of pred/phiblock.
18284 (phi_translate_1): Likewise.
18285 (phi_translate_set): Likewise. Insert all translated
18286 exprs for a value into the set, keeping possibly multiple
18287 expressions per value.
18288 (compute_antic_aux): Adjust for phi_translate changes.
18289 When intersecting union the expressions and prune those
18290 not in the final value set, keeping possibly multiple
18291 expressions per value. Do not use value-insertion
18292 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
18293 all expressions. Add verification that the value-sets
18294 only shrink during iteration.
18295 (compute_partial_antic_aux): Adjust for the phi_translate changes.
18296 (do_pre_regular_insertion): Likewise.
18297 (do_pre_partial_partial_insertion): Likewise.
18299 2018-03-02 Richard Biener <rguenther@suse.de>
18302 * config/darwin.c (saved_debug_info_level): New static global.
18303 (darwin_asm_lto_start): Disable debug info generation for LTO out.
18304 (darwin_asm_lto_end): Restore debug info generation settings.
18306 2018-03-01 Martin Liska <mliska@suse.cz>
18309 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
18310 volatile arguments.
18312 2018-03-01 Richard Biener <rguenther@suse.de>
18315 * dwarf2out.c (gen_variable_die): Properly handle late VLA
18316 type annotation with LTO when debug was disabled at compile-time.
18318 2018-03-01 Matthew Fortune <mfortune@gmail.com>
18320 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
18322 (mips_final_postscan_insn): Likewise.
18324 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
18326 PR rtl-optimization/84528
18327 * alias.c (init_alias_target): Add commentary.
18328 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
18329 a unique base value if the frame pointer is not eliminated
18330 to the stack pointer.
18332 2018-03-01 Tom de Vries <tom@codesourcery.com>
18334 PR rtl-optimization/83327
18335 * lra-int.h (hard_regs_spilled_into): Declare.
18336 * lra.c (hard_regs_spilled_into): Define.
18337 (init_reg_info): Init hard_regs_spilled_into.
18338 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
18339 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
18340 (process_bb_lives): Handle hard_regs_spilled_into.
18341 (lra_create_live_ranges_1): Before doing liveness propagation, clear
18342 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
18344 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
18346 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
18347 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
18348 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
18349 * config/rs6000/aix72.h: New file.
18351 2018-02-28 Jakub Jelinek <jakub@redhat.com>
18353 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
18354 instead of warning_at with conditional singular and plural messages
18358 * stor-layout.c (update_alignment_for_field): For
18359 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
18360 && !DECL_PACKED (field), do the alignment update, just use
18361 only desired_align instead of MAX (type_align, desired_align)
18363 (place_field): Don't do known_align < desired_align handling
18364 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
18365 is non-NULL, instead do it after rli->prev_field handling and
18366 only if not within a bitfield word. For DECL_PACKED (field)
18367 use type_align of BITS_PER_UNIT.
18369 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
18371 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
18372 superfluous parentheses and trailing spaces.
18374 2018-02-28 Richard Biener <rguenther@suse.de>
18376 PR tree-optimization/84584
18377 * graphite-scop-detection.c (scop_detection::add_scop): Discard
18378 SCoPs with fake exit edge.
18380 2018-02-28 Martin Liska <mliska@suse.cz>
18383 * timevar.c (timer::print): Fix format to properly print 100%
18386 2018-02-28 Richard Biener <rguenther@suse.de>
18388 PR middle-end/84607
18389 * genmatch.c (capture_info::walk_match): Do not mark
18390 captured expressions without operands as expr_p given
18391 they act more like predicates and should be subject to
18392 "lost tail" side-effect preserving.
18394 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
18396 PR rtl-optimization/81611
18397 * auto-inc-dec.c (attempt_change): Move dead note from
18398 mem_insn if it's the next use of regno
18399 (find_address): Take address use of reg holding
18400 non-incremented value. Add parm to limit search to the named
18402 (merge_in_block): Attempt to use a mem insn that is the next
18403 use of the original regno.
18405 2018-02-27 Martin Sebor <msebor@redhat.com>
18408 * doc/invoke.texi (-Wmissing-attributes): New option.
18409 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
18411 2018-02-27 Martin Sebor <msebor@redhat.com>
18413 PR translation/84207
18414 * diagnostic-core.h (warning_n, error_n, inform_n): Change
18415 n argument to unsigned HOST_WIDE_INT.
18416 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
18417 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
18418 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
18419 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
18421 2018-02-27 Richard Biener <rguenther@suse.de>
18423 PR tree-optimization/84512
18424 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
18425 Do not use the estimate returned from record_stmt_cost for
18426 the scalar iteration cost but sum properly using add_stmt_cost.
18428 2018-02-27 Richard Biener <rguenther@suse.de>
18430 PR tree-optimization/84466
18431 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
18432 Adjust last change to less strictly validate use operands.
18434 2018-02-27 Martin Liska <mliska@suse.cz>
18436 PR gcov-profile/84548
18437 * gcov.c (process_file): Allow partial overlap and consider it
18438 also as group functions.
18439 (output_lines): Properly calculate range of lines for a group.
18441 2018-02-27 Martin Liska <mliska@suse.cz>
18443 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
18444 'ggc' suffixes. Change first column width.
18445 (timer::print): Fix formatting of the column.
18447 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
18449 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
18450 preserve inline entry blocks for the sake of debug inline
18451 entry point markers alone.
18452 (remove_unused_locals): Suggest in comments a better place to
18453 force the preservation of inline entry blocks that are
18454 otherwise unused, but do not preserve them.
18456 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
18458 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
18460 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
18463 * config/i386/constraints.md (Bs): Replace
18464 ix86_indirect_branch_register with
18465 TARGET_INDIRECT_BRANCH_REGISTER.
18467 * config/i386/i386.md (indirect_jump): Likewise.
18468 (tablejump): Likewise.
18469 (*sibcall_memory): Likewise.
18470 (*sibcall_value_memory): Likewise.
18471 Peepholes of indirect call and jump via memory: Likewise.
18472 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
18473 (*sibcall_value_GOT_32): Likewise.
18474 * config/i386/predicates.md (indirect_branch_operand): Likewise.
18475 (GOT_memory_operand): Likewise.
18476 (call_insn_operand): Likewise.
18477 (sibcall_insn_operand): Likewise.
18478 (GOT32_symbol_operand): Likewise.
18479 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
18481 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
18483 PR rtl-optimization/83496
18484 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
18485 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
18486 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
18487 redundant insn, if any.
18488 (relax_delay_slots): Likewise.
18489 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
18491 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
18493 PR tree-optimization/83965
18494 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
18495 that grouped statements are part of a reduction chain. Return
18496 true if the statement is not marked as a reduction itself but
18497 is part of a group.
18498 (vect_recog_dot_prod_pattern): Don't check whether the statement
18499 is part of a group here.
18500 (vect_recog_sad_pattern): Likewise.
18501 (vect_recog_widen_sum_pattern): Likewise.
18503 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
18506 * final.c (rest_of_clean_state): Also look for calls inside sequences.
18508 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
18511 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
18513 (ix86_output_indirect_function_return): New prototype.
18514 (ix86_split_simple_return_pop_internal): Likewise.
18515 * config/i386/i386.c (indirect_return_via_cx): New.
18516 (indirect_return_via_cx_bnd): Likewise.
18517 (indirect_thunk_name): Handle return va CX_REG.
18518 (output_indirect_thunk_function): Create alias for
18519 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
18520 (ix86_output_indirect_jmp): Remove the bool argument.
18521 (ix86_output_indirect_function_return): New function.
18522 (ix86_split_simple_return_pop_internal): Likewise.
18523 * config/i386/i386.md (*indirect_jump): Don't pass false
18524 to ix86_output_indirect_jmp.
18525 (*tablejump_1): Likewise.
18526 (simple_return_pop_internal): Change it to define_insn_and_split.
18527 Call ix86_split_simple_return_pop_internal to split it for
18528 -mfunction-return=.
18529 (simple_return_indirect_internal): Call
18530 ix86_output_indirect_function_return instead of
18531 ix86_output_indirect_jmp.
18533 2018-02-26 Jakub Jelinek <jakub@redhat.com>
18536 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
18537 memset and value initialization afterwards.
18539 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
18541 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
18543 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
18546 * common/config/aarch64/aarch64-common.c
18547 (aarch_option_optimization_table[]): Switch
18548 off fomit-frame-pointer
18550 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
18551 Chung-Ju Wu <jasonwucj@gmail.com>
18553 * config/nds32/nds32-multiple.md (load_multiple): Disallow
18555 (store_multiple): Ditto.
18557 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
18559 * config.gcc: Add --with-cpu support for nds32 target.
18560 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
18561 * config/nds32/nds32.opt: Add -mcpu= option.
18563 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
18565 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
18566 isel=yes): Warn for these deprecated options.
18568 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
18570 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
18571 ISA_2_5_MASKS_EMBEDDED.
18573 2018-02-23 Jakub Jelinek <jakub@redhat.com>
18575 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
18576 p->max as pointers rather than using iterative_hash_expr.
18578 2018-02-23 Carl Love <cel@us.ibm.com>
18580 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
18581 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
18583 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
18584 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
18585 P8V_BUILTIN_VEC_VUNSIGNED2.
18587 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
18590 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
18591 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
18593 * lra-constraints.c (curr_insn_transform): Set up
18594 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
18596 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
18597 * lra-eliminations.c (spill_pseudos): Ditto.
18598 (process_insn_for_elimination): Ditto.
18599 * lra-lives.c (reg_early_clobber_p): Use the new macros.
18600 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
18601 LRA_NON_CLOBBERED_ALT.
18603 2018-02-22 Martin Sebor <msebor@redhat.com>
18605 PR tree-optimization/84480
18606 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
18607 to maybe_diag_stxncpy_trunc. Call it.
18608 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
18609 from gimple_fold_builtin_strcpy. Print inlining stack.
18610 (handle_builtin_stxncpy): Print inlining stack.
18611 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
18613 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
18616 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
18617 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
18618 and -fcheck-pointer-bounds are used together.
18619 (indirect_thunk_prefix): New enum.
18620 (indirect_thunk_need_prefix): New function.
18621 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
18622 "_nt" instead of "_bnd" for NOTRACK prefix.
18623 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
18624 (output_indirect_thunk_function): Likewise.
18626 (ix86_code_end): Update output_indirect_thunk_function calls.
18627 (ix86_output_indirect_branch_via_reg): Replace
18628 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
18629 (ix86_output_indirect_branch_via_push): Likewise.
18630 (ix86_output_function_return): Likewise.
18631 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
18632 incompatible with -fcf-protection=branch and
18633 -fcheck-pointer-bounds.
18635 2018-02-22 Steve Ellcey <sellcey@cavium.com>
18638 * config/aarch64/aarch64.c (aarch64_print_address_internal):
18639 Change gcc_assert call to output_operand_lossage.
18641 2018-02-22 Steve Ellcey <sellcey@cavium.com>
18643 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
18645 2018-02-22 DJ Delorie <dj@redhat.com>
18646 Sebastian Perta <sebastian.perta@renesas.com>
18647 Oleg Endo <olegendo@gcc.gnu.org>
18649 * config/rx/rx.c (rx_rtx_costs): New function.
18650 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
18652 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
18654 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
18656 2018-02-22 Martin Liska <mliska@suse.cz>
18659 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
18660 Add "native" as a possible value.
18662 2018-02-22 Martin Liska <mliska@suse.cz>
18665 * config/i386/i386.c (ix86_option_override_internal):
18666 Add "native" as a possible value for -march and -mtune.
18668 2018-02-22 Jakub Jelinek <jakub@redhat.com>
18671 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
18672 to all type variants.
18674 PR tree-optimization/84503
18675 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
18676 width as info->bitpos + info->bitsize - start.
18677 (merged_store_group::merge_overlapping): Simplify width computation.
18678 (check_no_overlap): New function.
18679 (imm_store_chain_info::try_coalesce_bswap): Compute expected
18680 start + width and last_order of the group, fail if check_no_overlap
18682 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
18683 to group if check_no_overlap fails.
18685 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
18687 * config/rs6000/altivec.md: Delete contraint arguments to
18688 define_expand, define_split, and define_peephole2, and in
18689 define_insn_and_split if always unused.
18690 * config/rs6000/darwin.md: Ditto.
18691 * config/rs6000/dfp.md: Ditto.
18692 * config/rs6000/rs6000.md: Ditto.
18693 * config/rs6000/sync.md: Ditto.
18694 * config/rs6000/vector.md: Ditto.
18695 * config/rs6000/vsx.md: Ditto.
18697 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
18699 * config/rs6000/altivec.md: Write output control strings as braced
18700 blocks instead of double-quoted strings.
18701 * config/rs6000/darwin.md: Ditto.
18702 * config/rs6000/rs6000.md: Ditto.
18703 * config/rs6000/vector.md: Ditto.
18704 * config/rs6000/vsx.md: Ditto.
18706 2018-02-21 Jason Merrill <jason@redhat.com>
18708 PR c++/84314 - ICE with templates and fastcall attribute.
18709 * attribs.c (build_type_attribute_qual_variant): Remove assert.
18711 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
18713 * ipa-cp.c (determine_versionability): Fix comment typos.
18715 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
18718 * ipa-cp.c (determine_versionability): Do not version functions caling
18721 2018-02-21 Martin Liska <mliska@suse.cz>
18724 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
18725 Add "native" as a possible value.
18726 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
18727 the macro when native cpu detection is available.
18729 2018-02-21 Martin Liska <mliska@suse.cz>
18732 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
18733 Add "native" as a possible value.
18734 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
18735 when native cpu detection is available.
18737 2018-02-21 Jakub Jelinek <jakub@redhat.com>
18738 Martin Sebor <msebor@redhat.com>
18740 PR tree-optimization/84478
18741 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
18743 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
18744 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
18745 support which is conservatively correct, for 2 only stay conservative
18746 for maxlen. Formatting and comment capitalization fixes. Add STRICT
18747 argument to the 2 argument get_range_strlen, adjust 6 arg
18748 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
18750 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
18751 (gimple_fold_builtin_strlen): Pass true as last argument to
18754 2018-02-20 Martin Sebor <msebor@redhat.com>
18756 PR middle-end/84095
18757 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
18758 (builtin_memref::set_base_and_offset): Same. Handle inner references.
18759 (builtin_memref::builtin_memref): Factor out parts into
18760 set_base_and_offset and call it.
18762 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
18764 PR middle-end/84406
18765 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
18766 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
18767 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
18768 search at the associated MODE_INT.
18770 2018-02-20 Jeff Law <law@redhat.com>
18772 PR middle-end/82123
18773 PR tree-optimization/81592
18774 PR middle-end/79257
18775 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
18776 for range data rather than using global data.
18777 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
18778 range data rather than using global data.
18779 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
18780 pass it to children as needed.
18781 (struct directive::fmtresult): Similarly.
18782 (struct directive::set_width): Similarly.
18783 (struct directive::set_precision): Similarly.
18784 (format_integer, format_directive, parse_directive): Similarly.
18785 (format_none): Accept unnamed vr_values parameter.
18786 (format_percent, format_floating, format_character): Similarly.
18787 (format_string, format_plain): Similarly.
18788 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
18789 the EVRP range analyzer for range data rather than using global data.
18790 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
18791 gimple-ssa-evrp-analyze.h
18792 (class sprintf_dom_walker): Add after_dom_children member function.
18793 Add evrp_range_analyzer member.
18794 (sprintf_dom_walker::before_dom_children): Call into the EVRP
18795 range analyzer as needed.
18796 (sprintf_dom_walker::after_dom_children): New member function.
18797 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
18799 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
18800 (evrp_range_analyzer::pop_to_marker): Likewise.
18802 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
18804 PR tree-optimization/84419
18805 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
18806 with the required type if its current type is compatible but
18809 2018-02-20 Jakub Jelinek <jakub@redhat.com>
18811 PR middle-end/82004
18812 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
18813 after vectorization.
18815 2018-02-20 Martin Liska <mliska@suse.cz>
18818 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
18819 possible values if we don't have a hint.
18821 2018-02-20 Martin Liska <mliska@suse.cz>
18825 * final.c (shorten_branches): Build align_tab array with one
18827 * opts.c (finish_options): Add alignment option limit check.
18828 (MAX_CODE_ALIGN): Likewise.
18829 (MAX_CODE_ALIGN_VALUE): Likewise.
18830 * doc/invoke.texi: Document maximum allowed option value for
18831 all -falign-* options.
18833 2018-02-19 Jakub Jelinek <jakub@redhat.com>
18836 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
18837 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
18838 * var-tracking.c (emit_note_insn_var_location): Remove all references
18839 to NOTE_INSN_CALL_ARG_LOCATION.
18840 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
18841 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
18842 Use copy_rtx_if_shared.
18843 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
18844 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
18845 (dwarf2out_var_location): Remove handling of
18846 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
18848 * final.c (final_scan_insn): Remove all references to
18849 NOTE_INSN_CALL_ARG_LOCATION.
18850 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
18851 before dumping final insns.
18852 * except.c (emit_note_eh_region_end): Remove all references to
18853 NOTE_INSN_CALL_ARG_LOCATION.
18854 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
18855 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
18856 * config/arc/arc.c (hwloop_optimize): Likewise.
18857 * config/arm/arm.c (create_fix_barrier): Likewise.
18858 * config/s390/s390.c (s390_chunkify_start): Likewise.
18859 * config/sh/sh.c (find_barrier): Likewise.
18860 * config/i386/i386.c (rest_of_insert_endbranch,
18861 ix86_seh_fixup_eh_fallthru): Likewise.
18862 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
18863 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
18864 * config/frv/frv.c (frv_function_prologue): Likewise.
18865 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
18867 (note_outside_basic_block_p): Remove all references to
18868 NOTE_INSN_CALL_ARG_LOCATION.
18869 * gengtype.c (adjust_field_rtx_def): Likewise.
18870 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
18872 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
18873 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
18876 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
18879 PR tree-optimization/84452
18880 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
18881 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
18884 2018-02-19 Martin Liska <mliska@suse.cz>
18887 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
18889 2018-02-19 Martin Liska <mliska@suse.cz>
18890 Richard Sandiford <richard.sandiford@linaro.org>
18892 PR tree-optimization/82491
18893 * gimple-fold.c (get_base_constructor): Make earlier bail out
18896 2018-02-19 Carl Love <cel@us.ibm.com>
18898 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
18899 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
18901 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
18902 P8V_BUILTIN_VEC_NEG.
18904 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
18906 * config/rl78/rl78.md (movdf): New define expand.
18908 2018-02-19 Martin Liska <mliska@suse.cz>
18911 * doc/invoke.texi: Fix typo.
18912 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
18914 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
18916 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
18917 handle rs6000_single_float and rs6000_double_float specially for
18920 2018-02-16 Jeff Law <law@redhat.com>
18922 * config/rx/rx.c (add_pop_cfi_notes): New function.;
18923 (pop_regs): Use it.
18925 2018-02-16 Jakub Jelinek <jakub@redhat.com>
18928 * ipa-inline.c (inline_small_functions): Fix a typo.
18930 2018-02-16 Nathan Sidwell <nathan@acm.org>
18932 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
18934 2018-02-16 Carl Love <cel@us.ibm.com>
18936 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
18937 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
18938 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
18939 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
18940 expansion to P8V_BUILTIN_VEC_FLOAT2.
18942 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
18944 PR rtl-optimization/70023
18945 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
18946 src_regno into account.
18948 2018-02-16 Carl Love <cel@us.ibm.com>
18950 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
18951 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
18952 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
18953 * config/rs6000/rs6000.c: Remove case statements for
18954 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
18955 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
18956 and P9V_BUILTIN_VEC_VINSERT4B.
18957 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
18958 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
18959 * config/rs6000/vsx.md:
18960 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
18963 2018-02-16 Carl Love <cel@us.ibm.com>
18965 * config/rs6000/altivec.h: Add builtin names vec_extract4b
18967 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
18969 * config/rs6000/rs6000-c.c: Add the definitions for
18970 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
18971 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
18972 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
18973 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
18974 definition for insert4b and define insn *insert3b_internal.
18975 * doc/extend.texi: Add documentation for vec_extract4b.
18977 2018-02-16 Nathan Sidwell <nathan@acm.org>
18979 * doc/extend.texi (Backwards Compatibility): Mention friend
18980 injection. Note for-scope is deprecated.
18981 * doc/invoke.texi (-ffriend-injection): Deprecate.
18983 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
18985 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
18986 that moved to I2, also allow destinations that are a paradoxical
18987 subreg (instead of a normal reg).
18989 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
18992 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
18995 2018-02-16 Richard Biener <rguenther@suse.de>
18997 PR tree-optimization/84037
18998 PR tree-optimization/84016
19000 * config/i386/i386.c (ix86_builtin_vectorization_cost):
19001 Adjust vec_construct for the fact we need additional higher latency
19002 128bit inserts for AVX256 and AVX512 vector builds.
19003 (ix86_add_stmt_cost): Scale vector construction cost for
19006 2018-02-16 Richard Biener <rguenther@suse.de>
19008 PR tree-optimization/84417
19009 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
19010 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
19011 (non_rewritable_lvalue_p): Likewise, use poly-ints.
19013 2018-02-16 Martin Liska <mliska@suse.cz>
19016 * internal-fn.def (ASAN_CHECK): Set proper flags.
19017 (ASAN_MARK): Likewise.
19019 2018-02-16 Julia Koval <julia.koval@intel.com>
19021 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
19022 from PTA_CANNONLAKE.
19024 2018-02-16 Jakub Jelinek <jakub@redhat.com>
19027 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
19028 Use ++iter rather than iter++ for std::list iterators.
19029 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
19030 defer deleting them until all nodes in the forest are processed. Do
19031 free even leaf nodes. Change to_process into auto_vec.
19034 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
19035 * vec.h (vec_default_construct): Use memset instead of placement new
19036 if BROKEN_VALUE_INITIALIZATION is defined.
19037 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
19038 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
19041 PR rtl-optimization/83723
19042 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
19043 * lra.c (lra_substitute_pseudo): Likewise. If true, use
19044 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
19046 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
19048 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
19050 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
19052 PR rtl-optimization/81443
19053 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
19054 from inner REGs to paradoxical SUBREGs.
19056 2018-02-16 Richard Biener <rguenther@suse.de>
19058 PR tree-optimization/84399
19059 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
19060 For operands we can analyze at their definition make sure we can
19061 analyze them at each use as well.
19063 2018-02-16 Richard Biener <rguenther@suse.de>
19065 PR tree-optimization/84190
19066 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
19067 volatile accesses if the decl isn't volatile.
19069 2018-02-15 Jason Merrill <jason@redhat.com>
19071 PR c++/84314 - ICE with templates and fastcall attribute.
19072 * attribs.c (build_type_attribute_qual_variant): Don't clobber
19073 TYPE_CANONICAL on an existing type.
19075 2018-02-15 Jakub Jelinek <jakub@redhat.com>
19077 PR tree-optimization/84383
19078 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
19079 dstoff nor call operand_equal_p if dstbase is NULL.
19081 PR tree-optimization/84334
19082 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
19083 also a CONSTANT_CLASS_P, punt.
19085 2018-02-14 Jim Wilson <jimw@sifive.com>
19087 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
19088 first SMALL_OPERAND check. New local min_second_step. Move assert
19089 to where locals are set. Add TARGET_RVC support.
19090 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
19092 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
19094 * doc/invoke.texi: Correct -Wformat-overflow code sample.
19096 2018-02-14 Martin Sebor <msebor@redhat.com>
19098 PR tree-optimization/83698
19099 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
19100 arrays constrain the offset range to their bounds.
19101 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
19102 (builtin_access::overlap): Avoid setting the size of overlap if it's
19104 (maybe_diag_overlap): Also consider arrays when deciding what values
19105 of offsets to include in diagnostics.
19107 2018-02-14 Martin Sebor <msebor@redhat.com>
19110 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
19111 that correspond to the kind of a declaration.
19113 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
19116 * config/pa/pa.md: Load address of PIC label using the linkage table
19117 if the label is nonlocal.
19119 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
19121 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
19122 warning message if user requests -maltivec=be.
19123 * doc/invoke.texi: Document deprecation of -maltivec=be.
19125 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
19128 * config/rs6000/rs6000-c.c: Update definitions for
19129 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
19130 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
19132 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
19135 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
19136 add _get_ssp intrinsics. Remove argument from
19137 __builtin_ia32_rdssp[d|q].
19138 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
19139 * config/i386/i386-builtin.def: Remove argument from
19140 __builtin_ia32_rdssp[d|q].
19141 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
19142 ix86_expand_special_args_builtin for _rdssp[d|q].
19143 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
19144 Clear register before usage.
19145 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
19146 Add documentation for new _get_ssp and _inc_ssp intrinsics.
19148 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
19150 PR tree-optimization/84357
19151 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
19152 operand 1 of an ARRAY_REF too.
19154 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
19157 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
19158 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
19160 (set_of_reg): New struct.
19161 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
19162 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
19163 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
19165 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
19166 Split into bitclr, bitset, bitinvert patterns if appropriate.
19167 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
19168 use rx_fuse_in_memory_bitop.
19169 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
19170 to named insn, correct maximum insn length.
19172 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
19175 * machmode.def: Define a complex mode for PARTIAL_INT.
19176 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
19178 * doc/rtl.texi: Document CSPImode.
19179 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
19181 (msp430_hard_regno_nregs_with_padding): Likewise.
19183 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
19186 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
19188 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
19190 PR rtl-optimization/84169
19191 * combine.c (try_combine): New variable split_i2i3. Set it to true if
19192 we generated a parallel as new i3 and we split that to new i2 and i3
19193 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
19194 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
19195 those to i2, not i1. Partially rewrite this scan code.
19197 2018-02-13 Jakub Jelinek <jakub@redhat.com>
19200 * stor-layout.c (place_field): For variable length fields, adjust
19201 offset_align afterwards not just based on the field's alignment,
19202 but also on the size.
19204 PR middle-end/84309
19205 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
19206 of exps and logs in the use_exp2 case.
19208 2018-02-13 Jeff Law <law@redhat.com>
19210 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
19211 entry for "vector".
19213 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
19216 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
19220 * common.opt (gas-loc-support, gas-locview-support): New.
19221 (ginline-points, ginternal-reset-location-views): New.
19222 * doc/invoke.texi: Document them. Use @itemx where intended.
19223 (gvariable-location-views): Adjust.
19224 * target.def (reset_location_view): New.
19225 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
19226 (TARGET_RESET_LOCATION_VIEW): New.
19227 * doc/tm.texi: Rebuilt.
19228 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
19229 (dwarf2out_default_as_locview_support): New.
19230 (output_asm_line_debug_info): Use option variables.
19231 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
19232 (output_loc_list): Likewise.
19233 (add_high_low_attributes): Check option variables.
19234 Don't output entry view attribute in strict mode.
19235 (gen_inlined_subroutine_die): Check option variables.
19236 (dwarf2out_inline_entry): Likewise.
19237 (init_sections_and_labels): Likewise.
19238 (dwarf2out_early_finish): Likewise.
19239 (maybe_reset_location_view): New, from...
19240 (dwarf2out_var_location): ... here. Call it.
19241 * debug.h (dwarf2out_default_as_loc_support): Declare.
19242 (dwarf2out_default_as_locview_support): Declare.
19243 * hooks.c (hook_int_rtx_insn_0): New.
19244 * hooks.h (hook_int_rtx_insn_0): Declare.
19245 * toplev.c (process_options): Take -gas-loc-support and
19246 -gas-locview-support from dwarf2out. Enable
19247 -gvariable-location-views by default only with locview
19248 assembler support. Enable -ginternal-reset-location-views by
19249 default only if the target defines the corresponding hook.
19250 Enable -ginline-points by default if location views are
19251 enabled; force it disabled if statement frontiers are
19253 * tree-inline.c (expand_call_inline): Check option variables.
19254 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
19256 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
19258 PR tree-optimization/84321
19259 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
19260 handling. Also check whether the anti-range contains any values
19261 that satisfy the mask; switch to a VR_RANGE if not.
19263 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
19266 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
19268 2018-02-13 Martin Jambor <mjambor@suse.cz>
19271 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
19272 of call statements, also set location of a load to a temporary.
19274 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
19276 * config/rl78/rl78.c (add_vector_labels): New function.
19277 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
19278 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
19279 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
19280 which checks that no arguments are passed.
19281 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
19282 * doc/extend.texi: Documentation for the new attribute.
19284 2018-02-13 Andreas Schwab <schwab@suse.de>
19286 * config/riscv/linux.h (CPP_SPEC): Define.
19288 2018-02-13 Jakub Jelinek <jakub@redhat.com>
19291 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
19292 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
19293 OPTION_MASK_ISA_AES as first argument to def_builtin_const
19294 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
19295 instead of OPTION_MASK_ISA_PCLMUL as first argument to
19296 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
19297 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
19298 temporarily for AES and PCLMUL builtins.
19300 PR tree-optimization/84339
19301 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
19302 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
19305 PR middle-end/84309
19306 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
19307 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
19308 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
19310 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
19312 * omp-simd-clone.h: New file.
19313 * omp-simd-clone.c: Include omp-simd-clone.h.
19314 (expand_simd_clones): No longer static.
19315 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
19316 cgraph.h and omp-simd-clone.h.
19317 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
19318 (vect_recog_widen_shift_pattern): Formatting fix.
19319 (vect_pattern_recog_1): Don't check optab for calls.
19322 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
19323 operands[2] into a REG before using gen_lowpart on it.
19325 2018-02-12 Jeff Law <law@redhat.com>
19328 * config/sh/sh.c (find_barrier): Consider a sibling call
19331 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
19332 successfully back substituting a reg.
19334 2018-02-12 Richard Biener <rguenther@suse.de>
19336 PR tree-optimization/84037
19337 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
19338 parameter, move visited init to caller.
19339 (vect_slp_analyze_operations): Separate cost from validity
19340 check, initialize visited once for all instances.
19341 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
19343 * tree-vect-stmts.c (vect_model_simple_cost): Make early
19345 (vect_model_promotion_demotion_cost): Likewise.
19346 (vectorizable_bswap): Guard cost modeling with !slp_node
19347 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
19349 (vectorizable_call): Likewise.
19350 (vectorizable_conversion): Likewise.
19351 (vectorizable_assignment): Likewise.
19352 (vectorizable_shift): Likewise.
19353 (vectorizable_operation): Likewise.
19354 (vectorizable_store): Likewise.
19355 (vectorizable_load): Likewise.
19356 (vectorizable_condition): Likewise.
19357 (vectorizable_comparison): Likewise.
19359 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
19362 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
19363 (ASAN_MARK): Fix fnspec to account for return value, change pointer
19364 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
19366 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
19368 PR middle-end/83665
19369 * params.def (inline-min-speedup): Increase from 8 to 15.
19370 (max-inline-insns-auto): Decrease from 40 to 30.
19371 * ipa-split.c (consider_split): Add some buffer for function to
19372 be considered inlining candidate.
19373 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
19376 2018-02-12 Richard Biener <rguenther@suse.de>
19378 PR tree-optimization/84037
19379 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
19380 matched stmts if we cannot swap the non-matched ones.
19382 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
19384 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
19385 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
19386 _mm_maskz_scalef_round_ss): New intrinsics.
19387 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
19388 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
19389 __builtin_ia32_scalefss_round): Remove.
19390 (__builtin_ia32_scalefsd_mask_round,
19391 __builtin_ia32_scalefss_mask_round): New intrinsics.
19392 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
19393 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
19394 ((match_operand:VF_128 2 "<round_nimm_predicate>"
19395 "<round_constraint>")): Changed to ...
19396 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
19397 "<round_scalar_constraint>")): ... this.
19398 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
19399 %0, %1, %2<round_op3>}"): Changed to ...
19400 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
19401 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
19402 %2<round_scalar_mask_op3>}"): ... this.
19403 * config/i386/subst.md (round_scalar_nimm_predicate): New.
19405 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
19407 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
19408 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
19409 (_mm_maskz_sqrt_round_ss): New intrinsics.
19410 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
19411 (__builtin_ia32_sqrtsd_mask_round)
19412 (__builtin_ia32_sqrtss_mask_round): New builtins.
19413 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
19414 (__builtin_ia32_sqrtss_round): Remove.
19415 (__builtin_ia32_sqrtsd_mask_round)
19416 (__builtin_ia32_sqrtss_mask_round): New builtins.
19417 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
19418 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
19419 ((match_operand:VF_128 1 "vector_operand"
19420 "xBm,<round_constraint>")): Changed to ...
19421 ((match_operand:VF_128 1 "vector_operand"
19422 "xBm,<round_scalar_constraint>")): ... this.
19423 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
19424 %0, %2, %<iptr>1<round_op3>}): Changed to ...
19425 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
19426 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
19427 %<iptr>1<round_scalar_mask_op3>}): ... this.
19428 ((set_attr "prefix" "<round_prefix>")): Changed to ...
19429 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
19431 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
19434 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
19435 Cast vec_cmpeq result to correct type.
19436 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
19437 Cast vec_cmpgt result to correct type.
19439 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
19441 * final.c (final_scan_insn_1): Renamed from...
19442 (final_scan_insn): ... this. New wrapper, to recover
19443 seen from the outermost call in recursive ones.
19444 * config/sparc/sparc.c (output_return): Drop seen from call.
19445 (output_sibcall): Likewise.
19446 * config/visium/visium.c (output_branch): Likewise.
19448 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
19450 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
19453 2018-02-10 Alan Modra <amodra@gmail.com>
19456 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
19457 Specify LR as an input.
19459 2018-02-10 Jakub Jelinek <jakub@redhat.com>
19462 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
19463 remove_member_access_dummy_vars): New functions.
19464 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
19465 lower_omp_1, execute_lower_omp): Use them.
19467 PR rtl-optimization/84308
19468 * shrink-wrap.c (spread_components): Release todo vector.
19470 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
19472 PR rtl-optimization/57193
19473 * ira-color.c (struct allocno_color_data): Add member
19474 conflict_allocno_hard_prefs.
19475 (update_conflict_allocno_hard_prefs): New.
19476 (bucket_allocno_compare_func): Add a preference based on
19477 conflict_allocno_hard_prefs.
19478 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
19479 (color_allocnos): Remove a dead code. Initiate
19480 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
19482 2018-02-09 Jakub Jelinek <jakub@redhat.com>
19485 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
19486 constraint from =wa to wa. Avoid a subreg on the output operand,
19487 instead use a pseudo and subreg it in a move.
19488 (p9_xxbrd_<mode>): Changed to ...
19489 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
19490 (p9_xxbrd_v2df): New expander.
19491 (p9_xxbrw_<mode>): Changed to ...
19492 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
19493 (p9_xxbrw_v4sf): New expander.
19495 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
19497 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
19499 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
19502 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
19503 multiply in 32-bit mode.
19504 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
19505 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
19508 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
19510 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
19511 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
19512 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
19513 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
19515 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
19518 * dwarf2out.c (is_trivial_indirect_ref): New function.
19519 (dwarf2out_late_global_decl): Do not generate a location
19520 attribute for variables that have a non-trivial DECL_VALUE_EXPR
19521 and that are not defined in the current unit.
19523 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
19525 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
19526 instead of a libcall for UNORDERED.
19528 2018-02-09 Tamar Christina <tamar.christina@arm.com>
19531 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
19532 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
19534 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
19537 * config/s390/s390.c (s390_set_current_function): Invoke
19538 s390_indirect_branch_settings also if fndecl didn't change.
19540 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
19542 * config/rs6000/rs6000.md (blockage): Set length to zero.
19544 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
19546 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
19548 2018-02-09 Jakub Jelinek <jakub@redhat.com>
19551 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
19552 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
19556 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
19557 PARALLEL incoming that failed vt_get_decl_and_offset check.
19559 PR middle-end/84237
19560 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
19561 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
19563 (get_variable_section): For decls in named .bss* sections pass true as
19564 second argument to bss_initializer_p.
19566 2018-02-09 Marek Polacek <polacek@redhat.com>
19567 Jakub Jelinek <jakub@redhat.com>
19570 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
19571 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
19572 Sync some changes from cxx_fold_indirect_ref.
19574 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
19576 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
19578 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
19579 (BLOCK_INLINE_ENTRY_LABEL): New.
19580 (dwarf2out_var_location): Disregard inline entry markers.
19581 (inline_entry_data): New struct.
19582 (inline_entry_data_hasher): New hashtable type.
19583 (inline_entry_data_hasher::hash): New.
19584 (inline_entry_data_hasher::equal): New.
19585 (inline_entry_data_table): New variable.
19586 (add_high_low_attributes): Add DW_AT_entry_pc and
19587 DW_AT_GNU_entry_view attributes if a pending entry is found
19588 in inline_entry_data_table. Add old entry_pc attribute only
19589 if debug nonbinding markers are disabled.
19590 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
19591 markers are enabled.
19592 (block_within_block_p, dwarf2out_inline_entry): New.
19593 (dwarf2out_finish): Check that no entries remained in
19594 inline_entry_data_table.
19595 * final.c (reemit_insn_block_notes): Handle inline entry notes.
19596 (final_scan_insn, notice_source_line): Likewise.
19597 (rest_of_clean_state): Skip inline entry markers.
19598 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
19600 * gimple.c (gimple_build_debug_inline_entry): New.
19601 * gimple.h (enum gimple_debug_subcode): Add
19602 GIMPLE_DEBUG_INLINE_ENTRY.
19603 (gimple_build_debug_inline_entry): Declare.
19604 (gimple_debug_inline_entry_p): New.
19605 (gimple_debug_nonbind_marker_p): Adjust.
19606 * insn-notes.def (INLINE_ENTRY): New.
19607 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
19608 inline entry marker notes.
19609 (print_insn): Likewise.
19610 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
19611 (INSN_DEBUG_MARKER_KIND): Likewise.
19612 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
19613 * tree-inline.c (expand_call_inline): Build and insert
19614 debug_inline_entry stmt.
19615 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
19616 inline entry blocks early, if nonbind markers are enabled.
19617 (dump_scope_block): Dump fragment info.
19618 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
19619 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
19620 (gimple_build_debug_inline_entry): New.
19621 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
19622 Enable/disable inline entry points too.
19623 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
19624 (DEBUG_INSN): Describe inline entry markers.
19626 * common.opt (gvariable-location-views): New.
19627 (gvariable-location-views=incompat5): New.
19628 * config.in: Rebuilt.
19629 * configure: Rebuilt.
19630 * configure.ac: Test assembler for view support.
19631 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
19632 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
19633 * dwarf2out.c (var_loc_view): New typedef.
19634 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
19635 (dwarf2out_locviews_in_attribute): New.
19636 (dwarf2out_locviews_in_loclist): New.
19637 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
19638 (enum dw_line_info_opcode): Add LI_adv_address.
19639 (struct dw_line_info_table): Add view.
19640 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
19641 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
19642 (zero_view_p): New variable.
19643 (ZERO_VIEW_P): New macro.
19644 (output_asm_line_debug_info): New.
19645 (struct var_loc_node): Add view.
19646 (add_AT_view_list, AT_loc_list): New.
19647 (add_var_loc_to_decl): Add view param. Test it against last.
19648 (new_loc_list): Add view params. Record them.
19649 (AT_loc_list_ptr): Handle loc and view lists.
19650 (view_list_to_loc_list_val_node): New.
19651 (print_dw_val): Handle dw_val_class_view_list.
19652 (size_of_die): Likewise.
19653 (value_format): Likewise.
19654 (loc_list_has_views): New.
19655 (gen_llsym): Set vl_symbol too.
19656 (maybe_gen_llsym, skip_loc_list_entry): New.
19657 (dwarf2out_maybe_output_loclist_view_pair): New.
19658 (output_loc_list): Output view list or entries too.
19659 (output_view_list_offset): New.
19660 (output_die): Handle dw_val_class_view_list.
19661 (output_dwarf_version): New.
19662 (output_compilation_unit_header): Use it.
19663 (output_skeleton_debug_sections): Likewise.
19664 (output_rnglists, output_line_info): Likewise.
19665 (output_pubnames, output_aranges): Update version comments.
19666 (output_one_line_info_table): Output view numbers in asm comments.
19667 (dw_loc_list): Determine current endview, pass it to new_loc_list.
19668 Call maybe_gen_llsym.
19669 (loc_list_from_tree_1): Adjust.
19670 (add_AT_location_description): Create view list attribute if
19671 needed, check it's absent otherwise.
19672 (convert_cfa_to_fb_loc_list): Adjust.
19673 (maybe_emit_file): Call output_asm_line_debug_info for test.
19674 (dwarf2out_var_location): Reset views as needed. Precompute
19675 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
19676 attribute. Set view.
19677 (new_line_info_table): Reset next view.
19678 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
19679 (dwarf2out_source_line): Likewise. Output view resets and labels to
19680 the assembler, or select appropriate line info opcodes.
19681 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
19682 (optimize_string_length): Catch it. Adjust.
19683 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
19684 dw_val_class_view_list, and remove it if no longer needed.
19685 (hash_loc_list): Hash view numbers.
19686 (loc_list_hasher::equal): Compare them.
19687 (optimize_location_lists): Check whether a view list symbol is
19688 needed, and whether the locview attribute is present, and
19689 whether they match. Remove the locview attribute if no longer
19691 (index_location_lists): Call skip_loc_list_entry for test.
19692 (dwarf2out_finish): Call output_asm_line_debug_info for test.
19693 Use output_dwarf_version.
19694 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
19695 (struct dw_val_node): Add val_view_list.
19696 * final.c (SEEN_NEXT_VIEW): New.
19697 (set_next_view_needed): New.
19698 (clear_next_view_needed): New.
19699 (maybe_output_next_view): New.
19700 (final_start_function): Rename to...
19701 (final_start_function_1): ... this. Take pointer to FIRST,
19702 add SEEN parameter. Emit param bindings in the initial view.
19703 (final_start_function): Reintroduce SEEN-less interface.
19704 (final): Rename to...
19705 (final_1): ... this. Take SEEN parameter. Output final pending
19706 next view at the end.
19707 (final): Reintroduce seen-less interface.
19708 (final_scan_insn): Output pending next view before switching
19709 sections or ending a block. Mark the next view as needed when
19710 outputting variable locations. Notify debug backend of section
19711 changes, and of location view changes.
19712 (rest_of_handle_final): Adjust.
19713 * toplev.c (process_options): Autodetect value for debug variable
19714 location views option. Warn on incompat5 without -gdwarf-5.
19715 * doc/invoke.texi (gvariable-location-views): New.
19716 (gvariable-location-views=incompat5): New.
19717 (gno-variable-location-views): New.
19719 2018-02-08 David Malcolm <dmalcolm@redhat.com>
19721 PR tree-optimization/84136
19722 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
19723 that the result of find_edge is non-NULL.
19725 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
19728 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
19729 storing integer register in SImode. Fix cost of 256 and 512
19730 byte aligned SSE register store.
19732 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
19734 * config/i386/i386.c (ix86_multiplication_cost): Fix
19735 multiplication cost for TARGET_AVX512DQ.
19737 2018-02-08 Marek Polacek <polacek@redhat.com>
19739 PR tree-optimization/84238
19740 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
19743 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
19745 PR tree-optimization/84265
19746 * tree-vect-stmts.c (vectorizable_store): Don't treat
19747 VMAT_CONTIGUOUS accesses as grouped.
19748 (vectorizable_load): Likewise.
19750 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
19752 PR tree-optimization/81635
19753 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
19754 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
19755 (test_round_for_mask): New functions.
19756 (wide_int_cc_tests): Call test_round_for_mask.
19757 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
19758 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
19759 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
19760 range returned by get_range_info.
19762 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
19765 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
19766 * symtab.c: Include builtins.h
19767 (symtab_node::output_to_lto_symbol_table_p): Move here
19768 from lto-streamer-out.c:output_symbol_p.
19769 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
19770 (output_symbol_p): Move all logic to symtab.c
19771 (produce_symtab): Update.
19773 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
19775 * config/s390/s390-opts.h (enum indirect_branch): Define.
19776 * config/s390/s390-protos.h (s390_return_addr_from_memory)
19777 (s390_indirect_branch_via_thunk)
19778 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
19779 (enum s390_indirect_branch_type): Define.
19780 * config/s390/s390.c (struct s390_frame_layout, struct
19781 machine_function): Remove.
19782 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
19783 (indirect_branch_table_label_no, indirect_branch_table_name):
19785 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
19786 (enum s390_indirect_branch_option): Define.
19787 (s390_return_addr_from_memory): New function.
19788 (s390_handle_string_attribute): New function.
19789 (s390_attribute_table): Add new attribute handler.
19790 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
19791 (s390_indirect_branch_via_thunk): New function.
19792 (s390_indirect_branch_via_inline_thunk): New function.
19793 (s390_function_ok_for_sibcall): When jumping via thunk disallow
19794 sibling call optimization for non z10 compiles.
19795 (s390_emit_call): Force indirect branch target to be a single
19796 register. Add r1 clobber for non-z10 compiles.
19797 (s390_emit_epilogue): Emit return jump via return_use expander.
19798 (s390_reorg): Handle JUMP_INSNs as execute targets.
19799 (s390_option_override_internal): Perform validity checks for the
19800 new command line options.
19801 (s390_indirect_branch_attrvalue): New function.
19802 (s390_indirect_branch_settings): New function.
19803 (s390_set_current_function): Invoke s390_indirect_branch_settings.
19804 (s390_output_indirect_thunk_function): New function.
19805 (s390_code_end): Implement target hook.
19806 (s390_case_values_threshold): Implement target hook.
19807 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
19809 * config/s390/s390.h (struct s390_frame_layout)
19810 (struct machine_function): Move here from s390.c.
19811 (TARGET_INDIRECT_BRANCH_NOBP_RET)
19812 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
19813 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
19814 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
19815 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
19816 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
19817 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
19818 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
19819 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
19820 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
19821 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
19822 (mnemonic attribute): Add values which aren't recognized
19824 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
19825 pattern for branch conversion. Fix mnemonic attribute.
19826 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
19827 indirect branch via thunk if requested.
19828 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
19829 ("*indirect_jump"): Disable for branch conversion using out of
19831 ("indirect_jump_via_thunk<mode>_z10")
19832 ("indirect_jump_via_thunk<mode>")
19833 ("indirect_jump_via_inlinethunk<mode>_z10")
19834 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
19835 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
19836 ("casesi_jump_via_inlinethunk<mode>_z10")
19837 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
19838 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
19839 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
19840 ("*indirect2_jump"): Disable for branch conversion.
19841 ("casesi_jump"): Turn into expander and expand patterns for branch
19843 ("return_use"): New expander.
19844 ("*return"): Emit return via thunk and rename it to ...
19845 ("*return<mode>"): ... this one.
19846 * config/s390/s390.opt: Add new options and and enum for the
19849 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
19851 * lra-constraints.c (match_reload): Unconditionally use
19852 gen_lowpart_SUBREG, rather than selecting between that
19853 and equivalent gen_rtx_SUBREG code.
19855 2018-02-08 Richard Biener <rguenther@suse.de>
19857 PR tree-optimization/84233
19858 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
19859 changed flag instead of boguously re-using phi_inserted.
19861 2018-02-08 Martin Jambor <mjambor@suse.cz>
19863 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
19864 static local variables.
19866 2018-02-08 Richard Biener <rguenther@suse.de>
19868 PR tree-optimization/84278
19869 * tree-vect-stmts.c (vectorizable_store): When looking for
19870 smaller vector types to perform grouped strided loads/stores
19871 make sure the mode is supported by the target.
19872 (vectorizable_load): Likewise.
19874 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
19876 * config/aarch64/aarch64.c (aarch64_components_for_bb):
19877 Increase LDP/STP opportunities by adding adjacent callee-saves.
19879 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
19881 PR rtl-optimization/84068
19882 PR rtl-optimization/83459
19883 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
19885 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
19887 PR tree-optimization/84224
19888 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
19889 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
19890 non-zero arguments.
19892 2018-02-07 Iain Sandoe <iain@codesourcery.com>
19895 * config/rs6000/altivec.md (*restore_world): Remove LR use.
19896 * config/rs6000/predicates.md (restore_world_operation): Adjust op
19897 count, remove one USE.
19899 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
19901 * doc/install.texi (Configuration): Document the
19902 --with-long-double-format={ibm,ieee} PowerPC configuration
19906 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
19907 Convert from define_expand to be define_insn_and_split. Rework
19908 float/double/_Float128 conversions to QI/HI/SImode to work with
19909 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
19910 conversions to QI/HImode types did a store and then a load to
19911 truncate the value. For conversions to VSX registers, don't split
19912 the insn, instead emit the code directly. Use the code iterator
19913 any_fix to combine signed and unsigned conversions.
19914 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
19915 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
19916 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
19917 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
19918 (fix_<mode>di2_hw): Likewise.
19919 (fixuns_<mode>di2_hw): Likewise.
19920 (fix_<mode>si2_hw): Likewise.
19921 (fixuns_<mode>si2_hw): Likewise.
19922 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
19923 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
19924 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
19925 fix<uns>_trunc<SFDF:mode>si2_p8.
19926 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
19927 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
19928 (fix<uns>_<mode>_mem): Likewise.
19929 (fctiw<u>z_<mode>_mem): Likewise.
19930 (fix<uns>_<mode>_mem): Likewise.
19931 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
19932 the register allocator from doing a direct move to the GPRs to do
19933 a store, and instead use the ISA 3.0 store byte/half-word from
19934 vector register instruction. For IEEE 128-bit floating point,
19935 also optimize stores of 32-bit ints.
19936 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
19938 2018-02-07 Alan Hayward <alan.hayward@arm.com>
19940 * genextract.c (push_pathstr_operand): New function to support
19942 (walk_rtx): Call push_pathstr_operand.
19943 (print_path): Support [a-zA-Z].
19945 2018-02-07 Richard Biener <rguenther@suse.de>
19947 PR tree-optimization/84037
19948 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
19949 (cse_and_gimplify_to_preheader): Declare.
19950 (vect_get_place_in_interleaving_chain): Likewise.
19951 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19953 (_loop_vec_info::~_loop_vec_info): Delete it.
19954 (cse_and_gimplify_to_preheader): New function.
19955 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
19956 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
19957 (vectorizable_load): Likewise. For grouped stores always base
19958 the IV on the first element.
19959 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
19960 condition before gimplifying.
19962 2018-02-07 Jakub Jelinek <jakub@redhat.com>
19964 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
19965 *DIV_EXPR and *MOD_EXPR.
19967 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
19970 * config/i386/i386.c (ix86_option_override_internal): Mask out
19971 the CF_SET bit when checking -fcf-protection.
19973 2018-02-07 Tom de Vries <tom@codesourcery.com>
19976 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
19979 2018-02-07 Richard Biener <rguenther@suse.de>
19981 PR tree-optimization/84204
19982 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
19985 PR tree-optimization/84205
19986 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
19987 special-case isl_ast_op_zdiv_r.
19989 PR tree-optimization/84223
19990 * graphite-scop-detection.c (gather_bbs::before_dom_children):
19991 Only add conditions from within the region.
19992 (gather_bbs::after_dom_children): Adjust.
19994 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
19997 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
19998 * config/avr/avr.md: Only post-reload split REG-REG moves if
19999 either register is GENERAL_REG_P.
20001 2018-02-07 Jakub Jelinek <jakub@redhat.com>
20003 PR tree-optimization/84235
20004 * tree-ssa-scopedtables.c
20005 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
20006 if the subtraction is performed in floating point type where NaNs are
20007 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
20008 build 1. Formatting fix.
20010 2018-02-06 Jakub Jelinek <jakub@redhat.com>
20013 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
20014 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
20015 and skip it regardless of bb boundaries. Use CALL_P macro,
20016 don't test INSN_P (insn) together with CALL_P or JUMP_P check
20017 unnecessarily, formatting fix.
20019 2018-02-06 Michael Collison <michael.collison@arm.com>
20021 * config/arm/thumb2.md:
20022 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
20023 (*thumb_mov_notscc): Ditto.
20025 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
20028 * config/rs6000/rs6000.md (su code attribute): Use "u" for
20029 unsigned_fix, not "s".
20031 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20033 * configure.ac (gcc_fn_eh_frame_ro): New function.
20034 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
20035 correct .eh_frame permissions.
20036 * configure: Regenerate.
20038 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
20040 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
20041 irrelevant options.
20043 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20045 * config/rs6000/rs6000.c (rs6000_option_override_internal):
20046 Display warning message for -mno-speculate-indirect-jumps.
20048 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
20050 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
20052 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
20054 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
20056 PR tree-optimization/84225
20057 * tree-eh.c (find_trapping_overflow): Only call
20058 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
20060 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
20063 * config/i386/i386.c: Reimplement the check of possible options
20064 -mibt/-mshstk conbination. Change error messages.
20065 * doc/invoke.texi: Fix a typo: remove extra '='.
20067 2018-02-06 Marek Polacek <polacek@redhat.com>
20069 PR tree-optimization/84228
20070 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
20072 2018-02-06 Tamar Christina <tamar.christina@arm.com>
20075 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
20076 emitted arch directives.
20077 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
20078 __ARM_FEATURE_COPROC before changing architectures.
20080 2018-02-06 Richard Biener <rguenther@suse.de>
20082 * config/i386/i386.c (print_reg): Fix typo.
20083 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
20085 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
20087 * configure: Regenerate.
20089 2018-02-05 Martin Sebor <msebor@redhat.com>
20091 PR tree-optimization/83369
20092 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
20095 2018-02-05 Martin Liska <mliska@suse.cz>
20097 * doc/invoke.texi: Cherry-pick upstream r323995.
20099 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
20101 * ira.c (ira_init_register_move_cost): Adjust comment.
20103 2018-02-05 Martin Liska <mliska@suse.cz>
20105 PR gcov-profile/84137
20106 * doc/gcov.texi: Fix typo in documentation.
20108 2018-02-05 Martin Liska <mliska@suse.cz>
20110 PR gcov-profile/83879
20111 * doc/gcov.texi: Document necessity of --dynamic-list-data when
20112 using dlopen functionality.
20114 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
20116 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
20117 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
20118 _mm_maskz_range_ss, _mm_mask_range_round_ss,
20119 _mm_maskz_range_round_ss): New intrinsics.
20120 (__builtin_ia32_rangesd128_round)
20121 (__builtin_ia32_rangess128_round): Remove.
20122 (__builtin_ia32_rangesd128_mask_round,
20123 __builtin_ia32_rangess128_mask_round): New builtins.
20124 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
20125 __builtin_ia32_rangess128_round): Remove.
20126 (__builtin_ia32_rangesd128_mask_round,
20127 __builtin_ia32_rangess128_mask_round): New builtins.
20128 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
20129 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
20130 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
20131 "<round_saeonly_constraint>")): Changed to ...
20132 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
20133 "<round_saeonly_scalar_constraint>")): ... this.
20134 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
20135 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
20136 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
20137 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
20138 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
20140 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
20142 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
20144 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
20145 Remove all values except native, 8540 and 8548.
20147 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
20149 * config/i386/i386.c (ix86_output_function_return): Pass
20150 INVALID_REGNUM, instead of -1, as invalid register number to
20151 indirect_thunk_name and output_indirect_thunk.
20153 2018-02-02 Julia Koval <julia.koval@intel.com>
20155 * config.gcc: Add -march=icelake.
20156 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
20157 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
20158 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
20159 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
20160 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
20161 (processor_target_table): Add icelake.
20162 (ix86_option_override_internal): Handle new PTAs.
20163 (get_builtin_code_for_version): Handle icelake.
20164 (M_INTEL_COREI7_ICELAKE): New.
20165 (fold_builtin_cpu): Handle icelake.
20166 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
20167 * doc/invoke.texi: Add -march=icelake.
20169 2018-02-02 Julia Koval <julia.koval@intel.com>
20171 * config/i386/i386.c (ix86_option_override_internal): Change flags type
20172 to wide_int_bitmask.
20173 * wide-int-bitmask.h: New.
20175 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
20178 * config/i386/i386.md: Replace Pmode with word_mode in
20179 builtin_setjmp_setup and builtin_longjmp to support x32.
20181 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
20185 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
20187 (rs6000_supported_cpu_names): New static variable.
20188 (linux_cpu_translation_table): Likewise.
20189 (elf_platform) <cpu>: Define new static variable and use it.
20190 Translate kernel AT_PLATFORM name to canonical name if needed.
20191 Error if platform name is unknown.
20193 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
20196 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
20198 2018-02-01 Jeff Law <law@redhat.com>
20201 * config/i386/i386.c (release_scratch_register_on_entry): Add new
20202 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
20203 the scratch if RELEASE_VIA_POP is false.
20204 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
20205 If we have to save a temporary register, decrement SIZE appropriately.
20206 Pass new arguments to release_scratch_register_on_entry.
20207 (ix86_adjust_stack_and_probe): Likewise.
20208 (ix86_emit_probe_stack_range): Pass new arguments to
20209 release_scratch_register_on_entry.
20211 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
20213 PR rtl-optimization/84157
20214 * combine.c (change_zero_ext): Use REG_P predicate in
20215 front of HARD_REGISTER_P predicate.
20217 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
20219 * config/avr/avr.c (avr_option_override): Move disabling of
20220 -fdelete-null-pointer-checks to...
20221 * common/config/avr/avr-common.c (avr_option_optimization_table):
20224 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20226 PR tree-optimization/81635
20227 * tree-data-ref.c (split_constant_offset_1): For types that
20228 wrap on overflow, try to use range info to prove that wrapping
20231 2018-02-01 Renlin Li <renlin.li@arm.com>
20234 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
20235 TAILCALL_ADDR_REGS.
20236 (aarch64_register_move_cost): Likewise.
20237 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
20238 TAILCALL_ADDR_REGS.
20239 (REG_CLASS_NAMES): Likewise.
20240 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
20241 TAILCALL_ADDR_REGS. Remove IP registers.
20242 * config/aarch64/aarch64.md (Ucs): Update register constraint.
20244 2018-02-01 Richard Biener <rguenther@suse.de>
20246 * domwalk.h (dom_walker::dom_walker): Add additional constructor
20247 for specifying RPO order and allow NULL for that.
20248 * domwalk.c (dom_walker::dom_walker): Likewise.
20249 (dom_walker::walk): Handle NULL RPO order.
20250 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
20252 (rewrite_update_dom_walker): Likewise.
20253 (mark_def_dom_walker): Likewise.
20255 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20257 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
20258 (aarch64_maybe_expand_sve_subreg_move): Declare.
20259 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
20260 * config/aarch64/predicates.md (aarch64_any_register_operand): New
20262 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
20263 that are semantically a reverse operation.
20264 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
20265 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
20266 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
20268 (aarch64_can_change_mode_class): For big-endian, forbid changes
20269 between two SVE modes if they have different element sizes.
20271 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20273 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
20274 the TImode handling for big-endian targets.
20276 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20278 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
20279 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
20281 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
20282 Remove BSWAP handing for big-endian targets and use the form of
20283 LD1RQ appropariate for the mode.
20285 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20287 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
20288 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
20289 duplicated element.
20291 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20294 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
20295 check for operands that need to go through aarch64_sve_reload_be.
20297 2018-02-01 Jakub Jelinek <jakub@redhat.com>
20299 PR tree-optimization/81661
20300 PR tree-optimization/84117
20301 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
20302 * tree-eh.c: Include gimplify.h.
20303 (find_trapping_overflow, replace_trapping_overflow,
20304 rewrite_to_non_trapping_overflow): New functions.
20305 * tree-vect-loop.c: Include tree-eh.h.
20306 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
20307 * tree-data-ref.c: Include tree-eh.h.
20308 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
20310 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
20312 PR rtl-optimization/84123
20313 * combine.c (change_zero_ext): Check if hard register satisfies
20314 can_change_dest_mode before calling gen_lowpart_SUBREG.
20316 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
20319 * ira.c (ira_init_register_move_cost): Remove assert.
20321 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
20323 PR rtl-optimization/84071
20324 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
20325 * doc/tm.texi: Regenerate.
20327 2018-01-31 Richard Biener <rguenther@suse.de>
20329 PR tree-optimization/84132
20330 * tree-data-ref.c (analyze_miv_subscript): Properly
20331 check whether evolution_function_is_affine_multivariate_p
20332 before calling gcd_of_steps_may_divide_p.
20334 2018-01-31 Julia Koval <julia.koval@intel.com>
20337 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
20338 * config/i386/i386.md (rdpid_rex64) New.
20339 (rdpid): Make 32bit only.
20341 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
20344 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
20345 an IDENTIFIER_NODE for FUNCTION_TYPE's.
20347 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
20350 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
20352 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
20354 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
20356 PR rtl-optimization/84071
20357 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
20358 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
20360 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
20362 * config/arc/arc.c (arc_handle_aux_attribute): New function.
20363 (arc_attribute_table): Add 'aux' attribute.
20364 (arc_in_small_data_p): Consider aux like variables.
20365 (arc_is_aux_reg_p): New function.
20366 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
20367 (arc_get_aux_arg): New function.
20368 (prepare_move_operands): Handle aux-register access.
20369 (arc_handle_aux_attribute): New function.
20370 * doc/extend.texi (ARC Variable attributes): Add subsection.
20372 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
20374 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
20375 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
20376 (arc_attribute_table): Add 'uncached' attribute.
20377 (arc_print_operand): Print '.di' flag for uncached memory
20379 (arc_in_small_data_p): Do not consider for small data the uncached
20381 (arc_is_uncached_mem_p): New function.
20382 * config/arc/predicates.md (compact_store_memory_operand): Check
20383 for uncached memory accesses.
20384 (nonvol_nonimm_operand): Likewise.
20385 * doc/extend.texi (ARC Type Attribute): New subsection.
20387 2018-01-31 Jakub Jelinek <jakub@redhat.com>
20390 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
20391 falign-loops=): Add Optimization flag.
20393 2018-01-30 Jeff Law <law@redhat.com>
20396 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
20397 INT_REGISTERS_SAVED. Check it prior to calling
20398 get_scratch_register_on_entry.
20399 (ix86_adjust_stack_and_probe): Similarly.
20400 (ix86_emit_probe_stack_range): Similarly.
20401 (ix86_expand_prologue): Corresponding changes.
20403 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20406 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
20407 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
20409 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
20412 * lra-constraints.c (curr_insn_transform): Process AND in the
20415 2018-01-30 Jakub Jelinek <jakub@redhat.com>
20417 PR rtl-optimization/83986
20418 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
20419 dependence against last_pending_memory_flush in addition to
20420 pending_jump_insns.
20422 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
20424 PR tree-optimization/81611
20425 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
20428 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20431 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
20434 2018-01-30 Richard Biener <rguenther@suse.de>
20435 Jakub Jelinek <jakub@redhat.com>
20437 PR tree-optimization/84111
20438 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
20439 inner loops added during recursion, as they don't have up-to-date
20442 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
20445 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
20446 (can_inline_edge_by_limits_p): ... here.
20447 (can_early_inline_edge_p, check_callers,
20448 update_caller_keys, update_callee_keys, recursive_inlining,
20449 add_new_edges_to_heap, speculation_useful_p,
20450 inline_small_functions,
20451 inline_small_functions, flatten_function,
20452 inline_to_all_callers_1): Update.
20454 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
20456 * profile-count.c (profile_count::combine_with_ipa_count): Handle
20459 2018-01-30 Richard Biener <rguenther@suse.de>
20461 PR tree-optimization/83008
20462 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
20463 invariant and constant vector uses in stmts when they need
20464 more than one stmt.
20466 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20469 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
20470 * configure: Regenerate.
20472 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
20474 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
20476 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
20477 Use gen_rtx_REG rather than gen_lowpart.
20479 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
20481 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
20482 rather than 0 when creating partial subregs.
20484 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
20486 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
20489 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
20492 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
20493 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
20494 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
20495 flags. This restores the settings used before the 2017-07-24.
20496 Turning off pre increment/decrement/modify allows IVOPTS to
20497 optimize DF/SF loops where the index is an int.
20499 2018-01-29 Richard Biener <rguenther@suse.de>
20500 Kelvin Nilsen <kelvin@gcc.gnu.org>
20503 * tree-vect-stmts.c (vectorizable_call): Don't call
20504 targetm.vectorize_builtin_md_vectorized_function if callee is
20507 2018-01-22 Carl Love <cel@us.ibm.com>
20509 * doc/extend.tex: Fix typo in second arg in
20510 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
20512 2018-01-29 Richard Biener <rguenther@suse.de>
20514 PR tree-optimization/84086
20515 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
20516 (flush_ssaname_freelist): When SSA names were released reset
20517 the SCEV hash table.
20519 2018-01-29 Richard Biener <rguenther@suse.de>
20521 PR tree-optimization/84057
20522 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
20523 removed paths when removing edges.
20525 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
20527 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
20528 -mfunction-return=@var{choice}.
20530 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
20532 PR diagnostic/84034
20533 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
20534 Handle CR like TAB.
20535 (layout::print_source_line): Likewise.
20536 (test_get_line_width_without_trailing_whitespace): Add test cases.
20538 2018-01-27 Jakub Jelinek <jakub@redhat.com>
20540 PR middle-end/84040
20541 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
20544 2018-01-26 Jim Wilson <jimw@sifive.com>
20546 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
20548 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
20551 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20553 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
20554 and CMP + SUB-immediate -> SUBS.
20556 2018-01-26 Martin Sebor <msebor@redhat.com>
20558 PR tree-optimization/83896
20559 * tree-ssa-strlen.c (get_string_len): Rename...
20560 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
20561 Avoid assuming length is constant.
20562 (handle_char_store): Use HOST_WIDE_INT for string length.
20564 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
20567 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
20568 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
20570 2018-01-26 Richard Biener <rguenther@suse.de>
20572 PR rtl-optimization/84003
20573 * dse.c (record_store): Only record redundant stores when
20574 the earlier store aliases at least all accesses the later one does.
20576 2018-01-26 Jakub Jelinek <jakub@redhat.com>
20578 PR rtl-optimization/83985
20579 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
20580 REG_CFA_RESTORE insns.
20581 (delete_unmarked_insns): Don't ignore separate shrink wrapping
20582 REG_CFA_RESTORE insns here.
20585 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
20586 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
20588 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20590 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
20591 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
20592 (arc_init): Likewise.
20593 (arc_override_options): Likewise.
20594 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
20596 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
20598 * config/arc/arc.h (TARGET_DBNZ): Define.
20599 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
20600 properly set the tune attribute.
20601 (dbnz): Use TARGET_DBNZ guard.
20602 * config/arc/arc.opt (mtune): Add core3 option.
20604 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20606 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
20607 recognize new pic like addresses.
20608 (arc_delegitimize_address): Clean up.
20610 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20612 * config/arc/arc-arches.def: Option mrf16 valid for all
20614 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
20615 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
20616 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
20617 * config/arc/arc-tables.opt: Regenerate.
20618 * config/arc/arc.c (arc_conditional_register_usage): Handle
20619 reduced register file case.
20620 (arc_file_start): Set must have build attributes.
20621 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
20622 mrf16 option value.
20623 * config/arc/arc.opt (mrf16): Add new option.
20624 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
20625 * config/arc/genmultilib.awk: Handle new mrf16 option.
20626 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
20627 * config/arc/t-multilib: Regenerate.
20628 * doc/invoke.texi (ARC Options): Document mrf16 option.
20630 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20632 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
20633 * config/arc/arc.c (arc_handle_secure_attribute): New function.
20634 (arc_attribute_table): Add 'secure_call' attribute.
20635 (arc_print_operand): Print secure call operand.
20636 (arc_function_ok_for_sibcall): Don't optimize tail calls when
20638 (arc_is_secure_call_p): New function. * config/arc/arc.md
20639 (call_i): Add support for sjli instruction.
20640 (call_value_i): Likewise.
20641 * config/arc/constraints.md (Csc): New constraint.
20643 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20644 John Eric Martin <John.Martin@emmicro-us.com>
20646 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
20647 * config/arc/arc.c (_arc_jli_section): New struct.
20648 (arc_jli_section): New type.
20649 (rc_jli_sections): New static variable.
20650 (arc_handle_jli_attribute): New function.
20651 (arc_attribute_table): Add jli_always and jli_fixed attribute.
20652 (arc_file_end): New function.
20653 (TARGET_ASM_FILE_END): Define.
20654 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
20655 (arc_add_jli_section): New function.
20656 (jli_call_scan): Likewise.
20657 (arc_reorg): Call jli_call_scan.
20658 (arc_output_addsi): Remove 'S' from printing asm operand.
20659 (arc_is_jli_call_p): New function.
20660 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
20662 (movhi_insn): Likewise.
20663 (movsi_insn): Likewise.
20664 (movsi_set_cc_insn): Likewise.
20665 (loadqi_update): Likewise.
20666 (load_zeroextendqisi_update): Likewise.
20667 (load_signextendqisi_update): Likewise.
20668 (loadhi_update): Likewise.
20669 (load_zeroextendhisi_update): Likewise.
20670 (load_signextendhisi_update): Likewise.
20671 (loadsi_update): Likewise.
20672 (loadsf_update): Likewise.
20673 (movsicc_insn): Likewise.
20674 (bset_insn): Likewise.
20675 (bxor_insn): Likewise.
20676 (bclr_insn): Likewise.
20677 (bmsk_insn): Likewise.
20678 (bicsi3_insn): Likewise.
20679 (cmpsi_cc_c_insn): Likewise.
20680 (movsi_ne): Likewise.
20681 (movsi_cond_exec): Likewise.
20682 (clrsbsi2): Likewise.
20683 (norm_f): Likewise.
20693 (call_i): Remove 'S' asm letter, add jli instruction.
20694 (call_value_i): Likewise.
20695 * config/arc/arc.op (mjli-always): New option.
20696 * config/arc/constraints.md (Cji): New constraint.
20697 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
20699 (subsf3_fpx): Likewise.
20700 (mulsf3_fpx): Likewise.
20701 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
20703 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
20704 function attrbutes.
20705 * doc/invoke.texi (ARC): Document mjli-always option.
20707 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
20709 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
20710 avoid addition with 0 and use incw and decw where possible.
20712 2018-01-26 Richard Biener <rguenther@suse.de>
20714 PR tree-optimization/81082
20715 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
20716 association if it requires casting to unsigned.
20717 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
20718 from fold_plusminus_mult_expr to catch important cases late when
20719 range info is available.
20721 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20723 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
20724 * configure.ac (hidden_linkonce): New test.
20725 * configure: Regenerate.
20726 * config.in: Regenerate.
20728 2018-01-26 Julia Koval <julia.koval@intel.com>
20730 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
20731 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
20732 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
20733 _mm_mask_bitshuffle_epi64_mask): Fix type.
20734 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
20735 USI_FTYPE_V4DI_V4DI_USI): Remove.
20736 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
20737 __builtin_ia32_vpshufbitqmb256_mask,
20738 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
20739 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
20740 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
20742 2018-01-26 Alan Modra <amodra@gmail.com>
20745 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
20746 UNSPEC_VBPERMQ. Sort other unspecs.
20748 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
20750 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
20752 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
20754 PR middle-end/83055
20755 * predict.c (drop_profile): Do not push/pop cfun; update also
20757 (handle_missing_profiles): Fix logic looking for zero profiles.
20759 2018-01-25 Jakub Jelinek <jakub@redhat.com>
20761 PR middle-end/83977
20762 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
20763 on functions with #pragma omp declare simd or functions with simd
20765 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
20766 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
20767 Remove trailing \n from warning_at calls.
20769 2018-01-25 Tom de Vries <tom@codesourcery.com>
20772 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
20773 for neutered workers.
20775 2018-01-24 Joseph Myers <joseph@codesourcery.com>
20778 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
20779 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
20781 2018-01-24 Jeff Law <law@redhat.com>
20784 * i386.c (get_probe_interval): Move to earlier point.
20785 (ix86_compute_frame_layout): If -fstack-clash-protection and
20786 the frame is larger than the probe interval, then use pushes
20787 to save registers rather than reg->mem moves.
20788 (ix86_expand_prologue): Remove conditional for int_registers_saved
20791 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
20794 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
20795 min/max for never referenced object.
20797 2018-01-24 Jakub Jelinek <jakub@redhat.com>
20799 PR middle-end/83977
20800 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
20802 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
20803 attributes from DECL_ATTRIBUTES (decl) without affecting
20804 DECL_ATTRIBUTES (current_function_decl).
20805 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
20806 functions with non-NULL DECL_ABSTRACT_ORIGIN.
20808 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
20810 PR tree-optimization/83979
20811 * fold-const.c (fold_comparison): Use constant_boolean_node
20812 instead of boolean_{true,false}_node.
20814 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
20816 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
20819 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20821 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
20822 Simplify the clause that sets the length attribute.
20823 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
20824 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
20825 clause that sets the length attribute.
20826 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
20828 2018-01-24 Tom de Vries <tom@codesourcery.com>
20831 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
20832 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
20833 Add strict parameter.
20834 (prevent_branch_around_nothing): Insert dummy insn between branch to
20835 label and label with no ptx insn inbetween.
20836 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
20838 2018-01-24 Tom de Vries <tom@codesourcery.com>
20841 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
20842 for neutered threads in warp.
20843 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
20845 2018-01-24 Richard Biener <rguenther@suse.de>
20847 PR tree-optimization/83176
20848 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
20851 2018-01-24 Richard Biener <rguenther@suse.de>
20853 PR tree-optimization/82819
20854 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
20855 code generating pluses that are no-ops in the target precision.
20857 2018-01-24 Richard Biener <rguenther@suse.de>
20859 PR middle-end/84000
20860 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
20862 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
20864 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
20865 to merge probabilities.
20866 * predict.c (probably_never_executed): Also mark as cold functions
20867 with global 0 profile and guessed local profile.
20868 * profile-count.c (profile_probability::combine_with_count): New
20870 * profile-count.h (profile_probability::operator*,
20871 profile_probability::operator*=, profile_probability::operator/,
20872 profile_probability::operator/=): Reduce precision to adjusted
20873 and set value to guessed on contradictory divisions.
20874 (profile_probability::combine_with_freq): Remove.
20875 (profile_probability::combine_wiht_count): Declare.
20876 (profile_count::force_nonzero):: Set to adjusted.
20877 (profile_count::probability_in):: Set quality to adjusted.
20878 * tree-ssa-tail-merge.c (replace_block_by): Use
20879 combine_with_count.
20881 2018-01-23 Andrew Waterman <andrew@sifive.com>
20882 Jim Wilson <jimw@sifive.com>
20884 * config/riscv/riscv.c (riscv_stack_boundary): New.
20885 (riscv_option_override): Set riscv_stack_boundary. Handle
20886 riscv_preferred_stack_boundary_arg.
20887 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
20888 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
20889 (STACK_BOUNDARY): Set to riscv_stack_boundary.
20890 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
20891 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
20892 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
20894 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
20897 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
20898 of struct ix86_frame.
20899 (ix86_expand_epilogue): Likewise. Add a local variable for
20900 the reg_save_offset field in struct ix86_frame.
20902 2018-01-23 Bin Cheng <bin.cheng@arm.com>
20904 PR tree-optimization/82604
20905 * tree-loop-distribution.c (enum partition_kind): New enum item
20906 PKIND_PARTIAL_MEMSET.
20907 (partition_builtin_p): Support above new enum item.
20908 (generate_code_for_partition): Ditto.
20909 (compute_access_range): Differentiate cases that equality can be
20910 proven at all loops, the innermost loops or no loops.
20911 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
20912 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
20913 (finalize_partitions, distribute_loop): Don't fuse partition of
20914 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
20915 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
20916 parloop is enabled.
20918 2018-01-23 Martin Liska <mliska@suse.cz>
20920 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
20921 order to ignore the predictor.
20922 (PRED_POLYMORPHIC_CALL): Likewise.
20923 (PRED_RECURSIVE_CALL): Likewise.
20925 2018-01-23 Martin Liska <mliska@suse.cz>
20927 * tree-profile.c (tree_profiling): Print function header to
20928 aware reader which function we are working on.
20929 * value-prof.c (gimple_find_values_to_profile): Do not print
20930 not interesting value histograms.
20932 2018-01-23 Martin Liska <mliska@suse.cz>
20934 * profile-count.h (enum profile_quality): Add
20935 profile_uninitialized as the first value. Do not number values
20936 as they are zero based.
20937 (profile_count::verify): Update sanity check.
20938 (profile_probability::verify): Likewise.
20940 2018-01-23 Nathan Sidwell <nathan@acm.org>
20942 * doc/invoke.texi (ffor-scope): Deprecate.
20944 2018-01-23 David Malcolm <dmalcolm@redhat.com>
20946 PR tree-optimization/83510
20947 * domwalk.c (set_all_edges_as_executable): New function.
20948 (dom_walker::dom_walker): Convert bool param
20949 "skip_unreachable_blocks" to enum reachability. Move setup of
20950 edge flags to set_all_edges_as_executable and only do it when
20951 reachability is REACHABLE_BLOCKS.
20952 * domwalk.h (enum dom_walker::reachability): New enum.
20953 (dom_walker::dom_walker): Convert bool param
20954 "skip_unreachable_blocks" to enum reachability.
20955 (set_all_edges_as_executable): New decl.
20956 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
20957 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
20959 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
20960 but converting true to REACHABLE_BLOCKS.
20961 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
20963 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
20964 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
20965 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
20967 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
20968 if check_all_array_refs will be called.
20970 2018-01-23 David Malcolm <dmalcolm@redhat.com>
20972 * tree.c (selftest::test_location_wrappers): Add more test
20975 2018-01-23 David Malcolm <dmalcolm@redhat.com>
20977 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
20978 (selftest::test_bit_in_range): Likewise.
20980 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
20983 * doc/sourcebuild.texi (vect_float): Say that the selector
20984 only describes the situation when -funsafe-math-optimizations is on.
20985 (vect_float_strict): Document.
20987 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
20989 PR tree-optimization/83965
20990 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
20991 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
20992 instead of checking only for a reduction.
20993 (vect_recog_widen_sum_pattern): Likewise.
20995 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
20997 * predict.c (probably_never_executed): Only use precise profile info.
20998 (compute_function_frequency): Skip after inlining hack since we now
20999 have quality checking.
21001 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
21003 * profile-count.h (profile_probability::very_unlikely,
21004 profile_probability::unlikely, profile_probability::even): Set
21005 precision to guessed.
21007 2018-01-23 Richard Biener <rguenther@suse.de>
21009 PR tree-optimization/83963
21010 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
21011 Properly terminate dominator walk when crossing the exit edge not
21012 when visiting its source block.
21014 2018-01-23 Jakub Jelinek <jakub@redhat.com>
21017 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
21018 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
21020 2018-01-22 Jakub Jelinek <jakub@redhat.com>
21022 PR tree-optimization/83957
21023 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
21024 semicolon after for body surrounded by braces.
21026 PR tree-optimization/83081
21027 * profile-count.h (profile_probability::split): New method.
21028 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
21029 Use profile_probability::split.
21030 (do_compare_rtx_and_jump): Fix adjustment of probabilities
21031 when splitting a single conditional jump into 2.
21033 2018-01-22 David Malcolm <dmalcolm@redhat.com>
21035 PR tree-optimization/69452
21036 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
21039 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21041 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
21042 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
21043 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
21045 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21047 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
21049 * config/rl78/rl78.md (movdi): New define_expand.
21050 * config/rl78/rl78.c (rl78_split_movdi): New function.
21052 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
21055 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
21057 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
21058 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
21059 128-bit to produce an UNSPEC move to get the double word with the
21060 signbit and then a shift directly to do signbit.
21061 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
21062 implementation with a new version that just does either a direct
21063 move or a regular move. Move memory interface to separate insns.
21064 Move insns so they are next to the expander.
21065 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
21066 with signbit move. Split big and little endian case.
21067 (signbit<mode>2_dm_mem_le): Likewise.
21068 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
21069 (signbit<mode>2_dm2): Likewise.
21071 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21073 * config/rl78/rl78.md (anddi3): New define_expand.
21075 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21077 * config/rl78/rl78.md (umindi3): New define_expand.
21079 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21081 * config/rl78/rl78.md (smindi3): New define_expand.
21083 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21085 * config/rl78/rl78.md (smaxdi3): New define_expand.
21087 2018-01-22 Carl Love <cel@us.ibm.com>
21089 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
21090 LVX_V1TI): Add macro expansion.
21091 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
21092 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
21093 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
21094 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
21095 Change check to determine if the instruction is a byte reversing
21096 entry. Fix typo in comment.
21097 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
21098 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
21099 Add def_builtin calls for new builtins.
21100 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
21101 Add define_insn expansion.
21103 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21105 * config/rl78/rl78.md (umaxdi3): New define_expand.
21107 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21109 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
21110 for non-QImode registers.
21112 2018-01-22 Richard Biener <rguenther@suse.de>
21114 PR tree-optimization/83963
21115 * graphite-scop-detection.c (scop_detection::get_sese): Delay
21116 including the loop exit block.
21117 (scop_detection::merge_sese): Likewise.
21118 (scop_detection::add_scop): Do it here instead.
21120 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21122 * doc/sourcebuild.texi (arm_softfloat): Document.
21124 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
21127 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
21128 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
21129 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
21131 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21132 David Edelsohn <dje.gcc@gmail.com>
21135 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
21136 Change "crset eq" to "crset 2".
21137 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
21138 (*call_indirect_aix<mode>_nospec): Likewise.
21139 (*call_value_indirect_aix<mode>_nospec): Likewise.
21140 (*call_indirect_elfv2<mode>_nospec): Likewise.
21141 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
21142 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
21143 change assembly output from . to $.
21144 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
21145 (indirect_jump<mode>_nospec): Change assembly output from . to $.
21146 (*tablejump<mode>_internal1_nospec): Likewise.
21148 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
21151 * config/sh/sh_optimize_sett_clrt.cc:
21152 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
21154 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
21156 PR tree-optimization/83940
21157 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
21158 offset_dt to vect_constant_def rather than vect_unknown_def_type.
21159 (vect_check_load_store_mask): Add a mask_dt_out parameter and
21160 use it to pass back the definition type.
21161 (vect_check_store_rhs): Likewise rhs_dt_out.
21162 (vect_build_gather_load_calls): Add a mask_dt argument and use
21163 it instead of a call to vect_is_simple_use.
21164 (vectorizable_store): Update calls to vect_check_load_store_mask
21165 and vect_check_store_rhs. Use the dt returned by the latter instead
21166 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
21167 instead of calls to vect_is_simple_use. Pass the scalar rather
21168 than the vector operand to vect_is_simple_use when handling
21169 second and subsequent copies of an rhs value.
21170 (vectorizable_load): Update calls to vect_check_load_store_mask
21171 and vect_build_gather_load_calls. Use the cached mask_dt and
21172 gs_info.offset_dt instead of calls to vect_is_simple_use.
21174 2018-01-20 Jakub Jelinek <jakub@redhat.com>
21176 PR middle-end/83945
21177 * tree-emutls.c: Include gimplify.h.
21178 (lower_emutls_2): New function.
21179 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
21180 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
21181 it before further processing.
21184 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
21185 UINTVAL (trueop1) instead of INTVAL (op1).
21187 2018-01-19 Jakub Jelinek <jakub@redhat.com>
21191 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
21192 INCOMING_FRAME_SP_OFFSET if not defined.
21193 (scan_trace): Add ENTRY argument. If true and
21194 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
21195 emit a note to adjust the CFA offset.
21196 (create_cfi_notes): Adjust scan_trace callers.
21197 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
21198 INCOMING_FRAME_SP_OFFSET in the CIE.
21199 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
21200 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
21202 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
21203 * doc/tm.texi: Regenerated.
21205 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
21207 PR rtl-optimization/83147
21208 * lra-constraints.c (remove_inheritance_pseudos): Use
21209 lra_substitute_pseudo_within_insn.
21211 2018-01-19 Tom de Vries <tom@codesourcery.com>
21212 Cesar Philippidis <cesar@codesourcery.com>
21215 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
21217 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
21220 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
21221 spaces for function labels.
21223 2018-01-19 Martin Liska <mliska@suse.cz>
21225 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
21226 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
21227 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
21228 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
21229 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
21230 (PRED_CONST_RETURN): Change from 69 to 65.
21231 (PRED_NULL_RETURN): Change from 91 to 71.
21232 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
21233 (PRED_LOOP_GUARD): Change from 66 to 73.
21235 2018-01-19 Martin Liska <mliska@suse.cz>
21237 * predict.c (predict_insn_def): Add new assert.
21238 (struct branch_predictor): Change type to signed integer.
21239 (test_prediction_value_range): Amend test to cover
21240 PROB_UNINITIALIZED.
21241 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
21242 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
21243 (PRED_LOOP_ITERATIONS_MAX): Likewise.
21244 (PRED_LOOP_IV_COMPARE): Likewise.
21245 * predict.h (PROB_UNINITIALIZED): Define new constant.
21247 2018-01-19 Martin Liska <mliska@suse.cz>
21249 * predict.c (dump_prediction): Add new format for
21250 analyze_brprob.py script which is enabled with -details
21252 * profile-count.h (precise_p): New function.
21254 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
21256 PR tree-optimization/83922
21257 * tree-vect-loop.c (vect_verify_full_masking): Return false if
21258 there are no statements that need masking.
21259 (vect_active_double_reduction_p): New function.
21260 (vect_analyze_loop_operations): Use it when handling phis that
21261 are not in the loop header.
21263 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
21265 PR tree-optimization/83914
21266 * tree-vect-loop.c (vectorizable_induction): Don't convert
21267 init_expr or apply the peeling adjustment for inductions
21268 that are nested within the vectorized loop.
21270 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21272 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
21275 2018-01-18 Jakub Jelinek <jakub@redhat.com>
21279 * function.h (gimplify_parameters): Add gimple_seq * argument.
21280 * function.c: Include gimple.h and options.h.
21281 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
21282 for the added local temporaries if needed.
21283 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
21284 if there are any parameter cleanups, wrap whole body into a
21285 try/finally with the cleanups.
21287 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
21290 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
21291 Use GET_MODE_CLASS for scalar floating point.
21293 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
21297 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
21298 Fix call of call_cgraph_insertion_hooks.
21300 2018-01-18 Martin Sebor <msebor@redhat.com>
21302 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
21304 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
21307 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
21310 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
21313 * common.opt: (-ffile-prefix-map): New option.
21314 * opts.c (common_handle_option): Defer it.
21315 * opts-global.c (handle_common_deferred_options): Handle it.
21316 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
21317 * file-prefix-map.h: New file.
21318 (remap_debug_filename, add_debug_prefix_map): ...here.
21319 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
21320 * final.c (debug_prefix_map, add_debug_prefix_map
21321 remap_debug_filename): Move to...
21322 * file-prefix-map.c: New file.
21323 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
21324 generalize, get rid of alloca(), use strrchr() instead of strchr().
21325 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
21326 Implement in terms of add_prefix_map().
21327 (remap_macro_filename, remap_debug_filename): Implement in term of
21329 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
21330 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
21331 * dbxout.c: Include file-prefix-map.h.
21332 * varasm.c: Likewise.
21333 * vmsdbgout.c: Likewise.
21334 * xcoffout.c: Likewise.
21335 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
21336 * doc/cppopts.texi (-fmacro-prefix-map): Document.
21337 * doc/invoke.texi (-ffile-prefix-map): Document.
21338 (-fdebug-prefix-map): Update description.
21340 2018-01-18 Martin Liska <mliska@suse.cz>
21342 * config/i386/i386.c (indirect_thunk_name): Document that also
21344 (output_indirect_thunk): Document why both instructions
21345 (pause and lfence) are generated.
21347 2018-01-18 Richard Biener <rguenther@suse.de>
21349 PR tree-optimization/83887
21350 * graphite-scop-detection.c
21351 (scop_detection::get_nearest_dom_with_single_entry): Remove.
21352 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
21353 (scop_detection::merge_sese): Re-implement with a flood-fill
21354 algorithm that properly finds a SESE region if it exists.
21356 2018-01-18 Jakub Jelinek <jakub@redhat.com>
21359 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
21360 pointer_diff optimizations use view_convert instead of convert.
21362 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21364 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
21365 Generate different code for -mno-speculate-indirect-jumps.
21366 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
21367 (*call_indirect_aix<mode>): Disable for
21368 -mno-speculate-indirect-jumps.
21369 (*call_indirect_aix<mode>_nospec): New define_insn.
21370 (*call_value_indirect_aix<mode>): Disable for
21371 -mno-speculate-indirect-jumps.
21372 (*call_value_indirect_aix<mode>_nospec): New define_insn.
21373 (*sibcall_nonlocal_sysv<mode>): Generate different code for
21374 -mno-speculate-indirect-jumps.
21375 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
21377 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
21379 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
21380 long double type, set the flags for noting the default long double
21381 type, even if we don't pass or return a long double type.
21383 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
21386 * ipa-inline.c (flatten_function): Do not overwrite final inlining
21389 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
21391 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
21392 support for merge[hl].
21393 (fold_mergehl_helper): New helper function.
21394 (tree-vector-builder.h): New #include for tree_vector_builder usage.
21395 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
21396 (altivec_vmrglw_direct): Add xxmrglw insn.
21398 2018-01-17 Andrew Waterman <andrew@sifive.com>
21400 * config/riscv/riscv.c (riscv_conditional_register_usage): If
21401 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
21403 2018-01-17 David Malcolm <dmalcolm@redhat.com>
21406 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
21407 call the lto_location_cache before reading the
21408 DECL_SOURCE_LOCATION of the types.
21410 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
21411 Richard Sandiford <richard.sandiford@linaro.org>
21413 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
21414 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
21415 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
21416 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
21417 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
21419 * config/aarch64/constraints.md (aarch64_movti_operand):
21421 * config/aarch64/predicates.md (Uti): Add new constraint.
21423 2018-01-17 Carl Love <cel@us.ibm.com>
21425 * config/rs6000/vsx.md (define_expand xl_len_r,
21426 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
21427 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
21429 (define_expand, define_insn): Move the shift left from the
21430 define_insn to the define_expand for lxvl and stxvl instructions.
21431 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
21432 and XL_LEN_R definitions to PURE.
21434 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
21436 * config/i386/i386.c (indirect_thunk_name): Declare regno
21437 as unsigned int. Compare regno with INVALID_REGNUM.
21438 (output_indirect_thunk): Ditto.
21439 (output_indirect_thunk_function): Ditto.
21440 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
21441 in the call to output_indirect_thunk_function.
21443 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
21445 PR middle-end/83884
21446 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
21447 rather than the size of inner_type to determine the stack slot size
21448 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
21450 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
21453 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
21456 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
21458 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
21459 endian Linux systems to optionally enable multilibs for selecting
21460 the long double type if the user configured an explicit type.
21461 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
21462 have no long double multilibs if not defined.
21463 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
21464 warn if the user used -mabi={ieee,ibm}longdouble and we built
21465 multilibs for long double.
21466 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
21467 appropriate multilib option.
21468 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
21470 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
21471 for building long double multilibs.
21472 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
21474 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
21476 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
21479 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
21481 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
21484 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
21487 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
21490 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
21492 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
21493 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
21495 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
21497 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
21498 different rtl trees depending on TARGET_64BIT.
21499 (rs6000_gen_lvx): Likewise.
21501 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
21503 * config/visium/visium.md (nop): Tweak comment.
21504 (hazard_nop): Likewise.
21506 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21508 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
21509 -mspeculate-indirect-jumps.
21510 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
21511 for -mno-speculate-indirect-jumps.
21512 (*call_indirect_elfv2<mode>_nospec): New define_insn.
21513 (*call_value_indirect_elfv2<mode>): Disable for
21514 -mno-speculate-indirect-jumps.
21515 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
21516 (indirect_jump): Emit different RTL for
21517 -mno-speculate-indirect-jumps.
21518 (*indirect_jump<mode>): Disable for
21519 -mno-speculate-indirect-jumps.
21520 (*indirect_jump<mode>_nospec): New define_insn.
21521 (tablejump): Emit different RTL for
21522 -mno-speculate-indirect-jumps.
21523 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
21524 (tablejumpsi_nospec): New define_expand.
21525 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
21526 (tablejumpdi_nospec): New define_expand.
21527 (*tablejump<mode>_internal1): Disable for
21528 -mno-speculate-indirect-jumps.
21529 (*tablejump<mode>_internal1_nospec): New define_insn.
21530 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
21533 2018-01-16 Artyom Skrobov tyomitch@gmail.com
21535 * caller-save.c (insert_save): Drop unnecessary parameter. All
21538 2018-01-16 Jakub Jelinek <jakub@redhat.com>
21539 Richard Biener <rguenth@suse.de>
21542 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
21543 return early, inline manually is_gimple_sizepos. Make sure if we
21544 call gimplify_expr we don't end up with a gimple constant.
21545 * tree.c (variably_modified_type_p): Don't return true for
21546 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
21547 * gimplify.h (is_gimple_sizepos): Remove.
21549 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21551 PR tree-optimization/83857
21552 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
21553 vectorizable_live_operation for pure SLP statements.
21554 (vectorizable_live_operation): Handle PHIs.
21556 2018-01-16 Richard Biener <rguenther@suse.de>
21558 PR tree-optimization/83867
21559 * tree-vect-stmts.c (vect_transform_stmt): Precompute
21560 nested_in_vect_loop_p since the scalar stmt may get invalidated.
21562 2018-01-16 Jakub Jelinek <jakub@redhat.com>
21565 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
21566 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
21567 If off is not INTEGER_CST, issue a may not be aligned warning
21568 rather than isn't aligned. Use isn%'t rather than isn't.
21569 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
21571 <case MULT_EXPR>: Improve the case when bottom and one of the
21572 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
21573 operand, in that case check if the other operand is multiple of
21574 bottom divided by the INTEGER_CST operand.
21576 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21579 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
21580 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
21581 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
21582 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
21583 * config/pa/pa.c (pa_function_arg_advance): Likewise.
21584 (pa_function_arg, pa_arg_partial_bytes): Likewise.
21585 (pa_function_arg_size): New function.
21587 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21589 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
21590 in a separate statement.
21592 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21594 PR tree-optimization/83847
21595 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
21596 group gathers and scatters.
21598 2018-01-16 Jakub Jelinek <jakub@redhat.com>
21600 PR rtl-optimization/86620
21601 * params.def (max-sched-ready-insns): Bump minimum value to 1.
21603 PR rtl-optimization/83213
21604 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
21605 to last if both are JUMP_INSNs.
21607 PR tree-optimization/83843
21608 * gimple-ssa-store-merging.c
21609 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
21610 store_immediate_info for bswap/nop orig_stores.
21612 2018-01-15 Andrew Waterman <andrew@sifive.com>
21614 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
21616 <UDIV>: Increase cost if !TARGET_DIV.
21618 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
21620 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
21621 (define_attr "cr_logical_3op"): New.
21622 (cceq_ior_compare): Adjust.
21623 (cceq_ior_compare_complement): Adjust.
21624 (*cceq_rev_compare): Adjust.
21625 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
21626 (is_cracked_insn): Adjust.
21627 (insn_must_be_first_in_group): Adjust.
21628 * config/rs6000/40x.md: Adjust.
21629 * config/rs6000/440.md: Adjust.
21630 * config/rs6000/476.md: Adjust.
21631 * config/rs6000/601.md: Adjust.
21632 * config/rs6000/603.md: Adjust.
21633 * config/rs6000/6xx.md: Adjust.
21634 * config/rs6000/7450.md: Adjust.
21635 * config/rs6000/7xx.md: Adjust.
21636 * config/rs6000/8540.md: Adjust.
21637 * config/rs6000/cell.md: Adjust.
21638 * config/rs6000/e300c2c3.md: Adjust.
21639 * config/rs6000/e500mc.md: Adjust.
21640 * config/rs6000/e500mc64.md: Adjust.
21641 * config/rs6000/e5500.md: Adjust.
21642 * config/rs6000/e6500.md: Adjust.
21643 * config/rs6000/mpc.md: Adjust.
21644 * config/rs6000/power4.md: Adjust.
21645 * config/rs6000/power5.md: Adjust.
21646 * config/rs6000/power6.md: Adjust.
21647 * config/rs6000/power7.md: Adjust.
21648 * config/rs6000/power8.md: Adjust.
21649 * config/rs6000/power9.md: Adjust.
21650 * config/rs6000/rs64.md: Adjust.
21651 * config/rs6000/titan.md: Adjust.
21653 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21655 * config/i386/predicates.md (indirect_branch_operand): Rewrite
21656 ix86_indirect_branch_register logic.
21658 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21660 * config/i386/constraints.md (Bs): Update
21661 ix86_indirect_branch_register check. Don't check
21662 ix86_indirect_branch_register with GOT_memory_operand.
21664 * config/i386/predicates.md (GOT_memory_operand): Don't check
21665 ix86_indirect_branch_register here.
21666 (GOT32_symbol_operand): Likewise.
21668 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21670 * config/i386/predicates.md (constant_call_address_operand):
21671 Rewrite ix86_indirect_branch_register logic.
21672 (sibcall_insn_operand): Likewise.
21674 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21676 * config/i386/constraints.md (Bs): Replace
21677 ix86_indirect_branch_thunk_register with
21678 ix86_indirect_branch_register.
21680 * config/i386/i386.md (indirect_jump): Likewise.
21681 (tablejump): Likewise.
21682 (*sibcall_memory): Likewise.
21683 (*sibcall_value_memory): Likewise.
21684 Peepholes of indirect call and jump via memory: Likewise.
21685 * config/i386/i386.opt: Likewise.
21686 * config/i386/predicates.md (indirect_branch_operand): Likewise.
21687 (GOT_memory_operand): Likewise.
21688 (call_insn_operand): Likewise.
21689 (sibcall_insn_operand): Likewise.
21690 (GOT32_symbol_operand): Likewise.
21692 2018-01-15 Jakub Jelinek <jakub@redhat.com>
21694 PR middle-end/83837
21695 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
21696 type rather than type addr's type points to.
21697 (expand_omp_atomic_mutex): Likewise.
21698 (expand_omp_atomic): Likewise.
21700 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21703 * config/i386/i386.c (output_indirect_thunk_function): Use
21704 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
21705 for __x86_return_thunk.
21707 2018-01-15 Richard Biener <rguenther@suse.de>
21709 PR middle-end/83850
21710 * expmed.c (extract_bit_field_1): Fix typo.
21712 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21715 * config/arm/iterators.md (VF): New mode iterator.
21716 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
21717 Remove integer-related logic from pattern.
21718 (neon_vabd<mode>_3): Likewise.
21720 2018-01-15 Jakub Jelinek <jakub@redhat.com>
21722 PR middle-end/82694
21723 * common.opt (fstrict-overflow): No longer an alias.
21724 (fwrapv-pointer): New option.
21725 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
21726 also for pointer types based on flag_wrapv_pointer.
21727 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
21728 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
21729 opts->x_flag_wrapv got set.
21730 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
21731 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
21732 POINTER_TYPE_OVERFLOW_UNDEFINED.
21733 * match.pd: Likewise in address comparison pattern.
21734 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
21736 2018-01-15 Richard Biener <rguenther@suse.de>
21739 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
21740 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
21741 Reset type names to their identifier if their TYPE_DECL doesn't
21742 have linkage (and thus is used for ODR and devirt).
21743 (save_debug_info_for_decl): Remove.
21744 (save_debug_info_for_type): Likewise.
21745 (add_tree_to_fld_list): Adjust.
21746 * tree-pretty-print.c (dump_generic_node): Make dumping of
21747 type names more robust.
21749 2018-01-15 Richard Biener <rguenther@suse.de>
21751 * BASE-VER: Bump to 8.0.1.
21753 2018-01-14 Martin Sebor <msebor@redhat.com>
21756 * builtins.c (check_access): Avoid warning when the no-warning bit
21759 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
21761 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
21762 * ira-color (allocno_hard_regs_compare): Likewise.
21764 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
21767 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
21768 Use .pushsection/.popsection.
21770 2018-01-14 Martin Sebor <msebor@redhat.com>
21773 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
21775 2018-01-14 Jakub Jelinek <jakub@redhat.com>
21777 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
21778 entry from extra_headers.
21779 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
21780 extra_headers, make the list bitwise identical to the i?86-*-* one.
21782 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21784 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
21785 -mcmodel=large with -mindirect-branch=thunk,
21786 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
21787 -mfunction-return=thunk-extern.
21788 * doc/invoke.texi: Document -mcmodel=large is incompatible with
21789 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
21790 -mfunction-return=thunk and -mfunction-return=thunk-extern.
21792 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21794 * config/i386/i386.c (print_reg): Print the name of the full
21795 integer register without '%'.
21796 (ix86_print_operand): Handle 'V'.
21797 * doc/extend.texi: Document 'V' modifier.
21799 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21801 * config/i386/constraints.md (Bs): Disallow memory operand for
21802 -mindirect-branch-register.
21804 * config/i386/predicates.md (indirect_branch_operand): Likewise.
21805 (GOT_memory_operand): Likewise.
21806 (call_insn_operand): Likewise.
21807 (sibcall_insn_operand): Likewise.
21808 (GOT32_symbol_operand): Likewise.
21809 * config/i386/i386.md (indirect_jump): Call convert_memory_address
21810 for -mindirect-branch-register.
21811 (tablejump): Likewise.
21812 (*sibcall_memory): Likewise.
21813 (*sibcall_value_memory): Likewise.
21814 Disallow peepholes of indirect call and jump via memory for
21815 -mindirect-branch-register.
21816 (*call_pop): Replace m with Bw.
21817 (*call_value_pop): Likewise.
21818 (*sibcall_pop_memory): Replace m with Bs.
21819 * config/i386/i386.opt (mindirect-branch-register): New option.
21820 * doc/invoke.texi: Document -mindirect-branch-register option.
21822 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21824 * config/i386/i386-protos.h (ix86_output_function_return): New.
21825 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
21826 set function_return_type.
21827 (indirect_thunk_name): Add ret_p to indicate thunk for function
21829 (output_indirect_thunk_function): Pass false to
21830 indirect_thunk_name.
21831 (ix86_output_indirect_branch_via_reg): Likewise.
21832 (ix86_output_indirect_branch_via_push): Likewise.
21833 (output_indirect_thunk_function): Create alias for function
21834 return thunk if regno < 0.
21835 (ix86_output_function_return): New function.
21836 (ix86_handle_fndecl_attribute): Handle function_return.
21837 (ix86_attribute_table): Add function_return.
21838 * config/i386/i386.h (machine_function): Add
21839 function_return_type.
21840 * config/i386/i386.md (simple_return_internal): Use
21841 ix86_output_function_return.
21842 (simple_return_internal_long): Likewise.
21843 * config/i386/i386.opt (mfunction-return=): New option.
21844 (indirect_branch): Mention -mfunction-return=.
21845 * doc/extend.texi: Document function_return function attribute.
21846 * doc/invoke.texi: Document -mfunction-return= option.
21848 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21850 * config/i386/i386-opts.h (indirect_branch): New.
21851 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
21852 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
21853 with local indirect jump when converting indirect call and jump.
21854 (ix86_set_indirect_branch_type): New.
21855 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
21856 (indirectlabelno): New.
21857 (indirect_thunk_needed): Likewise.
21858 (indirect_thunk_bnd_needed): Likewise.
21859 (indirect_thunks_used): Likewise.
21860 (indirect_thunks_bnd_used): Likewise.
21861 (INDIRECT_LABEL): Likewise.
21862 (indirect_thunk_name): Likewise.
21863 (output_indirect_thunk): Likewise.
21864 (output_indirect_thunk_function): Likewise.
21865 (ix86_output_indirect_branch_via_reg): Likewise.
21866 (ix86_output_indirect_branch_via_push): Likewise.
21867 (ix86_output_indirect_branch): Likewise.
21868 (ix86_output_indirect_jmp): Likewise.
21869 (ix86_code_end): Call output_indirect_thunk_function if needed.
21870 (ix86_output_call_insn): Call ix86_output_indirect_branch if
21872 (ix86_handle_fndecl_attribute): Handle indirect_branch.
21873 (ix86_attribute_table): Add indirect_branch.
21874 * config/i386/i386.h (machine_function): Add indirect_branch_type
21875 and has_local_indirect_jump.
21876 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
21878 (tablejump): Likewise.
21879 (*indirect_jump): Use ix86_output_indirect_jmp.
21880 (*tablejump_1): Likewise.
21881 (simple_return_indirect_internal): Likewise.
21882 * config/i386/i386.opt (mindirect-branch=): New option.
21883 (indirect_branch): New.
21886 (thunk-inline): Likewise.
21887 (thunk-extern): Likewise.
21888 * doc/extend.texi: Document indirect_branch function attribute.
21889 * doc/invoke.texi: Document -mindirect-branch= option.
21891 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
21894 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
21896 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
21898 * ipa-inline.c (want_inline_small_function_p): Return false if
21899 inlining has already failed with CIF_FINAL_ERROR.
21900 (update_caller_keys): Call want_inline_small_function_p before
21902 (update_callee_keys): Likewise.
21904 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
21906 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
21908 (rs6000_quadword_masked_address_p): Likewise.
21909 (quad_aligned_load_p): Likewise.
21910 (quad_aligned_store_p): Likewise.
21911 (const_load_sequence_p): Add comment to describe the outer-most loop.
21912 (mimic_memory_attributes_and_flags): New function.
21913 (rs6000_gen_stvx): Likewise.
21914 (replace_swapped_aligned_store): Likewise.
21915 (rs6000_gen_lvx): Likewise.
21916 (replace_swapped_aligned_load): Likewise.
21917 (replace_swapped_load_constant): Capitalize argument name in
21918 comment describing this function.
21919 (rs6000_analyze_swaps): Add a third pass to search for vector loads
21920 and stores that access quad-word aligned addresses and replace
21921 with stvx or lvx instructions when appropriate.
21922 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
21923 New function prototype.
21924 (rs6000_quadword_masked_address_p): Likewise.
21925 (rs6000_gen_lvx): Likewise.
21926 (rs6000_gen_stvx): Likewise.
21927 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
21928 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
21929 when memory address is aligned.
21930 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
21931 this split to select lvx instruction when memory address is aligned.
21932 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
21933 instruction when memory address is aligned.
21934 (*vsx_le_perm_load_v16qi): Likewise.
21935 (four unnamed splitters): Modify to select the stvx instruction
21936 when memory is aligned.
21938 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
21940 * predict.c (determine_unlikely_bbs): Handle correctly BBs
21941 which appears in the queue multiple times.
21943 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21944 Alan Hayward <alan.hayward@arm.com>
21945 David Sherwood <david.sherwood@arm.com>
21947 * tree-vectorizer.h (vec_lower_bound): New structure.
21948 (_loop_vec_info): Add check_nonzero and lower_bounds.
21949 (LOOP_VINFO_CHECK_NONZERO): New macro.
21950 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
21951 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
21952 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
21953 fields. Make seg_len the distance travelled, not including the
21955 (dr_direction_indicator): Declare.
21956 (dr_zero_step_indicator): Likewise.
21957 (dr_known_forward_stride_p): Likewise.
21958 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
21960 (runtime_alias_check_p): Allow runtime alias checks with
21962 (operator ==): Compare access_size and align.
21963 (prune_runtime_alias_test_list): Rework for new distinction between
21964 the access_size and seg_len.
21965 (create_intersect_range_checks_index): Likewise. Cope with polynomial
21967 (get_segment_min_max): New function.
21968 (create_intersect_range_checks): Use it.
21969 (dr_step_indicator): New function.
21970 (dr_direction_indicator): Likewise.
21971 (dr_zero_step_indicator): Likewise.
21972 (dr_known_forward_stride_p): Likewise.
21973 * tree-loop-distribution.c (data_ref_segment_size): Return
21974 DR_STEP * (niters - 1).
21975 (compute_alias_check_pairs): Update call to the dr_with_seg_len
21977 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
21978 (vect_preserves_scalar_order_p): New function, split out from...
21979 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
21980 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
21981 (vect_vfa_access_size): New function.
21982 (vect_vfa_align): Likewise.
21983 (vect_compile_time_alias): Take access_size_a and access_b arguments.
21984 (dump_lower_bound): New function.
21985 (vect_check_lower_bound): Likewise.
21986 (vect_small_gap_p): Likewise.
21987 (vectorizable_with_step_bound_p): Likewise.
21988 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
21989 depencies if the vectorization factor is 1. Convert the checks
21990 for nonzero steps into checks on the bounds of DR_STEP. Try using
21991 a bunds check for variable steps if the minimum required step is
21992 relatively small. Update calls to the dr_with_seg_len
21993 constructor and to vect_compile_time_alias.
21994 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
21996 (vect_loop_versioning): Call it.
21997 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
21999 (vect_estimate_min_profitable_iters): Account for any bounds checks.
22001 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22002 Alan Hayward <alan.hayward@arm.com>
22003 David Sherwood <david.sherwood@arm.com>
22005 * doc/sourcebuild.texi (vect_scatter_store): Document.
22006 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
22008 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
22010 * genopinit.c (main): Add supports_vec_scatter_store and
22011 supports_vec_scatter_store_cached to target_optabs.
22012 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
22013 IFN_MASK_SCATTER_STORE.
22014 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
22016 * internal-fn.h (internal_store_fn_p): Declare.
22017 (internal_fn_stored_value_index): Likewise.
22018 * internal-fn.c (scatter_store_direct): New macro.
22019 (expand_scatter_store_optab_fn): New function.
22020 (direct_scatter_store_optab_supported_p): New macro.
22021 (internal_store_fn_p): New function.
22022 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
22023 IFN_MASK_SCATTER_STORE.
22024 (internal_fn_mask_index): Likewise.
22025 (internal_fn_stored_value_index): New function.
22026 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
22027 for scatter stores.
22028 * optabs-query.h (supports_vec_scatter_store_p): Declare.
22029 * optabs-query.c (supports_vec_scatter_store_p): New function.
22030 * tree-vectorizer.h (vect_get_store_rhs): Declare.
22031 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
22032 true for scatter stores.
22033 (vect_gather_scatter_fn_p): Handle scatter stores too.
22034 (vect_check_gather_scatter): Consider using scatter stores if
22035 supports_vec_scatter_store_p.
22036 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
22037 scatter stores too.
22038 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
22039 internal_fn_stored_value_index.
22040 (check_load_store_masking): Handle scatter stores too.
22041 (vect_get_store_rhs): Make public.
22042 (vectorizable_call): Use internal_store_fn_p.
22043 (vectorizable_store): Handle scatter store internal functions.
22044 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
22045 when deciding whether the end of the group has been reached.
22046 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
22047 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
22048 (mask_scatter_store<mode>): New insns.
22050 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22051 Alan Hayward <alan.hayward@arm.com>
22052 David Sherwood <david.sherwood@arm.com>
22054 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
22055 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
22056 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
22058 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
22059 Use vect_truncate_gather_scatter_offset if we can't treat the
22060 operation as a normal gather load or scatter store.
22061 (get_group_load_store_type): Take the gather_scatter_info
22062 as argument. Try using a gather load or scatter store for
22063 single-element groups.
22064 (get_load_store_type): Update calls to get_group_load_store_type
22065 and vect_use_strided_gather_scatters_p.
22067 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22068 Alan Hayward <alan.hayward@arm.com>
22069 David Sherwood <david.sherwood@arm.com>
22071 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
22072 optional tree argument.
22073 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
22075 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
22076 but continue to use the current value as a fallback.
22077 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
22078 to compare the updates.
22079 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
22080 (get_load_store_type): Use it when handling a strided access.
22081 (vect_get_strided_load_store_ops): New function.
22082 (vect_get_data_ptr_increment): Likewise.
22083 (vectorizable_load): Handle strided gather loads. Always pass
22084 a step to vect_create_data_ref_ptr and bump_vector_ptr.
22086 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22087 Alan Hayward <alan.hayward@arm.com>
22088 David Sherwood <david.sherwood@arm.com>
22090 * doc/md.texi (gather_load@var{m}): Document.
22091 (mask_gather_load@var{m}): Likewise.
22092 * genopinit.c (main): Add supports_vec_gather_load and
22093 supports_vec_gather_load_cached to target_optabs.
22094 * optabs-tree.c (init_tree_optimization_optabs): Use
22095 ggc_cleared_alloc to allocate target_optabs.
22096 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
22097 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
22099 * internal-fn.h (internal_load_fn_p): Declare.
22100 (internal_gather_scatter_fn_p): Likewise.
22101 (internal_fn_mask_index): Likewise.
22102 (internal_gather_scatter_fn_supported_p): Likewise.
22103 * internal-fn.c (gather_load_direct): New macro.
22104 (expand_gather_load_optab_fn): New function.
22105 (direct_gather_load_optab_supported_p): New macro.
22106 (direct_internal_fn_optab): New function.
22107 (internal_load_fn_p): Likewise.
22108 (internal_gather_scatter_fn_p): Likewise.
22109 (internal_fn_mask_index): Likewise.
22110 (internal_gather_scatter_fn_supported_p): Likewise.
22111 * optabs-query.c (supports_at_least_one_mode_p): New function.
22112 (supports_vec_gather_load_p): Likewise.
22113 * optabs-query.h (supports_vec_gather_load_p): Declare.
22114 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
22115 and memory_type field.
22116 (NUM_PATTERNS): Bump to 15.
22117 * tree-vect-data-refs.c: Include internal-fn.h.
22118 (vect_gather_scatter_fn_p): New function.
22119 (vect_describe_gather_scatter_call): Likewise.
22120 (vect_check_gather_scatter): Try using internal functions for
22121 gather loads. Recognize existing calls to a gather load function.
22122 (vect_analyze_data_refs): Consider using gather loads if
22123 supports_vec_gather_load_p.
22124 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
22125 (vect_get_gather_scatter_offset_type): Likewise.
22126 (vect_convert_mask_for_vectype): Likewise.
22127 (vect_add_conversion_to_patterm): Likewise.
22128 (vect_try_gather_scatter_pattern): Likewise.
22129 (vect_recog_gather_scatter_pattern): New pattern recognizer.
22130 (vect_vect_recog_func_ptrs): Add it.
22131 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
22132 internal_fn_mask_index and internal_gather_scatter_fn_p.
22133 (check_load_store_masking): Take the gather_scatter_info as an
22134 argument and handle gather loads.
22135 (vect_get_gather_scatter_ops): New function.
22136 (vectorizable_call): Check internal_load_fn_p.
22137 (vectorizable_load): Likewise. Handle gather load internal
22139 (vectorizable_store): Update call to check_load_store_masking.
22140 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
22141 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
22142 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
22143 (aarch64_gather_scale_operand_d): New predicates.
22144 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
22145 (mask_gather_load<mode>): New insns.
22147 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22148 Alan Hayward <alan.hayward@arm.com>
22149 David Sherwood <david.sherwood@arm.com>
22151 * optabs.def (fold_left_plus_optab): New optab.
22152 * doc/md.texi (fold_left_plus_@var{m}): Document.
22153 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
22154 * internal-fn.c (fold_left_direct): Define.
22155 (expand_fold_left_optab_fn): Likewise.
22156 (direct_fold_left_optab_supported_p): Likewise.
22157 * fold-const-call.c (fold_const_fold_left): New function.
22158 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
22159 * tree-parloops.c (valid_reduction_p): New function.
22160 (gather_scalar_reductions): Use it.
22161 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
22162 (vect_finish_replace_stmt): Declare.
22163 * tree-vect-loop.c (fold_left_reduction_fn): New function.
22164 (needs_fold_left_reduction_p): New function, split out from...
22165 (vect_is_simple_reduction): ...here. Accept reductions that
22166 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
22167 (vect_force_simple_reduction): Also store the reduction type in
22168 the assignment's STMT_VINFO_REDUC_TYPE.
22169 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
22170 (merge_with_identity): New function.
22171 (vect_expand_fold_left): Likewise.
22172 (vectorize_fold_left_reduction): Likewise.
22173 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
22174 scalar phi in place for it. Check for target support and reject
22175 cases that would reassociate the operation. Defer the transform
22176 phase to vectorize_fold_left_reduction.
22177 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
22178 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
22179 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
22181 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22183 * tree-if-conv.c (predicate_mem_writes): Remove redundant
22184 call to ifc_temp_var.
22186 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22187 Alan Hayward <alan.hayward@arm.com>
22188 David Sherwood <david.sherwood@arm.com>
22190 * target.def (legitimize_address_displacement): Take the original
22191 offset as a poly_int.
22192 * targhooks.h (default_legitimize_address_displacement): Update
22194 * targhooks.c (default_legitimize_address_displacement): Likewise.
22195 * doc/tm.texi: Regenerate.
22196 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
22197 as an argument, moving assert of ad->disp == ad->disp_term to...
22198 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
22199 Try calling targetm.legitimize_address_displacement before expanding
22200 the address rather than afterwards, and adjust for the new interface.
22201 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
22202 Match the new hook interface. Handle SVE addresses.
22203 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
22204 new hook interface.
22206 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22208 * Makefile.in (OBJS): Add early-remat.o.
22209 * target.def (select_early_remat_modes): New hook.
22210 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
22211 * doc/tm.texi: Regenerate.
22212 * targhooks.h (default_select_early_remat_modes): Declare.
22213 * targhooks.c (default_select_early_remat_modes): New function.
22214 * timevar.def (TV_EARLY_REMAT): New timevar.
22215 * passes.def (pass_early_remat): New pass.
22216 * tree-pass.h (make_pass_early_remat): Declare.
22217 * early-remat.c: New file.
22218 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
22220 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
22222 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22223 Alan Hayward <alan.hayward@arm.com>
22224 David Sherwood <david.sherwood@arm.com>
22226 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
22227 vfm1 with a bound_epilog parameter.
22228 (vect_do_peeling): Update calls accordingly, and move the prologue
22229 call earlier in the function. Treat the base bound_epilog as 0 for
22230 fully-masked loops and retain vf - 1 for other loops. Add 1 to
22231 this base when peeling for gaps.
22232 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
22233 with fully-masked loops.
22234 (vect_estimate_min_profitable_iters): Handle the single peeled
22235 iteration in that case.
22237 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22238 Alan Hayward <alan.hayward@arm.com>
22239 David Sherwood <david.sherwood@arm.com>
22241 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
22242 single-element interleaving even if the size is not a power of 2.
22243 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
22244 accesses for single-element interleaving if the group size is
22247 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22248 Alan Hayward <alan.hayward@arm.com>
22249 David Sherwood <david.sherwood@arm.com>
22251 * doc/md.texi (fold_extract_last_@var{m}): Document.
22252 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
22253 * optabs.def (fold_extract_last_optab): New optab.
22254 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
22255 * internal-fn.c (fold_extract_direct): New macro.
22256 (expand_fold_extract_optab_fn): Likewise.
22257 (direct_fold_extract_optab_supported_p): Likewise.
22258 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
22259 * tree-vect-loop.c (vect_model_reduction_cost): Handle
22260 EXTRACT_LAST_REDUCTION.
22261 (get_initial_def_for_reduction): Do not create an initial vector
22262 for EXTRACT_LAST_REDUCTION reductions.
22263 (vectorizable_reduction): Leave the scalar phi in place for
22264 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
22265 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
22266 epilogue code for EXTRACT_LAST_REDUCTION and defer the
22267 transform phase to vectorizable_condition.
22268 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
22270 (vect_finish_stmt_generation): ...here.
22271 (vect_finish_replace_stmt): New function.
22272 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
22273 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
22275 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
22277 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22278 Alan Hayward <alan.hayward@arm.com>
22279 David Sherwood <david.sherwood@arm.com>
22281 * doc/md.texi (extract_last_@var{m}): Document.
22282 * optabs.def (extract_last_optab): New optab.
22283 * internal-fn.def (EXTRACT_LAST): New internal function.
22284 * internal-fn.c (cond_unary_direct): New macro.
22285 (expand_cond_unary_optab_fn): Likewise.
22286 (direct_cond_unary_optab_supported_p): Likewise.
22287 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
22288 loops using EXTRACT_LAST.
22289 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
22290 (extract_last_<mode>): ...this optab.
22291 (vec_extract<mode><Vel>): Update accordingly.
22293 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22294 Alan Hayward <alan.hayward@arm.com>
22295 David Sherwood <david.sherwood@arm.com>
22297 * target.def (empty_mask_is_expensive): New hook.
22298 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
22299 * doc/tm.texi: Regenerate.
22300 * targhooks.h (default_empty_mask_is_expensive): Declare.
22301 * targhooks.c (default_empty_mask_is_expensive): New function.
22302 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
22303 if the target says that empty masks are expensive.
22304 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
22306 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
22308 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22309 Alan Hayward <alan.hayward@arm.com>
22310 David Sherwood <david.sherwood@arm.com>
22312 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
22313 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
22314 (vect_use_loop_mask_for_alignment_p): New function.
22315 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
22316 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
22317 niters_skip argument. Make sure that the first niters_skip elements
22318 of the first iteration are inactive.
22319 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
22320 Update call to vect_set_loop_masks_directly.
22321 (get_misalign_in_elems): New function, split out from...
22322 (vect_gen_prolog_loop_niters): ...here.
22323 (vect_update_init_of_dr): Take a code argument that specifies whether
22324 the adjustment should be added or subtracted.
22325 (vect_update_init_of_drs): Likewise.
22326 (vect_prepare_for_masked_peels): New function.
22327 (vect_do_peeling): Skip prologue peeling if we're using a mask
22328 instead. Update call to vect_update_inits_of_drs.
22329 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
22331 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
22332 alignment. Do not include the number of peeled iterations in
22333 the minimum threshold in that case.
22334 (vectorizable_induction): Adjust the start value down by
22335 LOOP_VINFO_MASK_SKIP_NITERS iterations.
22336 (vect_transform_loop): Call vect_prepare_for_masked_peels.
22337 Take the number of skipped iterations into account when calculating
22339 * tree-vect-stmts.c (vect_gen_while_not): New function.
22341 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22342 Alan Hayward <alan.hayward@arm.com>
22343 David Sherwood <david.sherwood@arm.com>
22345 * doc/sourcebuild.texi (vect_fully_masked): Document.
22346 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
22347 default value to 0.
22348 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
22350 (vect_analyze_loop_2): ...here. Don't check the vectorization
22351 factor against the number of loop iterations if the loop is
22354 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22355 Alan Hayward <alan.hayward@arm.com>
22356 David Sherwood <david.sherwood@arm.com>
22358 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
22359 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
22360 (dump_groups): Update accordingly.
22361 (iv_use::mem_type): New member variable.
22362 (address_p): New function.
22363 (record_use): Add a mem_type argument and initialize the new
22365 (record_group_use): Add a mem_type argument. Use address_p.
22366 Remove obsolete null checks of base_object. Update call to record_use.
22367 (find_interesting_uses_op): Update call to record_group_use.
22368 (find_interesting_uses_cond): Likewise.
22369 (find_interesting_uses_address): Likewise.
22370 (get_mem_type_for_internal_fn): New function.
22371 (find_address_like_use): Likewise.
22372 (find_interesting_uses_stmt): Try find_address_like_use before
22373 calling find_interesting_uses_op.
22374 (addr_offset_valid_p): Use the iv mem_type field as the type
22375 of the addressed memory.
22376 (add_autoinc_candidates): Likewise.
22377 (get_address_cost): Likewise.
22378 (split_small_address_groups_p): Use address_p.
22379 (split_address_groups): Likewise.
22380 (add_iv_candidate_for_use): Likewise.
22381 (autoinc_possible_for_pair): Likewise.
22382 (rewrite_groups): Likewise.
22383 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
22384 (determine_group_iv_cost): Update after split of USE_ADDRESS.
22385 (get_alias_ptr_type_for_ptr_address): New function.
22386 (rewrite_use_address): Rewrite address uses in calls that were
22387 identified by find_address_like_use.
22389 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22390 Alan Hayward <alan.hayward@arm.com>
22391 David Sherwood <david.sherwood@arm.com>
22393 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
22395 * gimple-expr.h (is_gimple_addressable: Likewise.
22396 * gimple-expr.c (is_gimple_address): Likewise.
22397 * internal-fn.c (expand_call_mem_ref): New function.
22398 (expand_mask_load_optab_fn): Use it.
22399 (expand_mask_store_optab_fn): Likewise.
22401 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22402 Alan Hayward <alan.hayward@arm.com>
22403 David Sherwood <david.sherwood@arm.com>
22405 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
22406 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
22407 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
22408 (cond_umax@var{mode}): Document.
22409 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
22410 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
22411 (cond_umin_optab, cond_umax_optab): New optabs.
22412 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
22413 (COND_IOR, COND_XOR): New internal functions.
22414 * internal-fn.h (get_conditional_internal_fn): Declare.
22415 * internal-fn.c (cond_binary_direct): New macro.
22416 (expand_cond_binary_optab_fn): Likewise.
22417 (direct_cond_binary_optab_supported_p): Likewise.
22418 (get_conditional_internal_fn): New function.
22419 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
22420 Cope with reduction statements that are vectorized as calls rather
22422 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
22423 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
22424 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
22425 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
22426 (UNSPEC_COND_EOR): New unspecs.
22427 (optab): Add mappings for them.
22428 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
22429 (sve_int_op, sve_fp_op): New int attributes.
22431 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22432 Alan Hayward <alan.hayward@arm.com>
22433 David Sherwood <david.sherwood@arm.com>
22435 * optabs.def (while_ult_optab): New optab.
22436 * doc/md.texi (while_ult@var{m}@var{n}): Document.
22437 * internal-fn.def (WHILE_ULT): New internal function.
22438 * internal-fn.h (direct_internal_fn_supported_p): New override
22439 that takes two types as argument.
22440 * internal-fn.c (while_direct): New macro.
22441 (expand_while_optab_fn): New function.
22442 (convert_optab_supported_p): Likewise.
22443 (direct_while_optab_supported_p): New macro.
22444 * wide-int.h (wi::udiv_ceil): New function.
22445 * tree-vectorizer.h (rgroup_masks): New structure.
22446 (vec_loop_masks): New typedef.
22447 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
22448 and fully_masked_p.
22449 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
22450 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
22451 (vect_max_vf): New function.
22452 (slpeel_make_loop_iterate_ntimes): Delete.
22453 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
22454 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
22455 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
22456 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
22457 internal-fn.h, stor-layout.h and optabs-query.h.
22458 (vect_set_loop_mask): New function.
22459 (add_preheader_seq): Likewise.
22460 (add_header_seq): Likewise.
22461 (interleave_supported_p): Likewise.
22462 (vect_maybe_permute_loop_masks): Likewise.
22463 (vect_set_loop_masks_directly): Likewise.
22464 (vect_set_loop_condition_masked): Likewise.
22465 (vect_set_loop_condition_unmasked): New function, split out from
22466 slpeel_make_loop_iterate_ntimes.
22467 (slpeel_make_loop_iterate_ntimes): Rename to..
22468 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
22469 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
22470 (vect_do_peeling): Update call accordingly.
22471 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
22473 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
22474 mask_compare_type, can_fully_mask_p and fully_masked_p.
22475 (release_vec_loop_masks): New function.
22476 (_loop_vec_info): Use it to free the loop masks.
22477 (can_produce_all_loop_masks_p): New function.
22478 (vect_get_max_nscalars_per_iter): Likewise.
22479 (vect_verify_full_masking): Likewise.
22480 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
22481 retries, and free the mask rgroups before retrying. Check loop-wide
22482 reasons for disallowing fully-masked loops. Make the final decision
22483 about whether use a fully-masked loop or not.
22484 (vect_estimate_min_profitable_iters): Do not assume that peeling
22485 for the number of iterations will be needed for fully-masked loops.
22486 (vectorizable_reduction): Disable fully-masked loops.
22487 (vectorizable_live_operation): Likewise.
22488 (vect_halve_mask_nunits): New function.
22489 (vect_double_mask_nunits): Likewise.
22490 (vect_record_loop_mask): Likewise.
22491 (vect_get_loop_mask): Likewise.
22492 (vect_transform_loop): Handle the case in which the final loop
22493 iteration might handle a partial vector. Call vect_set_loop_condition
22494 instead of slpeel_make_loop_iterate_ntimes.
22495 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
22496 (check_load_store_masking): New function.
22497 (prepare_load_store_mask): Likewise.
22498 (vectorizable_store): Handle fully-masked loops.
22499 (vectorizable_load): Likewise.
22500 (supportable_widening_operation): Use vect_halve_mask_nunits for
22502 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
22503 (vect_gen_while): New function.
22504 * config/aarch64/aarch64.md (umax<mode>3): New expander.
22505 (aarch64_uqdec<mode>): New insn.
22507 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22508 Alan Hayward <alan.hayward@arm.com>
22509 David Sherwood <david.sherwood@arm.com>
22511 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
22512 (reduc_xor_scal_optab): New optabs.
22513 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
22514 (reduc_xor_scal_@var{m}): Document.
22515 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
22516 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
22517 internal functions.
22518 * fold-const-call.c (fold_const_call): Handle them.
22519 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
22520 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
22521 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
22522 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
22523 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
22524 (UNSPEC_XORV): New unspecs.
22525 (optab): Add entries for them.
22526 (BITWISEV): New int iterator.
22527 (bit_reduc_op): New int attributes.
22529 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22530 Alan Hayward <alan.hayward@arm.com>
22531 David Sherwood <david.sherwood@arm.com>
22533 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
22534 * internal-fn.def (VEC_SHL_INSERT): New internal function.
22535 * optabs.def (vec_shl_insert_optab): New optab.
22536 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
22537 (duplicate_and_interleave): Likewise.
22538 * tree-vect-loop.c: Include internal-fn.h.
22539 (neutral_op_for_slp_reduction): New function, split out from
22540 get_initial_defs_for_reduction.
22541 (get_initial_def_for_reduction): Handle option 2 for variable-length
22542 vectors by loading the neutral value into a vector and then shifting
22543 the initial value into element 0.
22544 (get_initial_defs_for_reduction): Replace the code argument with
22545 the neutral value calculated by neutral_op_for_slp_reduction.
22546 Use gimple_build_vector for constant-length vectors.
22547 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
22548 but the first group_size elements have a neutral value.
22549 Use duplicate_and_interleave otherwise.
22550 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
22551 Update call to get_initial_defs_for_reduction. Handle SLP
22552 reductions for variable-length vectors by creating one vector
22553 result for each scalar result, with the elements associated
22554 with other scalar results stubbed out with the neutral value.
22555 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
22556 Require IFN_VEC_SHL_INSERT for double reductions on
22557 variable-length vectors, or SLP reductions that have
22558 a neutral value. Require can_duplicate_and_interleave_p
22559 support for variable-length unchained SLP reductions if there
22560 is no neutral value, such as for MIN/MAX reductions. Also require
22561 the number of vector elements to be a multiple of the number of
22562 SLP statements when doing variable-length unchained SLP reductions.
22563 Update call to vect_create_epilog_for_reduction.
22564 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
22565 and remove initial values.
22566 (duplicate_and_interleave): Make public.
22567 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
22568 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
22570 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22571 Alan Hayward <alan.hayward@arm.com>
22572 David Sherwood <david.sherwood@arm.com>
22574 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
22575 (can_duplicate_and_interleave_p): New function.
22576 (vect_get_and_check_slp_defs): Take the vector of statements
22577 rather than just the current one. Remove excess parentheses.
22578 Restriction rejectinon of vect_constant_def and vect_external_def
22579 for variable-length vectors to boolean types, or types for which
22580 can_duplicate_and_interleave_p is false.
22581 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
22582 (duplicate_and_interleave): New function.
22583 (vect_get_constant_vectors): Use gimple_build_vector for
22584 constant-length vectors and suitable variable-length constant
22585 vectors. Use duplicate_and_interleave for other variable-length
22586 vectors. Don't defer the update when inserting new statements.
22588 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22589 Alan Hayward <alan.hayward@arm.com>
22590 David Sherwood <david.sherwood@arm.com>
22592 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
22593 min_profitable_iters doesn't go negative.
22595 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22596 Alan Hayward <alan.hayward@arm.com>
22597 David Sherwood <david.sherwood@arm.com>
22599 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
22600 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
22601 * optabs.def (vec_mask_load_lanes_optab): New optab.
22602 (vec_mask_store_lanes_optab): Likewise.
22603 * internal-fn.def (MASK_LOAD_LANES): New internal function.
22604 (MASK_STORE_LANES): Likewise.
22605 * internal-fn.c (mask_load_lanes_direct): New macro.
22606 (mask_store_lanes_direct): Likewise.
22607 (expand_mask_load_optab_fn): Handle masked operations.
22608 (expand_mask_load_lanes_optab_fn): New macro.
22609 (expand_mask_store_optab_fn): Handle masked operations.
22610 (expand_mask_store_lanes_optab_fn): New macro.
22611 (direct_mask_load_lanes_optab_supported_p): Likewise.
22612 (direct_mask_store_lanes_optab_supported_p): Likewise.
22613 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
22615 (vect_load_lanes_supported): Likewise.
22616 * tree-vect-data-refs.c (strip_conversion): New function.
22617 (can_group_stmts_p): Likewise.
22618 (vect_analyze_data_ref_accesses): Use it instead of checking
22619 for a pair of assignments.
22620 (vect_store_lanes_supported): Take a masked_p parameter.
22621 (vect_load_lanes_supported): Likewise.
22622 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
22623 vect_store_lanes_supported and vect_load_lanes_supported.
22624 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
22625 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
22626 parameter. Don't allow gaps for masked accesses.
22627 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
22628 and vect_load_lanes_supported.
22629 (get_load_store_type): Take a masked_p parameter and update
22630 call to get_group_load_store_type.
22631 (vectorizable_store): Update call to get_load_store_type.
22632 Handle IFN_MASK_STORE_LANES.
22633 (vectorizable_load): Update call to get_load_store_type.
22634 Handle IFN_MASK_LOAD_LANES.
22636 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22637 Alan Hayward <alan.hayward@arm.com>
22638 David Sherwood <david.sherwood@arm.com>
22640 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
22642 * config/aarch64/aarch64-protos.h
22643 (aarch64_sve_struct_memory_operand_p): Declare.
22644 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
22645 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
22646 (VPRED, vpred): Handle SVE structure modes.
22647 * config/aarch64/constraints.md (Utx): New constraint.
22648 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
22649 (aarch64_sve_struct_nonimmediate_operand): New predicates.
22650 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
22651 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
22652 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
22653 structure modes. Split into pieces after RA.
22654 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
22655 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
22657 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
22658 SVE structure modes.
22659 (aarch64_classify_address): Likewise.
22660 (sizetochar): Move earlier in file.
22661 (aarch64_print_operand): Handle SVE register lists.
22662 (aarch64_array_mode): New function.
22663 (aarch64_sve_struct_memory_operand_p): Likewise.
22664 (TARGET_ARRAY_MODE): Redefine.
22666 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22667 Alan Hayward <alan.hayward@arm.com>
22668 David Sherwood <david.sherwood@arm.com>
22670 * target.def (array_mode): New target hook.
22671 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
22672 * doc/tm.texi: Regenerate.
22673 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
22674 * hooks.c (hook_optmode_mode_uhwi_none): New function.
22675 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
22676 targetm.array_mode.
22677 * stor-layout.c (mode_for_array): Likewise. Support polynomial
22680 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22681 Alan Hayward <alan.hayward@arm.com>
22682 David Sherwood <david.sherwood@arm.com>
22684 * fold-const.c (fold_binary_loc): Check the argument types
22685 rather than the result type when testing for a vector operation.
22687 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22689 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
22690 * doc/tm.texi: Regenerate.
22692 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22693 Alan Hayward <alan.hayward@arm.com>
22694 David Sherwood <david.sherwood@arm.com>
22696 * doc/invoke.texi (-msve-vector-bits=): Document new option.
22697 (sve): Document new AArch64 extension.
22698 * doc/md.texi (w): Extend the description of the AArch64
22699 constraint to include SVE vectors.
22700 (Upl, Upa): Document new AArch64 predicate constraints.
22701 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
22703 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
22704 (msve-vector-bits=): New option.
22705 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
22706 SVE when these are disabled.
22707 (sve): New extension.
22708 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
22709 modes. Adjust their number of units based on aarch64_sve_vg.
22710 (MAX_BITSIZE_MODE_ANY_MODE): Define.
22711 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
22712 aarch64_addr_query_type.
22713 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
22714 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
22715 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
22716 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
22717 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
22718 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
22719 (aarch64_simd_imm_zero_p): Delete.
22720 (aarch64_check_zero_based_sve_index_immediate): Declare.
22721 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
22722 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
22723 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
22724 (aarch64_sve_float_mul_immediate_p): Likewise.
22725 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
22726 rather than an rtx.
22727 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
22728 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
22729 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
22730 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
22731 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
22732 (aarch64_regmode_natural_size): Likewise.
22733 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
22734 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
22736 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
22737 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
22738 for VG and the SVE predicate registers.
22739 (V_ALIASES): Add a "z"-prefixed alias.
22740 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
22741 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
22742 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
22743 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
22744 (REG_CLASS_NAMES): Add entries for them.
22745 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
22746 and the predicate registers.
22747 (aarch64_sve_vg): Declare.
22748 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
22749 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
22750 (REGMODE_NATURAL_SIZE): Define.
22751 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
22753 * config/aarch64/aarch64.c: Include cfgrtl.h.
22754 (simd_immediate_info): Add a constructor for series vectors,
22755 and an associated step field.
22756 (aarch64_sve_vg): New variable.
22757 (aarch64_dbx_register_number): Handle VG and the predicate registers.
22758 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
22759 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
22760 (VEC_ANY_DATA, VEC_STRUCT): New constants.
22761 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
22762 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
22763 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
22764 (aarch64_get_mask_mode): New functions.
22765 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
22766 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
22767 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
22768 predicate modes and predicate registers. Explicitly restrict
22769 GPRs to modes of 16 bytes or smaller. Only allow FP registers
22770 to store a vector mode if it is recognized by
22771 aarch64_classify_vector_mode.
22772 (aarch64_regmode_natural_size): New function.
22773 (aarch64_hard_regno_caller_save_mode): Return the original mode
22775 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
22776 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
22777 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
22778 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
22780 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
22781 does not overlap dest if the function is frame-related. Handle
22783 (aarch64_split_add_offset): New function.
22784 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
22785 them aarch64_add_offset.
22786 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
22787 and update call to aarch64_sub_sp.
22788 (aarch64_add_cfa_expression): New function.
22789 (aarch64_expand_prologue): Pass extra temporary registers to the
22790 functions above. Handle the case in which we need to emit new
22791 DW_CFA_expressions for registers that were originally saved
22792 relative to the stack pointer, but now have to be expressed
22793 relative to the frame pointer.
22794 (aarch64_output_mi_thunk): Pass extra temporary registers to the
22796 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
22797 IP0 and IP1 values for SVE frames.
22798 (aarch64_expand_vec_series): New function.
22799 (aarch64_expand_sve_widened_duplicate): Likewise.
22800 (aarch64_expand_sve_const_vector): Likewise.
22801 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
22802 Handle SVE constants. Use emit_move_insn to move a force_const_mem
22803 into the register, rather than emitting a SET directly.
22804 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
22805 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
22806 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
22807 (offset_9bit_signed_scaled_p): New functions.
22808 (aarch64_replicate_bitmask_imm): New function.
22809 (aarch64_bitmask_imm): Use it.
22810 (aarch64_cannot_force_const_mem): Reject expressions involving
22811 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
22812 (aarch64_classify_index): Handle SVE indices, by requiring
22813 a plain register index with a scale that matches the element size.
22814 (aarch64_classify_address): Handle SVE addresses. Assert that
22815 the mode of the address is VOIDmode or an integer mode.
22816 Update call to aarch64_classify_symbol.
22817 (aarch64_classify_symbolic_expression): Update call to
22818 aarch64_classify_symbol.
22819 (aarch64_const_vec_all_in_range_p): New function.
22820 (aarch64_print_vector_float_operand): Likewise.
22821 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
22822 "vN" for FP registers with SVE modes. Handle (const ...) vectors
22823 and the FP immediates 1.0 and 0.5.
22824 (aarch64_print_address_internal): Handle SVE addresses.
22825 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
22826 (aarch64_regno_regclass): Handle predicate registers.
22827 (aarch64_secondary_reload): Handle big-endian reloads of SVE
22829 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
22830 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
22831 (aarch64_convert_sve_vector_bits): New function.
22832 (aarch64_override_options): Use it to handle -msve-vector-bits=.
22833 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
22834 rather than an rtx.
22835 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
22836 Handle SVE vector and predicate modes. Accept VL-based constants
22837 that need only one temporary register, and VL offsets that require
22838 no temporary registers.
22839 (aarch64_conditional_register_usage): Mark the predicate registers
22840 as fixed if SVE isn't available.
22841 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
22842 Return true for SVE vector and predicate modes.
22843 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
22844 rather than an unsigned int. Handle SVE modes.
22845 (aarch64_preferred_simd_mode): Update call accordingly. Handle
22847 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
22849 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
22850 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
22851 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
22852 (aarch64_sve_float_mul_immediate_p): New functions.
22853 (aarch64_sve_valid_immediate): New function.
22854 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
22855 Explicitly reject structure modes. Check for INDEX constants.
22856 Handle PTRUE and PFALSE constants.
22857 (aarch64_check_zero_based_sve_index_immediate): New function.
22858 (aarch64_simd_imm_zero_p): Delete.
22859 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
22860 vector modes. Accept constants in the range of CNT[BHWD].
22861 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
22862 ask for an Advanced SIMD mode.
22863 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
22864 (aarch64_simd_vector_alignment): Handle SVE predicates.
22865 (aarch64_vectorize_preferred_vector_alignment): New function.
22866 (aarch64_simd_vector_alignment_reachable): Use it instead of
22868 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
22869 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
22871 (MAX_VECT_LEN): Delete.
22872 (expand_vec_perm_d): Add a vec_flags field.
22873 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
22874 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
22875 (aarch64_evpc_ext): Don't apply a big-endian lane correction
22877 (aarch64_evpc_rev): Rename to...
22878 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
22879 (aarch64_evpc_rev_global): New function.
22880 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
22881 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
22883 (aarch64_evpc_sve_tbl): New function.
22884 (aarch64_expand_vec_perm_const_1): Update after rename of
22885 aarch64_evpc_rev. Handle SVE permutes too, trying
22886 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
22887 than aarch64_evpc_tbl.
22888 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
22889 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
22890 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
22891 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
22892 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
22893 (aarch64_expand_sve_vcond): New functions.
22894 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
22895 of aarch64_vector_mode_p.
22896 (aarch64_dwarf_poly_indeterminate_value): New function.
22897 (aarch64_compute_pressure_classes): Likewise.
22898 (aarch64_can_change_mode_class): Likewise.
22899 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
22900 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
22901 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
22902 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
22903 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
22904 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
22905 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
22906 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
22908 (Dn, Dl, Dr): Accept const as well as const_vector.
22909 (Dz): Likewise. Compare against CONST0_RTX.
22910 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
22911 of "vector" where appropriate.
22912 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
22913 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
22914 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
22915 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
22916 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
22917 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
22918 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
22919 (v_int_equiv): Extend to SVE modes.
22920 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
22922 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
22923 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
22924 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
22925 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
22926 (SVE_COND_FP_CMP): New int iterators.
22927 (perm_hilo): Handle the new unpack unspecs.
22928 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
22930 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
22931 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
22932 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
22933 (aarch64_equality_operator, aarch64_constant_vector_operand)
22934 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
22935 (aarch64_sve_nonimmediate_operand): Likewise.
22936 (aarch64_sve_general_operand): Likewise.
22937 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
22938 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
22939 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
22940 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
22941 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
22942 (aarch64_sve_float_arith_immediate): Likewise.
22943 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
22944 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
22945 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
22946 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
22947 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
22948 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
22949 (aarch64_sve_float_arith_operand): Likewise.
22950 (aarch64_sve_float_arith_with_sub_operand): Likewise.
22951 (aarch64_sve_float_mul_operand): Likewise.
22952 (aarch64_sve_vec_perm_operand): Likewise.
22953 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
22954 (aarch64_mov_operand): Accept const_poly_int and const_vector.
22955 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
22956 as well as const_vector.
22957 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
22958 in file. Use CONST0_RTX and CONSTM1_RTX.
22959 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
22960 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
22961 Use aarch64_simd_imm_zero.
22962 * config/aarch64/aarch64-sve.md: New file.
22963 * config/aarch64/aarch64.md: Include it.
22964 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
22965 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
22966 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
22967 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
22968 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
22969 (sve): New attribute.
22970 (enabled): Disable instructions with the sve attribute unless
22972 (movqi, movhi): Pass CONST_POLY_INT operaneds through
22973 aarch64_expand_mov_immediate.
22974 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
22975 CNT[BHSD] immediates.
22976 (movti): Split CONST_POLY_INT moves into two halves.
22977 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
22978 Split additions that need a temporary here if the destination
22979 is the stack pointer.
22980 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
22981 (*add<mode>3_poly_1): New instruction.
22982 (set_clobber_cc): New expander.
22984 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22986 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
22987 parameter and use it instead of GET_MODE_SIZE (innermode). Use
22988 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
22989 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
22990 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
22991 Change innermode from fixed_mode_size to machine_mode.
22992 (simplify_subreg): Update call accordingly. Handle a constant-sized
22993 subreg of a variable-length CONST_VECTOR.
22995 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22996 Alan Hayward <alan.hayward@arm.com>
22997 David Sherwood <david.sherwood@arm.com>
22999 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
23000 (add_offset_to_base): New function, split out from...
23001 (create_mem_ref): ...here. When handling a scale other than 1,
23002 check first whether the address is valid without the offset.
23003 Add it into the base if so, leaving the index and scale as-is.
23005 2018-01-12 Jakub Jelinek <jakub@redhat.com>
23008 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
23009 fold_for_warn before checking if arg2 is INTEGER_CST.
23011 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
23013 * config/rs6000/predicates.md (load_multiple_operation): Delete.
23014 (store_multiple_operation): Delete.
23015 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
23016 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
23017 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
23018 guarded by TARGET_STRING.
23019 (rs6000_output_load_multiple): Delete.
23020 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
23021 OPTION_MASK_STRING / TARGET_STRING handling.
23022 (print_operand) <'N', 'O'>: Add comment that these are unused now.
23023 (const rs6000_opt_masks) <"string">: Change mask to 0.
23024 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
23025 (MASK_STRING): Delete.
23026 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
23028 (load_multiple): Delete.
23035 (store_multiple): Delete.
23042 (movmemsi_8reg): Delete.
23043 (corresponding unnamed define_insn): Delete.
23044 (movmemsi_6reg): Delete.
23045 (corresponding unnamed define_insn): Delete.
23046 (movmemsi_4reg): Delete.
23047 (corresponding unnamed define_insn): Delete.
23048 (movmemsi_2reg): Delete.
23049 (corresponding unnamed define_insn): Delete.
23050 (movmemsi_1reg): Delete.
23051 (corresponding unnamed define_insn): Delete.
23052 * config/rs6000/rs6000.opt (mno-string): New.
23053 (mstring): Replace by deprecation warning stub.
23054 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
23056 2018-01-12 Jakub Jelinek <jakub@redhat.com>
23058 * regrename.c (regrename_do_replace): If replacing the same
23059 reg multiple times, try to reuse last created gen_raw_REG.
23062 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
23063 main to workaround a bug in GDB.
23065 2018-01-12 Tom de Vries <tom@codesourcery.com>
23068 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
23070 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
23072 PR rtl-optimization/80481
23073 * ira-color.c (get_cap_member): New function.
23074 (allocnos_conflict_by_live_ranges_p): Use it.
23075 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
23076 (setup_slot_coalesced_allocno_live_ranges): Ditto.
23078 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
23081 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
23082 (*saddl_se_1): Ditto.
23083 (*ssubsi_1): Ditto.
23084 (*ssubl_se_1): Ditto.
23086 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
23088 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
23089 rather than wi::to_widest for DR_INITs.
23090 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
23091 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
23092 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
23094 (vect_analyze_group_access_1): Note that here.
23096 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
23098 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
23099 polynomial type sizes.
23101 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
23103 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
23104 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
23105 (gimple_add_tmp_var): Likewise.
23107 2018-01-12 Martin Liska <mliska@suse.cz>
23109 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
23110 (gimple_alloc_sizes): Likewise.
23111 (dump_gimple_statistics): Use PRIu64 in printf format.
23112 * gimple.h: Change uint64_t to int.
23114 2018-01-12 Martin Liska <mliska@suse.cz>
23116 * tree-core.h: Use uint64_t instead of int.
23117 * tree.c (tree_node_counts): Likewise.
23118 (tree_node_sizes): Likewise.
23119 (dump_tree_statistics): Use PRIu64 in printf format.
23121 2018-01-12 Martin Liska <mliska@suse.cz>
23123 * Makefile.in: As qsort_chk is implemented in vec.c, add
23124 vec.o to linkage of gencfn-macros.
23125 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
23126 passing the info to record_node_allocation_statistics.
23127 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
23129 * ggc-common.c (struct ggc_usage): Add operator== and use
23130 it in operator< and compare function.
23131 * mem-stats.h (struct mem_usage): Likewise.
23132 * vec.c (struct vec_usage): Remove operator< and compare
23133 function. Can be simply inherited.
23135 2018-01-12 Martin Jambor <mjambor@suse.cz>
23138 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
23139 * tree-ssa-math-opts.c: Include domwalk.h.
23140 (convert_mult_to_fma_1): New function.
23141 (fma_transformation_info): New type.
23142 (fma_deferring_state): Likewise.
23143 (cancel_fma_deferring): New function.
23144 (result_of_phi): Likewise.
23145 (last_fma_candidate_feeds_initial_phi): Likewise.
23146 (convert_mult_to_fma): Added deferring logic, split actual
23147 transformation to convert_mult_to_fma_1.
23148 (math_opts_dom_walker): New type.
23149 (math_opts_dom_walker::after_dom_children): New method, body moved
23150 here from pass_optimize_widening_mul::execute, added deferring logic
23152 (pass_optimize_widening_mul::execute): Moved most of code to
23153 math_opts_dom_walker::after_dom_children.
23154 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
23155 * config/i386/i386.c (ix86_option_override_internal): Added
23156 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
23158 2018-01-12 Richard Biener <rguenther@suse.de>
23161 * dwarf2out.c (gen_variable_die): Do not reset old_die for
23162 inline instance vars.
23164 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
23167 * config/rx/rx.c (rx_is_restricted_memory_address):
23168 Handle SUBREG case.
23170 2018-01-12 Richard Biener <rguenther@suse.de>
23172 PR tree-optimization/80846
23173 * target.def (split_reduction): New target hook.
23174 * targhooks.c (default_split_reduction): New function.
23175 * targhooks.h (default_split_reduction): Declare.
23176 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
23177 target requests first reduce vectors by combining low and high
23179 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
23180 (get_vectype_for_scalar_type_and_size): Export.
23181 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
23182 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
23183 * doc/tm.texi: Regenerate.
23184 * config/i386/i386.c (ix86_split_reduction): Implement
23185 TARGET_VECTORIZE_SPLIT_REDUCTION.
23187 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
23190 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
23191 in PIC mode except for TARGET_VXWORKS_RTP.
23192 * config/sparc/sparc.c: Include cfgrtl.h.
23193 (TARGET_INIT_PIC_REG): Define.
23194 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
23195 (sparc_pic_register_p): New predicate.
23196 (sparc_legitimate_address_p): Use it.
23197 (sparc_legitimize_pic_address): Likewise.
23198 (sparc_delegitimize_address): Likewise.
23199 (sparc_mode_dependent_address_p): Likewise.
23200 (gen_load_pcrel_sym): Remove 4th parameter.
23201 (load_got_register): Adjust call to above. Remove obsolete stuff.
23202 (sparc_expand_prologue): Do not call load_got_register here.
23203 (sparc_flat_expand_prologue): Likewise.
23204 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
23205 (sparc_use_pseudo_pic_reg): New function.
23206 (sparc_init_pic_reg): Likewise.
23207 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
23208 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
23210 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
23212 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
23213 Add item for branch_cost.
23215 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
23217 PR rtl-optimization/83565
23218 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
23219 not extend the result to a larger mode for rotate operations.
23220 (num_sign_bit_copies1): Likewise.
23222 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
23225 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
23227 Use values-Xc.o for -pedantic.
23228 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
23230 2018-01-12 Martin Liska <mliska@suse.cz>
23233 * ipa-devirt.c (final_warning_record::grow_type_warnings):
23235 (possible_polymorphic_call_targets): Use it.
23236 (ipa_devirt): Likewise.
23238 2018-01-12 Martin Liska <mliska@suse.cz>
23240 * profile-count.h (enum profile_quality): Use 0 as invalid
23241 enum value of profile_quality.
23243 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
23245 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
23246 -mext-string options.
23248 2018-01-12 Richard Biener <rguenther@suse.de>
23250 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
23251 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
23252 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
23254 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
23256 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
23258 * configure.ac (--with-long-double-format): Add support for the
23259 configuration option to change the default long double format on
23261 * config.gcc (powerpc*-linux*-*): Likewise.
23262 * configure: Regenerate.
23263 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
23264 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
23265 used without modification.
23267 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23269 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
23270 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
23271 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
23272 MISC_BUILTIN_SPEC_BARRIER.
23273 (rs6000_init_builtins): Likewise.
23274 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
23276 (speculation_barrier): New define_insn.
23277 * doc/extend.texi: Document __builtin_speculation_barrier.
23279 2018-01-11 Jakub Jelinek <jakub@redhat.com>
23282 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
23283 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
23284 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
23286 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
23287 integral modes instead of "ss" and "sd".
23288 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
23289 vectors with 32-bit and 64-bit elements.
23290 (vecdupssescalarmodesuffix): New mode attribute.
23291 (vec_dup<mode>): Use it.
23293 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
23296 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
23297 frame if argument is passed on stack.
23299 2018-01-11 Jakub Jelinek <jakub@redhat.com>
23302 * ree.c (combine_reaching_defs): Optimize also
23303 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
23304 reg2=any_extend(exp); reg1=reg2;, formatting fix.
23306 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
23308 PR middle-end/83189
23309 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
23311 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
23313 PR middle-end/83718
23314 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
23315 after they are computed.
23317 2018-01-11 Bin Cheng <bin.cheng@arm.com>
23319 PR tree-optimization/83695
23320 * gimple-loop-linterchange.cc
23321 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
23322 reset cached scev information after interchange.
23323 (pass_linterchange::execute): Remove call to scev_reset_htab.
23325 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23327 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
23328 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
23329 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
23330 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
23331 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
23332 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
23333 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
23334 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
23335 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
23336 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
23337 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
23338 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
23339 (V_lane_reg): Likewise.
23340 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
23342 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
23343 (vfmal_lane_low<mode>_intrinsic,
23344 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
23345 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
23346 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
23347 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
23348 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
23349 vfmsl_lane_high<mode>_intrinsic): New define_insns.
23351 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23353 * config/arm/arm-cpus.in (fp16fml): New feature.
23354 (ALL_SIMD): Add fp16fml.
23355 (armv8.2-a): Add fp16fml as an option.
23356 (armv8.3-a): Likewise.
23357 (armv8.4-a): Add fp16fml as part of fp16.
23358 * config/arm/arm.h (TARGET_FP16FML): Define.
23359 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
23361 * config/arm/arm-modes.def (V2HF): Define.
23362 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
23363 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
23364 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
23365 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
23366 vfmsl_low, vfmsl_high): New set of builtins.
23367 * config/arm/iterators.md (PLUSMINUS): New code iterator.
23368 (vfml_op): New code attribute.
23369 (VFMLHALVES): New int iterator.
23370 (VFML, VFMLSEL): New mode attributes.
23371 (V_reg): Define mapping for V2HF.
23372 (V_hi, V_lo): New mode attributes.
23373 (VF_constraint): Likewise.
23374 (vfml_half, vfml_half_selector): New int attributes.
23375 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
23377 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
23378 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
23380 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
23381 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
23382 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
23383 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
23385 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
23386 Document new effective target and option set.
23388 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23390 * config/arm/arm-cpus.in (armv8_4): New feature.
23391 (ARMv8_4a): New fgroup.
23392 (armv8.4-a): New arch.
23393 * config/arm/arm-tables.opt: Regenerate.
23394 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
23395 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
23396 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
23397 Add matching rules for -march=armv8.4-a and extensions.
23398 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
23400 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
23403 * config/rx/rx.md (BW): New mode attribute.
23404 (sync_lock_test_and_setsi): Add mode suffix to insn output.
23406 2018-01-11 Richard Biener <rguenther@suse.de>
23408 PR tree-optimization/83435
23409 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
23410 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
23411 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
23413 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23414 Alan Hayward <alan.hayward@arm.com>
23415 David Sherwood <david.sherwood@arm.com>
23417 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
23419 (aarch64_classify_address): Initialize it. Track polynomial offsets.
23420 (aarch64_print_address_internal): Use it to check for a zero offset.
23422 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23423 Alan Hayward <alan.hayward@arm.com>
23424 David Sherwood <david.sherwood@arm.com>
23426 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
23427 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
23428 Return a poly_int64 rather than a HOST_WIDE_INT.
23429 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
23430 rather than a HOST_WIDE_INT.
23431 * config/aarch64/aarch64.h (aarch64_frame): Protect with
23432 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
23433 hard_fp_offset, frame_size, initial_adjust, callee_offset and
23434 final_offset from HOST_WIDE_INT to poly_int64.
23435 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
23436 to_constant when getting the number of units in an Advanced SIMD
23438 (aarch64_builtin_vectorized_function): Check for a constant number
23440 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
23442 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
23443 attribute instead of GET_MODE_NUNITS.
23444 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
23445 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
23446 GET_MODE_SIZE for fixed-size registers.
23447 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
23448 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
23449 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
23450 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
23451 (aarch64_print_operand, aarch64_print_address_internal)
23452 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
23453 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
23454 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
23455 Handle polynomial GET_MODE_SIZE.
23456 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
23457 wider than SImode without modification.
23458 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
23459 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
23460 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
23461 passing and returning SVE modes.
23462 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
23463 rather than GEN_INT.
23464 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
23465 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
23466 (aarch64_allocate_and_probe_stack_space): Likewise.
23467 (aarch64_layout_frame): Cope with polynomial offsets.
23468 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
23469 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
23470 polynomial offsets.
23471 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
23472 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
23473 poly_int64 rather than a HOST_WIDE_INT.
23474 (aarch64_get_separate_components, aarch64_process_components)
23475 (aarch64_expand_prologue, aarch64_expand_epilogue)
23476 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
23477 (aarch64_anchor_offset): New function, split out from...
23478 (aarch64_legitimize_address): ...here.
23479 (aarch64_builtin_vectorization_cost): Handle polynomial
23480 TYPE_VECTOR_SUBPARTS.
23481 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
23483 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
23484 number of elements from the PARALLEL rather than the mode.
23485 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
23486 rather than GET_MODE_BITSIZE.
23487 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
23488 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
23489 (aarch64_expand_vec_perm_const_1): Handle polynomial
23490 d->perm.length () and d->perm elements.
23491 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
23492 Apply to_constant to d->perm elements.
23493 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
23494 polynomial CONST_VECTOR_NUNITS.
23495 (aarch64_move_pointer): Take amount as a poly_int64 rather
23497 (aarch64_progress_pointer): Avoid temporary variable.
23498 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
23499 the mode attribute instead of GET_MODE.
23501 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23502 Alan Hayward <alan.hayward@arm.com>
23503 David Sherwood <david.sherwood@arm.com>
23505 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
23506 x exists before using it.
23507 (aarch64_add_constant_internal): Rename to...
23508 (aarch64_add_offset_1): ...this. Replace regnum with separate
23509 src and dest rtxes. Handle the case in which they're different,
23510 including when the offset is zero. Replace scratchreg with an rtx.
23511 Use 2 additions if there is no spare register into which we can
23512 move a 16-bit constant.
23513 (aarch64_add_constant): Delete.
23514 (aarch64_add_offset): Replace reg with separate src and dest
23515 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
23516 Use aarch64_add_offset_1.
23517 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
23518 an rtx rather than an int. Take the delta as a poly_int64
23519 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
23520 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
23521 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
23522 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
23523 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
23524 and aarch64_add_sp.
23525 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
23526 aarch64_add_constant.
23528 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23530 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
23531 Use scalar_float_mode.
23533 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23535 * config/aarch64/aarch64-simd.md
23536 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
23537 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
23538 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
23539 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
23540 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
23541 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
23542 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
23543 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
23544 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
23545 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
23547 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23550 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
23551 targ_options->x_arm_arch_string is non NULL.
23553 2018-01-11 Tamar Christina <tamar.christina@arm.com>
23555 * config/aarch64/aarch64.h
23556 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
23558 2018-01-11 Sudakshina Das <sudi.das@arm.com>
23561 * expmed.c (emit_store_flag_force): Swap if const op0
23562 and change VOIDmode to mode of op0.
23564 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23566 PR rtl-optimization/83761
23567 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
23568 than bytes to mode_for_size.
23570 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
23572 PR middle-end/83189
23573 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
23574 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
23577 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
23579 PR middle-end/83575
23580 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
23581 when in layout mode.
23582 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
23583 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
23586 2018-01-10 Michael Collison <michael.collison@arm.com>
23588 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
23589 * config/aarch64/aarch64-option-extension.def: Add
23590 AARCH64_OPT_EXTENSION of 'fp16fml'.
23591 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23592 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
23593 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
23594 * config/aarch64/constraints.md (Ui7): New constraint.
23595 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
23596 (VFMLA_SEL_W): Ditto.
23599 (VFMLA16_LOW): New int iterator.
23600 (VFMLA16_HIGH): Ditto.
23601 (UNSPEC_FMLAL): New unspec.
23602 (UNSPEC_FMLSL): Ditto.
23603 (UNSPEC_FMLAL2): Ditto.
23604 (UNSPEC_FMLSL2): Ditto.
23605 (f16mac): New code attribute.
23606 * config/aarch64/aarch64-simd-builtins.def
23607 (aarch64_fmlal_lowv2sf): Ditto.
23608 (aarch64_fmlsl_lowv2sf): Ditto.
23609 (aarch64_fmlalq_lowv4sf): Ditto.
23610 (aarch64_fmlslq_lowv4sf): Ditto.
23611 (aarch64_fmlal_highv2sf): Ditto.
23612 (aarch64_fmlsl_highv2sf): Ditto.
23613 (aarch64_fmlalq_highv4sf): Ditto.
23614 (aarch64_fmlslq_highv4sf): Ditto.
23615 (aarch64_fmlal_lane_lowv2sf): Ditto.
23616 (aarch64_fmlsl_lane_lowv2sf): Ditto.
23617 (aarch64_fmlal_laneq_lowv2sf): Ditto.
23618 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
23619 (aarch64_fmlalq_lane_lowv4sf): Ditto.
23620 (aarch64_fmlsl_lane_lowv4sf): Ditto.
23621 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
23622 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
23623 (aarch64_fmlal_lane_highv2sf): Ditto.
23624 (aarch64_fmlsl_lane_highv2sf): Ditto.
23625 (aarch64_fmlal_laneq_highv2sf): Ditto.
23626 (aarch64_fmlsl_laneq_highv2sf): Ditto.
23627 (aarch64_fmlalq_lane_highv4sf): Ditto.
23628 (aarch64_fmlsl_lane_highv4sf): Ditto.
23629 (aarch64_fmlalq_laneq_highv4sf): Ditto.
23630 (aarch64_fmlsl_laneq_highv4sf): Ditto.
23631 * config/aarch64/aarch64-simd.md:
23632 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
23633 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
23634 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
23635 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
23636 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
23637 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
23638 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
23639 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
23640 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
23641 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
23642 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
23643 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
23644 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
23645 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
23646 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
23647 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
23648 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
23649 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
23650 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
23651 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
23652 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
23653 (vfmlsl_low_u32): Ditto.
23654 (vfmlalq_low_u32): Ditto.
23655 (vfmlslq_low_u32): Ditto.
23656 (vfmlal_high_u32): Ditto.
23657 (vfmlsl_high_u32): Ditto.
23658 (vfmlalq_high_u32): Ditto.
23659 (vfmlslq_high_u32): Ditto.
23660 (vfmlal_lane_low_u32): Ditto.
23661 (vfmlsl_lane_low_u32): Ditto.
23662 (vfmlal_laneq_low_u32): Ditto.
23663 (vfmlsl_laneq_low_u32): Ditto.
23664 (vfmlalq_lane_low_u32): Ditto.
23665 (vfmlslq_lane_low_u32): Ditto.
23666 (vfmlalq_laneq_low_u32): Ditto.
23667 (vfmlslq_laneq_low_u32): Ditto.
23668 (vfmlal_lane_high_u32): Ditto.
23669 (vfmlsl_lane_high_u32): Ditto.
23670 (vfmlal_laneq_high_u32): Ditto.
23671 (vfmlsl_laneq_high_u32): Ditto.
23672 (vfmlalq_lane_high_u32): Ditto.
23673 (vfmlslq_lane_high_u32): Ditto.
23674 (vfmlalq_laneq_high_u32): Ditto.
23675 (vfmlslq_laneq_high_u32): Ditto.
23676 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
23677 (AARCH64_FL_FOR_ARCH8_4): New.
23678 (AARCH64_ISA_F16FML): New ISA flag.
23679 (TARGET_F16FML): New feature flag for fp16fml.
23680 (doc/invoke.texi): Document new fp16fml option.
23682 2018-01-10 Michael Collison <michael.collison@arm.com>
23684 * config/aarch64/aarch64-builtins.c:
23685 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
23686 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23687 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
23688 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
23689 (AARCH64_ISA_SHA3): New ISA flag.
23690 (TARGET_SHA3): New feature flag for sha3.
23691 * config/aarch64/iterators.md (sha512_op): New int attribute.
23692 (CRYPTO_SHA512): New int iterator.
23693 (UNSPEC_SHA512H): New unspec.
23694 (UNSPEC_SHA512H2): Ditto.
23695 (UNSPEC_SHA512SU0): Ditto.
23696 (UNSPEC_SHA512SU1): Ditto.
23697 * config/aarch64/aarch64-simd-builtins.def
23698 (aarch64_crypto_sha512hqv2di): New builtin.
23699 (aarch64_crypto_sha512h2qv2di): Ditto.
23700 (aarch64_crypto_sha512su0qv2di): Ditto.
23701 (aarch64_crypto_sha512su1qv2di): Ditto.
23702 (aarch64_eor3qv8hi): Ditto.
23703 (aarch64_rax1qv2di): Ditto.
23704 (aarch64_xarqv2di): Ditto.
23705 (aarch64_bcaxqv8hi): Ditto.
23706 * config/aarch64/aarch64-simd.md:
23707 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
23708 (aarch64_crypto_sha512su0qv2di): Ditto.
23709 (aarch64_crypto_sha512su1qv2di): Ditto.
23710 (aarch64_eor3qv8hi): Ditto.
23711 (aarch64_rax1qv2di): Ditto.
23712 (aarch64_xarqv2di): Ditto.
23713 (aarch64_bcaxqv8hi): Ditto.
23714 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
23715 (vsha512h2q_u64): Ditto.
23716 (vsha512su0q_u64): Ditto.
23717 (vsha512su1q_u64): Ditto.
23718 (veor3q_u16): Ditto.
23719 (vrax1q_u64): Ditto.
23720 (vxarq_u64): Ditto.
23721 (vbcaxq_u16): Ditto.
23722 * config/arm/types.md (crypto_sha512): New type attribute.
23723 (crypto_sha3): Ditto.
23724 (doc/invoke.texi): Document new sha3 option.
23726 2018-01-10 Michael Collison <michael.collison@arm.com>
23728 * config/aarch64/aarch64-builtins.c:
23729 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
23730 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23731 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
23732 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
23733 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
23734 (AARCH64_ISA_SM4): New ISA flag.
23735 (TARGET_SM4): New feature flag for sm4.
23736 * config/aarch64/aarch64-simd-builtins.def
23737 (aarch64_sm3ss1qv4si): Ditto.
23738 (aarch64_sm3tt1aq4si): Ditto.
23739 (aarch64_sm3tt1bq4si): Ditto.
23740 (aarch64_sm3tt2aq4si): Ditto.
23741 (aarch64_sm3tt2bq4si): Ditto.
23742 (aarch64_sm3partw1qv4si): Ditto.
23743 (aarch64_sm3partw2qv4si): Ditto.
23744 (aarch64_sm4eqv4si): Ditto.
23745 (aarch64_sm4ekeyqv4si): Ditto.
23746 * config/aarch64/aarch64-simd.md:
23747 (aarch64_sm3ss1qv4si): Ditto.
23748 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
23749 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
23750 (aarch64_sm4eqv4si): Ditto.
23751 (aarch64_sm4ekeyqv4si): Ditto.
23752 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
23753 (sm3part_op): Ditto.
23754 (CRYPTO_SM3TT): Ditto.
23755 (CRYPTO_SM3PART): Ditto.
23756 (UNSPEC_SM3SS1): New unspec.
23757 (UNSPEC_SM3TT1A): Ditto.
23758 (UNSPEC_SM3TT1B): Ditto.
23759 (UNSPEC_SM3TT2A): Ditto.
23760 (UNSPEC_SM3TT2B): Ditto.
23761 (UNSPEC_SM3PARTW1): Ditto.
23762 (UNSPEC_SM3PARTW2): Ditto.
23763 (UNSPEC_SM4E): Ditto.
23764 (UNSPEC_SM4EKEY): Ditto.
23765 * config/aarch64/constraints.md (Ui2): New constraint.
23766 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
23767 * config/arm/types.md (crypto_sm3): New type attribute.
23768 (crypto_sm4): Ditto.
23769 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
23770 (vsm3tt1aq_u32): Ditto.
23771 (vsm3tt1bq_u32): Ditto.
23772 (vsm3tt2aq_u32): Ditto.
23773 (vsm3tt2bq_u32): Ditto.
23774 (vsm3partw1q_u32): Ditto.
23775 (vsm3partw2q_u32): Ditto.
23776 (vsm4eq_u32): Ditto.
23777 (vsm4ekeyq_u32): Ditto.
23778 (doc/invoke.texi): Document new sm4 option.
23780 2018-01-10 Michael Collison <michael.collison@arm.com>
23782 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
23783 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
23784 (AARCH64_FL_FOR_ARCH8_4): New.
23785 (AARCH64_FL_V8_4): New flag.
23786 (doc/invoke.texi): Document new armv8.4-a option.
23788 2018-01-10 Michael Collison <michael.collison@arm.com>
23790 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23791 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
23792 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
23793 * config/aarch64/aarch64-option-extension.def: Add
23794 AARCH64_OPT_EXTENSION of 'sha2'.
23795 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
23796 (crypto): Disable sha2 and aes if crypto disabled.
23797 (crypto): Enable aes and sha2 if enabled.
23798 (simd): Disable sha2 and aes if simd disabled.
23799 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
23801 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
23802 (TARGET_SHA2): New feature flag for sha2.
23803 (TARGET_AES): New feature flag for aes.
23804 * config/aarch64/aarch64-simd.md:
23805 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
23806 conditional on TARGET_AES.
23807 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
23808 (aarch64_crypto_sha1hsi): Make pattern conditional
23810 (aarch64_crypto_sha1hv4si): Ditto.
23811 (aarch64_be_crypto_sha1hv4si): Ditto.
23812 (aarch64_crypto_sha1su1v4si): Ditto.
23813 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
23814 (aarch64_crypto_sha1su0v4si): Ditto.
23815 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
23816 (aarch64_crypto_sha256su0v4si): Ditto.
23817 (aarch64_crypto_sha256su1v4si): Ditto.
23818 (doc/invoke.texi): Document new aes and sha2 options.
23820 2018-01-10 Martin Sebor <msebor@redhat.com>
23822 PR tree-optimization/83781
23823 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
23826 2018-01-11 Martin Sebor <msebor@gmail.com>
23827 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23829 PR tree-optimization/83501
23830 PR tree-optimization/81703
23832 * tree-ssa-strlen.c (get_string_cst): Rename...
23833 (get_string_len): ...to this. Handle global constants.
23834 (handle_char_store): Adjust.
23836 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
23837 Jim Wilson <jimw@sifive.com>
23839 * config/riscv/riscv-protos.h (riscv_output_return): New.
23840 * config/riscv/riscv.c (struct machine_function): New naked_p field.
23841 (riscv_attribute_table, riscv_output_return),
23842 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
23843 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
23844 (riscv_compute_frame_info): Only compute frame->mask if not a naked
23846 (riscv_expand_prologue): Add early return for naked function.
23847 (riscv_expand_epilogue): Likewise.
23848 (riscv_function_ok_for_sibcall): Return false for naked function.
23849 (riscv_set_current_function): New.
23850 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
23851 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
23852 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
23853 * doc/extend.texi (RISC-V Function Attributes): New.
23855 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
23857 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
23858 check for 128-bit long double before checking TCmode.
23859 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
23860 128-bit long doubles before checking TFmode or TCmode.
23861 (FLOAT128_IBM_P): Likewise.
23863 2018-01-10 Martin Sebor <msebor@redhat.com>
23865 PR tree-optimization/83671
23866 * builtins.c (c_strlen): Unconditionally return zero for the empty
23868 Use -Warray-bounds for warnings.
23869 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
23870 for non-constant array indices with COMPONENT_REF, arrays of
23871 arrays, and pointers to arrays.
23872 (gimple_fold_builtin_strlen): Determine and set length range for
23873 non-constant character arrays.
23875 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
23877 PR middle-end/81897
23878 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
23881 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
23883 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
23885 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
23888 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
23889 VECTOR_MEM_ALTIVEC_OR_VSX_P.
23890 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
23891 indexed_or_indirect_operand predicate.
23892 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
23893 (*vsx_le_perm_load_v8hi): Likewise.
23894 (*vsx_le_perm_load_v16qi): Likewise.
23895 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
23896 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
23897 (*vsx_le_perm_store_v8hi): Likewise.
23898 (*vsx_le_perm_store_v16qi): Likewise.
23899 (eight unnamed splitters): Likewise.
23901 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
23903 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
23904 * config/rs6000/emmintrin.h: Likewise.
23905 * config/rs6000/mmintrin.h: Likewise.
23906 * config/rs6000/xmmintrin.h: Likewise.
23908 2018-01-10 David Malcolm <dmalcolm@redhat.com>
23911 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
23913 * tree.c (tree_nop_conversion): Return true for location wrapper
23915 (maybe_wrap_with_location): New function.
23916 (selftest::check_strip_nops): New function.
23917 (selftest::test_location_wrappers): New function.
23918 (selftest::tree_c_tests): Call it.
23919 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
23920 (maybe_wrap_with_location): New decl.
23921 (EXPR_LOCATION_WRAPPER_P): New macro.
23922 (location_wrapper_p): New inline function.
23923 (tree_strip_any_location_wrapper): New inline function.
23925 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
23928 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
23929 stack_realign_offset for the largest alignment of stack slot
23931 (ix86_find_max_used_stack_alignment): New function.
23932 (ix86_finalize_stack_frame_flags): Use it. Set
23933 max_used_stack_alignment if we don't realign stack.
23934 * config/i386/i386.h (machine_function): Add
23935 max_used_stack_alignment.
23937 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
23939 * config/arm/arm.opt (-mbranch-cost): New option.
23940 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
23943 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
23946 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
23947 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
23949 2018-01-10 Richard Biener <rguenther@suse.de>
23952 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
23953 early out so it also covers the case where we have a non-NULL
23956 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
23958 PR tree-optimization/83753
23959 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
23960 for non-strided grouped accesses if the number of elements is 1.
23962 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
23965 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
23966 * i386.h (TARGET_USE_GATHER): Define.
23967 * x86-tune.def (X86_TUNE_USE_GATHER): New.
23969 2018-01-10 Martin Liska <mliska@suse.cz>
23972 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
23973 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
23975 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
23976 CLEANUP_NO_PARTITIONING is not set.
23978 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
23980 * doc/rtl.texi: Remove documentation of (const ...) wrappers
23981 for vectors, as a partial revert of r254296.
23982 * rtl.h (const_vec_p): Delete.
23983 (const_vec_duplicate_p): Don't test for vector CONSTs.
23984 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
23985 * expmed.c (make_tree): Likewise.
23988 * common.md (E, F): Use CONSTANT_P instead of checking for
23990 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
23991 checking for CONST_VECTOR.
23993 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
23995 PR middle-end/83575
23996 * predict.c (force_edge_cold): Handle in more sane way edges
23997 with no prediction.
23999 2018-01-09 Carl Love <cel@us.ibm.com>
24001 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
24003 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
24004 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
24005 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
24006 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
24007 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
24008 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
24009 * config/rs6000/rs6000-protos.h: Add extern defition for
24010 rs6000_generate_float2_double_code.
24011 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
24013 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
24014 (float2_v2df): Add define_expand.
24016 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
24019 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
24020 op_mode in the force_to_mode call.
24022 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
24024 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
24025 instead of checking each element individually.
24026 (aarch64_evpc_uzp): Likewise.
24027 (aarch64_evpc_zip): Likewise.
24028 (aarch64_evpc_ext): Likewise.
24029 (aarch64_evpc_rev): Likewise.
24030 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
24031 instead of checking each element individually. Return true without
24033 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
24034 whether all selected elements come from the same input, instead of
24035 checking each element individually. Remove calls to gen_rtx_REG,
24036 start_sequence and end_sequence and instead assert that no rtl is
24039 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
24041 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
24042 order of HIGH and CONST checks.
24044 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
24046 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
24047 if the destination isn't an SSA_NAME.
24049 2018-01-09 Richard Biener <rguenther@suse.de>
24051 PR tree-optimization/83668
24052 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
24054 (canonicalize_loop_form): ... here, renamed from ...
24055 (canonicalize_loop_closed_ssa_form): ... this and amended to
24056 swap successor edges for loop exit blocks to make us use
24057 the RPO order we need for initial schedule generation.
24059 2018-01-09 Joseph Myers <joseph@codesourcery.com>
24061 PR tree-optimization/64811
24062 * match.pd: When optimizing comparisons with Inf, avoid
24063 introducing or losing exceptions from comparisons with NaN.
24065 2018-01-09 Martin Liska <mliska@suse.cz>
24068 * asan.c (shadow_mem_size): Add gcc_assert.
24070 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
24072 Don't save registers in main().
24075 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
24076 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
24077 * config/avr/avr.c (avr_set_current_function): Don't error if
24078 naked, OS_task or OS_main are specified at the same time.
24079 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
24081 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
24083 * common/config/avr/avr-common.c (avr_option_optimization_table):
24084 Switch on -mmain-is-OS_task for optimizing compilations.
24086 2018-01-09 Richard Biener <rguenther@suse.de>
24088 PR tree-optimization/83572
24089 * graphite.c: Include cfganal.h.
24090 (graphite_transform_loops): Connect infinite loops to exit
24091 and remove fake edges at the end.
24093 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
24095 * ipa-inline.c (edge_badness): Revert accidental checkin.
24097 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
24100 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
24101 symbols; not inline clones.
24103 2018-01-09 Jakub Jelinek <jakub@redhat.com>
24106 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
24107 hard registers. Formatting fixes.
24109 PR preprocessor/83722
24110 * gcc.c (try_generate_repro): Pass
24111 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
24112 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
24115 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
24116 Kito Cheng <kito.cheng@gmail.com>
24118 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
24119 (riscv_leaf_function_p): Delete.
24120 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
24122 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
24124 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
24126 (do_ifelse): New function.
24127 (do_isel): New function.
24128 (do_sub3): New function.
24129 (do_add3): New function.
24130 (do_load_mask_compare): New function.
24131 (do_overlap_load_compare): New function.
24132 (expand_compare_loop): New function.
24133 (expand_block_compare): Call expand_compare_loop() when appropriate.
24134 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
24135 option description.
24136 (-mblock-compare-inline-loop-limit): New option.
24138 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
24141 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
24142 Reverse order of second and third operands in first alternative.
24143 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
24144 of first and second elements in UNSPEC_VPERMR vector.
24145 (altivec_expand_vec_perm_le): Likewise.
24147 2018-01-08 Jeff Law <law@redhat.com>
24149 PR rtl-optimizatin/81308
24150 * tree-switch-conversion.c (cfg_altered): New file scoped static.
24151 (process_switch): If group_case_labels makes a change, then set
24153 (pass_convert_switch::execute): If a switch is converted, then
24154 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
24156 PR rtl-optimization/81308
24157 * recog.c (split_all_insns): Conditionally cleanup the CFG after
24160 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
24162 PR target/83663 - Revert r255946
24163 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
24164 generation for cases where splatting a value is not useful.
24165 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
24166 across a vec_duplicate and a paradoxical subreg forming a vector
24167 mode to a vec_concat.
24169 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24171 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
24172 -march=armv8.3-a variants.
24173 * config/arm/t-multilib: Likewise.
24174 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
24176 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
24178 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
24180 (cceq_ior_compare_complement): Give it a name so I can use it, and
24181 change boolean_or_operator predicate to boolean_operator so it can
24182 be used to generate a crand.
24183 (eqne): New code iterator.
24184 (bd/bd_neg): New code_attrs.
24185 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
24186 a single define_insn.
24187 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
24188 decrement (bdnzt/bdnzf/bdzt/bdzf).
24189 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
24190 with the new names of the branch decrement patterns, and added the
24191 names of the branch decrement conditional patterns.
24193 2018-01-08 Richard Biener <rguenther@suse.de>
24195 PR tree-optimization/83563
24196 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
24199 2018-01-08 Richard Biener <rguenther@suse.de>
24201 PR middle-end/83713
24202 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
24204 2018-01-08 Richard Biener <rguenther@suse.de>
24206 PR tree-optimization/83685
24207 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
24208 references to abnormals.
24210 2018-01-08 Richard Biener <rguenther@suse.de>
24213 * dwarf2out.c (output_indirect_strings): Handle empty
24214 skeleton_debug_str_hash.
24215 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
24217 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
24219 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
24220 (emit_store_direct): Likewise.
24221 (arc_trampoline_adjust_address): Likewise.
24222 (arc_asm_trampoline_template): New function.
24223 (arc_initialize_trampoline): Use asm_trampoline_template.
24224 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
24225 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
24226 * config/arc/arc.md (flush_icache): Delete pattern.
24228 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
24230 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
24231 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
24234 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
24237 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
24238 by not USED_FOR_TARGET.
24239 (make_pass_resolve_sw_modes): Likewise.
24241 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
24243 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
24246 2018-01-08 Richard Biener <rguenther@suse.de>
24248 PR middle-end/83580
24249 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
24251 2018-01-08 Richard Biener <rguenther@suse.de>
24253 PR middle-end/83517
24254 * match.pd ((t * 2) / 2) -> t): Add missing :c.
24256 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
24258 PR middle-end/81897
24259 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
24260 basic blocks with a small number of successors.
24261 (convert_control_dep_chain_into_preds): Improve handling of
24263 (dump_predicates): Split apart into...
24264 (dump_pred_chain): ...here...
24265 (dump_pred_info): ...and here.
24266 (can_one_predicate_be_invalidated_p): Add debugging printfs.
24267 (can_chain_union_be_invalidated_p): Improve check for invalidation
24269 (uninit_uses_cannot_happen): Avoid unnecessary if
24270 convert_control_dep_chain_into_preds yielded nothing.
24272 2018-01-06 Martin Sebor <msebor@redhat.com>
24274 PR tree-optimization/83640
24275 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
24276 subtracting negative offset from size.
24277 (builtin_access::overlap): Adjust offset bounds of the access to fall
24278 within the size of the object if possible.
24280 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
24282 PR rtl-optimization/83699
24283 * expmed.c (extract_bit_field_1): Restrict the vector usage of
24284 extract_bit_field_as_subreg to cases in which the extracted
24285 value is also a vector.
24287 * lra-constraints.c (process_alt_operands): Test for the equivalence
24288 substitutions when detecting a possible reload cycle.
24290 2018-01-06 Jakub Jelinek <jakub@redhat.com>
24293 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
24294 by default if flag_selective_schedling{,2}. Formatting fixes.
24296 PR rtl-optimization/83682
24297 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
24298 if it has non-VECTOR_MODE element mode.
24299 (vec_duplicate_p): Likewise.
24301 PR middle-end/83694
24302 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
24303 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
24305 2018-01-05 Jakub Jelinek <jakub@redhat.com>
24308 * config/i386/i386-builtin.def
24309 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
24310 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
24311 Require also OPTION_MASK_ISA_AVX512F in addition to
24312 OPTION_MASK_ISA_GFNI.
24313 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
24314 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
24315 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
24316 to OPTION_MASK_ISA_GFNI.
24317 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
24318 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
24319 OPTION_MASK_ISA_AVX512BW.
24320 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
24321 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
24322 addition to OPTION_MASK_ISA_GFNI.
24323 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
24324 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
24325 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
24326 to OPTION_MASK_ISA_GFNI.
24327 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
24328 a requirement for all ISAs rather than any of them with a few
24330 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
24332 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
24333 bitmasks to be enabled with 3 exceptions, instead of requiring any
24334 enabled ISA with lots of exceptions.
24335 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
24336 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
24337 Change avx512bw in isa attribute to avx512f.
24338 * config/i386/sgxintrin.h: Add license boilerplate.
24339 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
24340 to __AVX512F__ and __AVX512VL to __AVX512VL__.
24341 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
24342 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
24344 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
24345 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
24346 temporarily sse2 rather than sse if not enabled already.
24349 * config/i386/sse.md (VI248_VLBW): Rename to ...
24350 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
24351 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
24352 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
24353 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
24354 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
24355 mode iterator instead of VI248_VLBW.
24357 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
24359 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
24360 (record_modified): Skip clobbers; add debug output.
24361 (param_change_prob): Use sreal frequencies.
24363 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
24365 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
24366 punt for user-aligned variables.
24368 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
24370 * tree-chrec.c (chrec_contains_symbols): Return true for
24373 2018-01-05 Sudakshina Das <sudi.das@arm.com>
24376 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
24377 of (x|y) == x for BICS pattern.
24379 2018-01-05 Jakub Jelinek <jakub@redhat.com>
24381 PR tree-optimization/83605
24382 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
24383 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
24386 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
24388 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
24389 * config/epiphany/rtems.h: New file.
24391 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24392 Uros Bizjak <ubizjak@gmail.com>
24395 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
24396 QIreg_operand instead of register_operand predicate.
24397 * config/i386/i386.c (ix86_rop_should_change_byte_p,
24398 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
24399 comments instead of -fmitigate[-_]rop.
24401 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
24404 * cgraphunit.c (symbol_table::compile): Switch to text_section
24405 before calling assembly_start debug hook.
24406 * run-rtl-passes.c (run_rtl_passes): Likewise.
24409 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24411 * tree-vrp.c (extract_range_from_binary_expr_1): Check
24412 range_int_cst_p rather than !symbolic_range_p before calling
24413 extract_range_from_multiplicative_op_1.
24415 2018-01-04 Jeff Law <law@redhat.com>
24417 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
24418 redundant test in assertion.
24420 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24422 * doc/rtl.texi: Document machine_mode wrapper classes.
24424 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24426 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
24427 using tree_to_uhwi.
24429 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24431 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
24432 the VEC_PERM_EXPR fold to fail.
24434 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24437 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
24438 to switched_sections.
24440 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24443 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
24444 test for d.testing.
24446 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
24449 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
24450 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
24452 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24455 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
24456 is BLKmode and bitpos not zero or mode change is needed.
24458 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24461 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
24464 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
24467 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
24468 instead of MULT rtx. Update all corresponding splitters.
24469 (*saddl_se): Ditto.
24470 (*ssub<modesuffix>): Ditto.
24471 (*ssubl_se): Ditto.
24472 (*cmp_sadd_di): Update split patterns.
24473 (*cmp_sadd_si): Ditto.
24474 (*cmp_sadd_sidi): Ditto.
24475 (*cmp_ssub_di): Ditto.
24476 (*cmp_ssub_si): Ditto.
24477 (*cmp_ssub_sidi): Ditto.
24478 * config/alpha/predicates.md (const23_operand): New predicate.
24479 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
24480 Look for ASHIFT, not MULT inner operand.
24481 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
24483 2018-01-04 Martin Liska <mliska@suse.cz>
24485 PR gcov-profile/83669
24486 * gcov.c (output_intermediate_file): Add version to intermediate
24488 * doc/gcov.texi: Document new field 'version' in intermediate
24489 file format. Fix location of '-k' option of gcov command.
24491 2018-01-04 Martin Liska <mliska@suse.cz>
24494 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
24496 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24498 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
24500 2018-01-03 Martin Sebor <msebor@redhat.com>
24502 PR tree-optimization/83655
24503 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
24504 checking calls with invalid arguments.
24506 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24508 * tree-vect-stmts.c (vect_get_store_rhs): New function.
24509 (vectorizable_mask_load_store): Delete.
24510 (vectorizable_call): Return false for masked loads and stores.
24511 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
24512 instead of gimple_assign_rhs1.
24513 (vectorizable_load): Handle IFN_MASK_LOAD.
24514 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
24516 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24518 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
24520 (vectorizable_mask_load_store): ...here.
24521 (vectorizable_load): ...and here.
24523 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24525 * tree-vect-stmts.c (vect_build_all_ones_mask)
24526 (vect_build_zero_merge_argument): New functions, split out from...
24527 (vectorizable_load): ...here.
24529 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24531 * tree-vect-stmts.c (vect_check_store_rhs): New function,
24533 (vectorizable_mask_load_store): ...here.
24534 (vectorizable_store): ...and here.
24536 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24538 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
24540 (vectorizable_mask_load_store): ...here.
24542 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24544 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
24545 (vect_model_store_cost): Take a vec_load_store_type instead of a
24547 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
24548 (vect_model_store_cost): Take a vec_load_store_type instead of a
24550 (vectorizable_mask_load_store): Update accordingly.
24551 (vectorizable_store): Likewise.
24552 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
24554 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24556 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
24557 IFN_MASK_LOAD calls here rather than...
24558 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
24560 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24561 Alan Hayward <alan.hayward@arm.com>
24562 David Sherwood <david.sherwood@arm.com>
24564 * expmed.c (extract_bit_field_1): For vector extracts,
24565 fall back to extract_bit_field_as_subreg if vec_extract
24568 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24569 Alan Hayward <alan.hayward@arm.com>
24570 David Sherwood <david.sherwood@arm.com>
24572 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
24573 they are variable or constant sized.
24574 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
24575 slots for constant-sized data.
24577 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24578 Alan Hayward <alan.hayward@arm.com>
24579 David Sherwood <david.sherwood@arm.com>
24581 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
24582 handling COND_EXPRs with boolean comparisons, try to find a better
24583 basis for the mask type than the boolean itself.
24585 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24587 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
24588 is calculated and how it can be overridden.
24589 * genmodes.c (max_bitsize_mode_any_mode): New variable.
24590 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
24592 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
24595 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24596 Alan Hayward <alan.hayward@arm.com>
24597 David Sherwood <david.sherwood@arm.com>
24599 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
24600 Remove the mode argument.
24601 (aarch64_simd_valid_immediate): Remove the mode and inverse
24603 * config/aarch64/iterators.md (bitsize): New iterator.
24604 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
24605 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
24606 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
24607 aarch64_simd_valid_immediate.
24608 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
24609 (aarch64_reg_or_bic_imm): Likewise.
24610 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
24611 with an insn_type enum and msl with a modifier_type enum.
24612 Replace element_width with a scalar_mode. Change the shift
24613 to unsigned int. Add constructors for scalar_float_mode and
24614 scalar_int_mode elements.
24615 (aarch64_vect_float_const_representable_p): Delete.
24616 (aarch64_can_const_movi_rtx_p)
24617 (aarch64_simd_scalar_immediate_valid_for_move)
24618 (aarch64_simd_make_constant): Update call to
24619 aarch64_simd_valid_immediate.
24620 (aarch64_advsimd_valid_immediate_hs): New function.
24621 (aarch64_advsimd_valid_immediate): Likewise.
24622 (aarch64_simd_valid_immediate): Remove mode and inverse
24623 arguments. Rewrite to use the above. Use const_vec_duplicate_p
24624 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
24625 and aarch64_float_const_representable_p on the result.
24626 (aarch64_output_simd_mov_immediate): Remove mode argument.
24627 Update call to aarch64_simd_valid_immediate and use of
24628 simd_immediate_info.
24629 (aarch64_output_scalar_simd_mov_immediate): Update call
24632 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24633 Alan Hayward <alan.hayward@arm.com>
24634 David Sherwood <david.sherwood@arm.com>
24636 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
24637 (mode_nunits): Likewise CONST_MODE_NUNITS.
24638 * machmode.def (ADJUST_NUNITS): Document.
24639 * genmodes.c (mode_data::need_nunits_adj): New field.
24640 (blank_mode): Update accordingly.
24641 (adj_nunits): New variable.
24642 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
24644 (emit_mode_size_inline): Set need_bytesize_adj for all modes
24645 listed in adj_nunits.
24646 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
24647 listed in adj_nunits. Don't emit case statements for such modes.
24648 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
24649 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
24650 nothing if adj_nunits is nonnull.
24651 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
24652 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
24653 (emit_mode_fbit): Update use of print_maybe_const_decl.
24654 (emit_move_size): Likewise. Treat the array as non-const
24656 (emit_mode_adjustments): Handle adj_nunits.
24658 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24660 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
24661 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
24662 (VECTOR_MODES): Use it.
24663 (make_vector_modes): Take the prefix as an argument.
24665 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24666 Alan Hayward <alan.hayward@arm.com>
24667 David Sherwood <david.sherwood@arm.com>
24669 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
24670 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
24671 for MODE_VECTOR_BOOL.
24672 * machmode.def (VECTOR_BOOL_MODE): Document.
24673 * genmodes.c (VECTOR_BOOL_MODE): New macro.
24674 (make_vector_bool_mode): New function.
24675 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
24677 * lto-streamer-in.c (lto_input_mode_table): Likewise.
24678 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
24680 * stor-layout.c (int_mode_for_mode): Likewise.
24681 * tree.c (build_vector_type_for_mode): Likewise.
24682 * varasm.c (output_constant_pool_2): Likewise.
24683 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
24684 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
24685 for MODE_VECTOR_BOOL.
24686 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
24687 of mode class checks.
24688 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
24689 instead of a list of mode class checks.
24690 (expand_vector_scalar_condition): Likewise.
24691 (type_for_widest_vector_mode): Handle BImode as an inner mode.
24693 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24694 Alan Hayward <alan.hayward@arm.com>
24695 David Sherwood <david.sherwood@arm.com>
24697 * machmode.h (mode_size): Change from unsigned short to
24699 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
24700 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
24701 or if measurement_type is not polynomial.
24702 (fixed_size_mode::includes_p): Check for constant-sized modes.
24703 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
24704 return a poly_uint16 rather than an unsigned short.
24705 (emit_mode_size): Change the type of mode_size from unsigned short
24706 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
24707 (emit_mode_adjustments): Cope with polynomial vector sizes.
24708 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24710 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24712 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
24713 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
24714 * caller-save.c (setup_save_areas): Likewise.
24715 (replace_reg_with_saved_mem): Likewise.
24716 * calls.c (emit_library_call_value_1): Likewise.
24717 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
24718 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
24719 (gen_lowpart_for_combine): Likewise.
24720 * convert.c (convert_to_integer_1): Likewise.
24721 * cse.c (equiv_constant, cse_insn): Likewise.
24722 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
24723 (cselib_subst_to_values): Likewise.
24724 * dce.c (word_dce_process_block): Likewise.
24725 * df-problems.c (df_word_lr_mark_ref): Likewise.
24726 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
24727 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
24728 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
24729 (rtl_for_decl_location): Likewise.
24730 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
24731 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
24732 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
24733 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
24734 (expand_expr_real_1): Likewise.
24735 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
24736 (pad_below): Likewise.
24737 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
24738 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
24739 * ira.c (get_subreg_tracking_sizes): Likewise.
24740 * ira-build.c (ira_create_allocno_objects): Likewise.
24741 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
24742 (ira_sort_regnos_for_alter_reg): Likewise.
24743 * ira-costs.c (record_operand_costs): Likewise.
24744 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
24745 (resolve_simple_move): Likewise.
24746 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
24747 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
24748 (lra_constraints): Likewise.
24749 (CONST_POOL_OK_P): Reject variable-sized modes.
24750 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
24751 (add_pseudo_to_slot, lra_spill): Likewise.
24752 * omp-low.c (omp_clause_aligned_alignment): Likewise.
24753 * optabs-query.c (get_best_extraction_insn): Likewise.
24754 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
24755 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
24756 (expand_mult_highpart, valid_multiword_target_p): Likewise.
24757 * recog.c (offsettable_address_addr_space_p): Likewise.
24758 * regcprop.c (maybe_mode_change): Likewise.
24759 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
24760 * regrename.c (build_def_use): Likewise.
24761 * regstat.c (dump_reg_info): Likewise.
24762 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
24763 (find_reloads, find_reloads_subreg_address): Likewise.
24764 * reload1.c (eliminate_regs_1): Likewise.
24765 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
24766 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
24767 (simplify_binary_operation_1, simplify_subreg): Likewise.
24768 * targhooks.c (default_function_arg_padding): Likewise.
24769 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
24770 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
24771 (verify_gimple_assign_ternary): Likewise.
24772 * tree-inline.c (estimate_move_cost): Likewise.
24773 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24774 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
24775 (get_address_cost_ainc): Likewise.
24776 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
24777 (vect_supportable_dr_alignment): Likewise.
24778 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
24779 (vectorizable_reduction): Likewise.
24780 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
24781 (vectorizable_operation, vectorizable_load): Likewise.
24782 * tree.c (build_same_sized_truth_vector_type): Likewise.
24783 * valtrack.c (cleanup_auto_inc_dec): Likewise.
24784 * var-tracking.c (emit_note_insn_var_location): Likewise.
24785 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
24786 (ADDR_VEC_ALIGN): Likewise.
24788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24789 Alan Hayward <alan.hayward@arm.com>
24790 David Sherwood <david.sherwood@arm.com>
24792 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
24794 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
24795 or if measurement_type is polynomial.
24796 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
24797 * combine.c (make_extraction): Likewise.
24798 * dse.c (find_shift_sequence): Likewise.
24799 * dwarf2out.c (mem_loc_descriptor): Likewise.
24800 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
24801 (extract_bit_field, extract_low_bits): Likewise.
24802 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
24803 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
24804 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
24805 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
24806 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
24807 * reload.c (find_reloads): Likewise.
24808 * reload1.c (alter_reg): Likewise.
24809 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
24810 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
24811 * tree-if-conv.c (predicate_mem_writes): Likewise.
24812 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
24813 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
24814 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
24815 * valtrack.c (dead_debug_insert_temp): Likewise.
24816 * varasm.c (mergeable_constant_section): Likewise.
24817 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
24819 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24820 Alan Hayward <alan.hayward@arm.com>
24821 David Sherwood <david.sherwood@arm.com>
24823 * expr.c (expand_assignment): Cope with polynomial mode sizes
24824 when assigning to a CONCAT.
24826 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24827 Alan Hayward <alan.hayward@arm.com>
24828 David Sherwood <david.sherwood@arm.com>
24830 * machmode.h (mode_precision): Change from unsigned short to
24832 (mode_to_precision): Return a poly_uint16 rather than an unsigned
24834 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
24835 or if measurement_type is not polynomial.
24836 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
24837 in which the mode is already known to be a scalar_int_mode.
24838 * genmodes.c (emit_mode_precision): Change the type of mode_precision
24839 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
24841 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24842 for GET_MODE_PRECISION.
24843 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24844 for GET_MODE_PRECISION.
24845 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
24847 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
24848 (expand_field_assignment, make_extraction): Likewise.
24849 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
24850 (get_last_value): Likewise.
24851 * convert.c (convert_to_integer_1): Likewise.
24852 * cse.c (cse_insn): Likewise.
24853 * expr.c (expand_expr_real_1): Likewise.
24854 * lra-constraints.c (simplify_operand_subreg): Likewise.
24855 * optabs-query.c (can_atomic_load_p): Likewise.
24856 * optabs.c (expand_atomic_load): Likewise.
24857 (expand_atomic_store): Likewise.
24858 * ree.c (combine_reaching_defs): Likewise.
24859 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
24860 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
24861 * tree.h (type_has_mode_precision_p): Likewise.
24862 * ubsan.c (instrument_si_overflow): Likewise.
24864 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24865 Alan Hayward <alan.hayward@arm.com>
24866 David Sherwood <david.sherwood@arm.com>
24868 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
24869 polynomial numbers of units.
24870 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
24871 (valid_vector_subparts_p): New function.
24872 (build_vector_type): Remove temporary shim and take the number
24873 of units as a poly_uint64 rather than an int.
24874 (build_opaque_vector_type): Take the number of units as a
24875 poly_uint64 rather than an int.
24876 * tree.c (build_vector_from_ctor): Handle polynomial
24877 TYPE_VECTOR_SUBPARTS.
24878 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
24879 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
24880 (build_vector_from_val): If the number of units is variable,
24881 use build_vec_duplicate_cst for constant operands and
24882 VEC_DUPLICATE_EXPR otherwise.
24883 (make_vector_type): Remove temporary is_constant ().
24884 (build_vector_type, build_opaque_vector_type): Take the number of
24885 units as a poly_uint64 rather than an int.
24886 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
24888 * cfgexpand.c (expand_debug_expr): Likewise.
24889 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
24890 (store_constructor, expand_expr_real_1): Likewise.
24891 (const_scalar_mask_from_tree): Likewise.
24892 * fold-const-call.c (fold_const_reduction): Likewise.
24893 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
24894 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
24895 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
24896 (fold_relational_const): Likewise.
24897 (native_interpret_vector): Likewise. Change the size from an
24898 int to an unsigned int.
24899 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
24900 TYPE_VECTOR_SUBPARTS.
24901 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
24902 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
24903 duplicating a non-constant operand into a variable-length vector.
24904 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
24905 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
24906 * ipa-icf.c (sem_variable::equals): Likewise.
24907 * match.pd: Likewise.
24908 * omp-simd-clone.c (simd_clone_subparts): Likewise.
24909 * print-tree.c (print_node): Likewise.
24910 * stor-layout.c (layout_type): Likewise.
24911 * targhooks.c (default_builtin_vectorization_cost): Likewise.
24912 * tree-cfg.c (verify_gimple_comparison): Likewise.
24913 (verify_gimple_assign_binary): Likewise.
24914 (verify_gimple_assign_ternary): Likewise.
24915 (verify_gimple_assign_single): Likewise.
24916 * tree-pretty-print.c (dump_generic_node): Likewise.
24917 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24918 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
24919 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
24920 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
24921 (vect_shift_permute_load_chain): Likewise.
24922 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
24923 (expand_vector_condition, optimize_vector_constructor): Likewise.
24924 (lower_vec_perm, get_compute_type): Likewise.
24925 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
24926 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
24927 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
24928 (vect_recog_mask_conversion_pattern): Likewise.
24929 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
24930 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
24931 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
24932 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
24933 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
24934 (vectorizable_shift, vectorizable_operation, vectorizable_store)
24935 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
24936 (supportable_widening_operation): Likewise.
24937 (supportable_narrowing_operation): Likewise.
24938 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
24940 * varasm.c (output_constant): Likewise.
24942 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24943 Alan Hayward <alan.hayward@arm.com>
24944 David Sherwood <david.sherwood@arm.com>
24946 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
24947 so that both the length == 3 and length != 3 cases set up their
24948 own permute vectors. Add comments explaining why we know the
24949 number of elements is constant.
24950 (vect_permute_load_chain): Likewise.
24952 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24953 Alan Hayward <alan.hayward@arm.com>
24954 David Sherwood <david.sherwood@arm.com>
24956 * machmode.h (mode_nunits): Change from unsigned char to
24958 (ONLY_FIXED_SIZE_MODES): New macro.
24959 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
24960 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
24961 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
24963 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
24964 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
24965 or if measurement_type is not polynomial.
24966 * genmodes.c (ZERO_COEFFS): New macro.
24967 (emit_mode_nunits_inline): Make mode_nunits_inline return a
24969 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
24970 Use ZERO_COEFFS when emitting initializers.
24971 * data-streamer.h (bp_pack_poly_value): New function.
24972 (bp_unpack_poly_value): Likewise.
24973 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24974 for GET_MODE_NUNITS.
24975 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24976 for GET_MODE_NUNITS.
24977 * tree.c (make_vector_type): Remove temporary shim and make
24978 the real function take the number of units as a poly_uint64
24979 rather than an int.
24980 (build_vector_type_for_mode): Handle polynomial nunits.
24981 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
24982 * emit-rtl.c (const_vec_series_p_1): Likewise.
24983 (gen_rtx_CONST_VECTOR): Likewise.
24984 * fold-const.c (test_vec_duplicate_folding): Likewise.
24985 * genrecog.c (validate_pattern): Likewise.
24986 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
24987 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
24988 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
24989 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
24990 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
24991 * rtlanal.c (subreg_get_info): Likewise.
24992 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
24993 (vect_grouped_load_supported): Likewise.
24994 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
24995 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
24996 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
24997 (simplify_const_unary_operation, simplify_binary_operation_1)
24998 (simplify_const_binary_operation, simplify_ternary_operation)
24999 (test_vector_ops_duplicate, test_vector_ops): Likewise.
25000 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
25001 instead of CONST_VECTOR_NUNITS.
25002 * varasm.c (output_constant_pool_2): Likewise.
25003 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
25004 explicit-encoded elements in the XVEC for variable-length vectors.
25006 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25008 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
25010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25011 Alan Hayward <alan.hayward@arm.com>
25012 David Sherwood <david.sherwood@arm.com>
25014 * coretypes.h (fixed_size_mode): Declare.
25015 (fixed_size_mode_pod): New typedef.
25016 * builtins.h (target_builtins::x_apply_args_mode)
25017 (target_builtins::x_apply_result_mode): Change type to
25018 fixed_size_mode_pod.
25019 * builtins.c (apply_args_size, apply_result_size, result_vector)
25020 (expand_builtin_apply_args_1, expand_builtin_apply)
25021 (expand_builtin_return): Update accordingly.
25023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25025 * cse.c (hash_rtx_cb): Hash only the encoded elements.
25026 * cselib.c (cselib_hash_rtx): Likewise.
25027 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
25028 CONST_VECTOR encoding.
25030 2018-01-03 Jakub Jelinek <jakub@redhat.com>
25031 Jeff Law <law@redhat.com>
25034 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
25035 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
25036 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
25037 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
25040 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
25041 explicitly probe *sp in a noreturn function if there were any callee
25042 register saves or frame pointer is needed.
25044 2018-01-03 Jakub Jelinek <jakub@redhat.com>
25047 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
25048 BLKmode for ternary, binary or unary expressions.
25051 * var-tracking.c (delete_vta_debug_insn): New inline function.
25052 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
25053 insns from get_insns () to NULL instead of each bb separately.
25054 Use delete_vta_debug_insn. No longer static.
25055 (vt_debug_insns_local, variable_tracking_main_1): Adjust
25056 delete_vta_debug_insns callers.
25057 * rtl.h (delete_vta_debug_insns): Declare.
25058 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
25059 instead of variable_tracking_main.
25061 2018-01-03 Martin Sebor <msebor@redhat.com>
25063 PR tree-optimization/83603
25064 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
25065 arguments past the endof the argument list in functions declared
25066 without a prototype.
25067 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
25068 Avoid checking when arguments are null.
25070 2018-01-03 Martin Sebor <msebor@redhat.com>
25073 * doc/extend.texi (attribute const): Fix a typo.
25074 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
25075 issuing -Wsuggest-attribute for void functions.
25077 2018-01-03 Martin Sebor <msebor@redhat.com>
25079 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
25080 offset_int::from instead of wide_int::to_shwi.
25081 (maybe_diag_overlap): Remove assertion.
25082 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
25083 * gimple-ssa-sprintf.c (format_directive): Same.
25084 (parse_directive): Same.
25085 (sprintf_dom_walker::compute_format_length): Same.
25086 (try_substitute_return_value): Same.
25088 2018-01-03 Jeff Law <law@redhat.com>
25090 PR middle-end/83654
25091 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
25092 non-constant residual for zero at runtime and avoid probing in
25093 that case. Reorganize code for trailing problem to mirror handling
25096 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
25098 PR tree-optimization/83501
25099 * tree-ssa-strlen.c (get_string_cst): New.
25100 (handle_char_store): Call get_string_cst.
25102 2018-01-03 Martin Liska <mliska@suse.cz>
25104 PR tree-optimization/83593
25105 * tree-ssa-strlen.c: Include tree-cfg.h.
25106 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
25107 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
25108 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
25110 (strlen_dom_walker::before_dom_children): Call
25111 gimple_purge_dead_eh_edges. Dump tranformation with details
25113 (strlen_dom_walker::before_dom_children): Update call by adding
25114 new argument cleanup_eh.
25115 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
25117 2018-01-03 Martin Liska <mliska@suse.cz>
25120 * cif-code.def (VARIADIC_THUNK): New enum value.
25121 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
25124 2018-01-03 Jan Beulich <jbeulich@suse.com>
25126 * sse.md (mov<mode>_internal): Tighten condition for when to use
25127 vmovdqu<ssescalarsize> for TI and OI modes.
25129 2018-01-03 Jakub Jelinek <jakub@redhat.com>
25131 Update copyright years.
25133 2018-01-03 Martin Liska <mliska@suse.cz>
25136 * ipa-visibility.c (function_and_variable_visibility): Skip
25137 functions with noipa attribure.
25139 2018-01-03 Jakub Jelinek <jakub@redhat.com>
25141 * gcc.c (process_command): Update copyright notice dates.
25142 * gcov-dump.c (print_version): Ditto.
25143 * gcov.c (print_version): Ditto.
25144 * gcov-tool.c (print_version): Ditto.
25145 * gengtype.c (create_file): Ditto.
25146 * doc/cpp.texi: Bump @copying's copyright year.
25147 * doc/cppinternals.texi: Ditto.
25148 * doc/gcc.texi: Ditto.
25149 * doc/gccint.texi: Ditto.
25150 * doc/gcov.texi: Ditto.
25151 * doc/install.texi: Ditto.
25152 * doc/invoke.texi: Ditto.
25154 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25156 * vector-builder.h (vector_builder::m_full_nelts): Change from
25157 unsigned int to poly_uint64.
25158 (vector_builder::full_nelts): Update prototype accordingly.
25159 (vector_builder::new_vector): Likewise.
25160 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
25161 (vector_builder::operator ==): Likewise.
25162 (vector_builder::finalize): Likewise.
25163 * int-vector-builder.h (int_vector_builder::int_vector_builder):
25164 Take the number of elements as a poly_uint64 rather than an
25166 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
25167 from unsigned int to poly_uint64.
25168 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
25169 (vec_perm_indices::new_vector): Likewise.
25170 (vec_perm_indices::length): Likewise.
25171 (vec_perm_indices::nelts_per_input): Likewise.
25172 (vec_perm_indices::input_nelts): Likewise.
25173 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
25174 number of elements per input as a poly_uint64 rather than an
25175 unsigned int. Use the original encoding for variable-length
25176 vectors, rather than clamping each individual element.
25177 For the second and subsequent elements in each pattern,
25178 clamp the step and base before clamping their sum.
25179 (vec_perm_indices::series_p): Handle polynomial element counts.
25180 (vec_perm_indices::all_in_range_p): Likewise.
25181 (vec_perm_indices_to_tree): Likewise.
25182 (vec_perm_indices_to_rtx): Likewise.
25183 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
25184 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
25185 (tree_vector_builder::new_binary_operation): Handle polynomial
25186 element counts. Return false if we need to know the number
25187 of elements at compile time.
25188 * fold-const.c (fold_vec_perm): Punt if the number of elements
25189 isn't known at compile time.
25191 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25193 * vec-perm-indices.h (vec_perm_builder): Change element type
25194 from HOST_WIDE_INT to poly_int64.
25195 (vec_perm_indices::element_type): Update accordingly.
25196 (vec_perm_indices::clamp): Handle polynomial element_types.
25197 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
25198 (vec_perm_indices::all_in_range_p): Likewise.
25199 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
25201 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
25202 polynomial vec_perm_indices element types.
25203 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
25204 * fold-const.c (fold_vec_perm): Likewise.
25205 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
25206 * tree-vect-generic.c (lower_vec_perm): Likewise.
25207 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
25208 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
25209 element type to HOST_WIDE_INT.
25211 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25212 Alan Hayward <alan.hayward@arm.com>
25213 David Sherwood <david.sherwood@arm.com>
25215 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
25216 rather than an int. Use plus_constant.
25217 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
25218 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
25220 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25221 Alan Hayward <alan.hayward@arm.com>
25222 David Sherwood <david.sherwood@arm.com>
25224 * calls.c (emit_call_1, expand_call): Change struct_value_size from
25225 a HOST_WIDE_INT to a poly_int64.
25227 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25228 Alan Hayward <alan.hayward@arm.com>
25229 David Sherwood <david.sherwood@arm.com>
25231 * calls.c (load_register_parameters): Cope with polynomial
25232 mode sizes. Require a constant size for BLKmode parameters
25233 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
25234 forces a parameter to be padded at the lsb end in order to
25235 fill a complete number of words, require the parameter size
25236 to be ordered wrt UNITS_PER_WORD.
25238 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25239 Alan Hayward <alan.hayward@arm.com>
25240 David Sherwood <david.sherwood@arm.com>
25242 * reload1.c (spill_stack_slot_width): Change element type
25243 from unsigned int to poly_uint64_pod.
25244 (alter_reg): Treat mode sizes as polynomial.
25246 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25247 Alan Hayward <alan.hayward@arm.com>
25248 David Sherwood <david.sherwood@arm.com>
25250 * reload.c (complex_word_subreg_p): New function.
25251 (reload_inner_reg_of_subreg, push_reload): Use it.
25253 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25254 Alan Hayward <alan.hayward@arm.com>
25255 David Sherwood <david.sherwood@arm.com>
25257 * lra-constraints.c (process_alt_operands): Reject matched
25258 operands whose sizes aren't ordered.
25259 (match_reload): Refer to this check here.
25261 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25262 Alan Hayward <alan.hayward@arm.com>
25263 David Sherwood <david.sherwood@arm.com>
25265 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
25266 that the mode size is in the set {1, 2, 4, 8, 16}.
25268 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25269 Alan Hayward <alan.hayward@arm.com>
25270 David Sherwood <david.sherwood@arm.com>
25272 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
25273 Use plus_constant instead of gen_rtx_PLUS.
25275 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25276 Alan Hayward <alan.hayward@arm.com>
25277 David Sherwood <david.sherwood@arm.com>
25279 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
25280 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
25281 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
25282 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
25283 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
25284 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
25285 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
25286 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
25287 * config/i386/i386.c (ix86_push_rounding): ...this new function.
25288 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
25290 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
25291 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
25292 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
25293 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
25294 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
25295 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
25296 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
25297 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
25298 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
25299 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
25301 * expr.c (emit_move_resolve_push): Treat the input and result
25302 of PUSH_ROUNDING as a poly_int64.
25303 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
25304 (emit_push_insn): Likewise.
25305 * lra-eliminations.c (mark_not_eliminable): Likewise.
25306 * recog.c (push_operand): Likewise.
25307 * reload1.c (elimination_effects): Likewise.
25308 * rtlanal.c (nonzero_bits1): Likewise.
25309 * calls.c (store_one_arg): Likewise. Require the padding to be
25310 known at compile time.
25312 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25313 Alan Hayward <alan.hayward@arm.com>
25314 David Sherwood <david.sherwood@arm.com>
25316 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
25317 Use plus_constant instead of gen_rtx_PLUS.
25319 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25320 Alan Hayward <alan.hayward@arm.com>
25321 David Sherwood <david.sherwood@arm.com>
25323 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
25324 rather than an int.
25326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25327 Alan Hayward <alan.hayward@arm.com>
25328 David Sherwood <david.sherwood@arm.com>
25330 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
25331 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
25332 via stack temporaries. Treat the mode size as polynomial too.
25334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25335 Alan Hayward <alan.hayward@arm.com>
25336 David Sherwood <david.sherwood@arm.com>
25338 * expr.c (expand_expr_real_2): When handling conversions involving
25339 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
25340 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
25341 as a poly_uint64 too.
25343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25344 Alan Hayward <alan.hayward@arm.com>
25345 David Sherwood <david.sherwood@arm.com>
25347 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
25349 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25350 Alan Hayward <alan.hayward@arm.com>
25351 David Sherwood <david.sherwood@arm.com>
25353 * combine.c (can_change_dest_mode): Handle polynomial
25354 REGMODE_NATURAL_SIZE.
25355 * expmed.c (store_bit_field_1): Likewise.
25356 * expr.c (store_constructor): Likewise.
25357 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
25358 and polynomial REGMODE_NATURAL_SIZE.
25359 (gen_lowpart_common): Likewise.
25360 * reginfo.c (record_subregs_of_mode): Likewise.
25361 * rtlanal.c (read_modify_subreg_p): Likewise.
25363 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25364 Alan Hayward <alan.hayward@arm.com>
25365 David Sherwood <david.sherwood@arm.com>
25367 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
25368 numbers of elements.
25370 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25371 Alan Hayward <alan.hayward@arm.com>
25372 David Sherwood <david.sherwood@arm.com>
25374 * match.pd: Cope with polynomial numbers of vector elements.
25376 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25377 Alan Hayward <alan.hayward@arm.com>
25378 David Sherwood <david.sherwood@arm.com>
25380 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
25381 in a POINTER_PLUS_EXPR.
25383 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25384 Alan Hayward <alan.hayward@arm.com>
25385 David Sherwood <david.sherwood@arm.com>
25387 * omp-simd-clone.c (simd_clone_subparts): New function.
25388 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
25389 (ipa_simd_modify_function_body): Likewise.
25391 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25392 Alan Hayward <alan.hayward@arm.com>
25393 David Sherwood <david.sherwood@arm.com>
25395 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
25396 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
25397 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
25398 (expand_vector_condition, vector_element): Likewise.
25399 (subparts_gt): New function.
25400 (get_compute_type): Use subparts_gt.
25401 (count_type_subparts): Delete.
25402 (expand_vector_operations_1): Use subparts_gt instead of
25403 count_type_subparts.
25405 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25406 Alan Hayward <alan.hayward@arm.com>
25407 David Sherwood <david.sherwood@arm.com>
25409 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
25410 (vect_compile_time_alias): ...this new function. Do the calculation
25411 on poly_ints rather than trees.
25412 (vect_prune_runtime_alias_test_list): Update call accordingly.
25414 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25415 Alan Hayward <alan.hayward@arm.com>
25416 David Sherwood <david.sherwood@arm.com>
25418 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
25420 (vect_schedule_slp_instance): Likewise.
25422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25423 Alan Hayward <alan.hayward@arm.com>
25424 David Sherwood <david.sherwood@arm.com>
25426 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
25427 constant and extern definitions for variable-length vectors.
25428 (vect_get_constant_vectors): Note that the number of units
25429 is known to be constant.
25431 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25432 Alan Hayward <alan.hayward@arm.com>
25433 David Sherwood <david.sherwood@arm.com>
25435 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
25436 of units as polynomial. Choose between WIDE and NARROW based
25439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25440 Alan Hayward <alan.hayward@arm.com>
25441 David Sherwood <david.sherwood@arm.com>
25443 * tree-vect-stmts.c (simd_clone_subparts): New function.
25444 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
25446 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25447 Alan Hayward <alan.hayward@arm.com>
25448 David Sherwood <david.sherwood@arm.com>
25450 * tree-vect-stmts.c (vectorizable_call): Treat the number of
25451 vectors as polynomial. Use build_index_vector for
25452 IFN_GOMP_SIMD_LANE.
25454 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25455 Alan Hayward <alan.hayward@arm.com>
25456 David Sherwood <david.sherwood@arm.com>
25458 * tree-vect-stmts.c (get_load_store_type): Treat the number of
25459 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
25460 for variable-length vectors.
25461 (vectorizable_mask_load_store): Treat the number of units as
25462 polynomial, asserting that it is constant if the condition has
25463 already been enforced.
25464 (vectorizable_store, vectorizable_load): Likewise.
25466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25467 Alan Hayward <alan.hayward@arm.com>
25468 David Sherwood <david.sherwood@arm.com>
25470 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
25471 of units as polynomial. Punt if we can't tell at compile time
25472 which vector contains the final result.
25474 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25475 Alan Hayward <alan.hayward@arm.com>
25476 David Sherwood <david.sherwood@arm.com>
25478 * tree-vect-loop.c (vectorizable_induction): Treat the number
25479 of units as polynomial. Punt on SLP inductions. Use an integer
25480 VEC_SERIES_EXPR for variable-length integer reductions. Use a
25481 cast of such a series for variable-length floating-point
25484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25485 Alan Hayward <alan.hayward@arm.com>
25486 David Sherwood <david.sherwood@arm.com>
25488 * tree.h (build_index_vector): Declare.
25489 * tree.c (build_index_vector): New function.
25490 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
25491 of units as polynomial, forcibly converting it to a constant if
25492 vectorizable_reduction has already enforced the condition.
25493 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
25494 to create a {1,2,3,...} vector.
25495 (vectorizable_reduction): Treat the number of units as polynomial.
25496 Choose vectype_in based on the largest scalar element size rather
25497 than the smallest number of units. Enforce the restrictions
25500 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25501 Alan Hayward <alan.hayward@arm.com>
25502 David Sherwood <david.sherwood@arm.com>
25504 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
25505 number of units as polynomial.
25507 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25508 Alan Hayward <alan.hayward@arm.com>
25509 David Sherwood <david.sherwood@arm.com>
25511 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
25512 * target.def (autovectorize_vector_sizes): Return the vector sizes
25513 by pointer, using vector_sizes rather than a bitmask.
25514 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
25515 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
25516 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
25518 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
25519 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
25520 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
25521 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
25522 * omp-general.c (omp_max_vf): Likewise.
25523 * omp-low.c (omp_clause_aligned_alignment): Likewise.
25524 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
25525 * tree-vect-loop.c (vect_analyze_loop): Likewise.
25526 * tree-vect-slp.c (vect_slp_bb): Likewise.
25527 * doc/tm.texi: Regenerate.
25528 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
25530 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
25531 the vector size as a poly_uint64 rather than an unsigned int.
25532 (current_vector_size): Change from an unsigned int to a poly_uint64.
25533 (get_vectype_for_scalar_type): Update accordingly.
25534 * tree.h (build_truth_vector_type): Take the size and number of
25535 units as a poly_uint64 rather than an unsigned int.
25536 (build_vector_type): Add a temporary overload that takes
25537 the number of units as a poly_uint64 rather than an unsigned int.
25538 * tree.c (make_vector_type): Likewise.
25539 (build_truth_vector_type): Take the number of units as a poly_uint64
25540 rather than an unsigned int.
25542 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25543 Alan Hayward <alan.hayward@arm.com>
25544 David Sherwood <david.sherwood@arm.com>
25546 * target.def (get_mask_mode): Take the number of units and length
25547 as poly_uint64s rather than unsigned ints.
25548 * targhooks.h (default_get_mask_mode): Update accordingly.
25549 * targhooks.c (default_get_mask_mode): Likewise.
25550 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
25551 * doc/tm.texi: Regenerate.
25553 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25554 Alan Hayward <alan.hayward@arm.com>
25555 David Sherwood <david.sherwood@arm.com>
25557 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
25558 * omp-general.c (omp_max_vf): Likewise.
25559 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
25560 (expand_omp_simd): Handle polynomial safelen.
25561 * omp-low.c (omplow_simd_context): Add a default constructor.
25562 (omplow_simd_context::max_vf): Change from int to poly_uint64.
25563 (lower_rec_simd_input_clauses): Update accordingly.
25564 (lower_rec_input_clauses): Likewise.
25566 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25567 Alan Hayward <alan.hayward@arm.com>
25568 David Sherwood <david.sherwood@arm.com>
25570 * tree-vectorizer.h (vect_nunits_for_cost): New function.
25571 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
25572 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
25573 (vect_analyze_slp_cost): Likewise.
25574 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
25575 (vect_model_load_cost): Likewise.
25577 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25578 Alan Hayward <alan.hayward@arm.com>
25579 David Sherwood <david.sherwood@arm.com>
25581 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
25582 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
25583 from an unsigned int * to a poly_uint64_pod *.
25584 (calculate_unrolling_factor): New function.
25585 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
25587 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25588 Alan Hayward <alan.hayward@arm.com>
25589 David Sherwood <david.sherwood@arm.com>
25591 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
25592 from an unsigned int to a poly_uint64.
25593 (_loop_vec_info::slp_unrolling_factor): Likewise.
25594 (_loop_vec_info::vectorization_factor): Change from an int
25596 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
25597 (vect_get_num_vectors): New function.
25598 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
25599 (vect_get_num_copies): Use vect_get_num_vectors.
25600 (vect_analyze_data_ref_dependences): Change max_vf from an int *
25601 to an unsigned int *.
25602 (vect_analyze_data_refs): Change min_vf from an int * to a
25604 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
25605 than an unsigned HOST_WIDE_INT.
25606 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
25607 (vect_analyze_data_ref_dependence): Change max_vf from an int *
25608 to an unsigned int *.
25609 (vect_analyze_data_ref_dependences): Likewise.
25610 (vect_compute_data_ref_alignment): Handle polynomial vf.
25611 (vect_enhance_data_refs_alignment): Likewise.
25612 (vect_prune_runtime_alias_test_list): Likewise.
25613 (vect_shift_permute_load_chain): Likewise.
25614 (vect_supportable_dr_alignment): Likewise.
25615 (dependence_distance_ge_vf): Take the vectorization factor as a
25616 poly_uint64 rather than an unsigned HOST_WIDE_INT.
25617 (vect_analyze_data_refs): Change min_vf from an int * to a
25619 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
25620 vfm1 as a poly_uint64 rather than an int. Make the same change
25621 for the returned bound_scalar.
25622 (vect_gen_vector_loop_niters): Handle polynomial vf.
25623 (vect_do_peeling): Likewise. Update call to
25624 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
25625 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
25627 * tree-vect-loop.c (vect_determine_vectorization_factor)
25628 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
25629 (vect_get_known_peeling_cost): Likewise.
25630 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
25631 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
25632 (vect_transform_loop): Likewise. Use the lowest possible VF when
25633 updating the upper bounds of the loop.
25634 (vect_min_worthwhile_factor): Make static. Return an unsigned int
25635 rather than an int.
25636 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
25637 polynomial unroll factors.
25638 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
25639 (vect_make_slp_decision): Likewise.
25640 (vect_supported_load_permutation_p): Likewise, and polynomial
25642 (vect_analyze_slp_cost): Handle polynomial vf.
25643 (vect_slp_analyze_node_operations): Likewise.
25644 (vect_slp_analyze_bb_1): Likewise.
25645 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
25646 than an unsigned HOST_WIDE_INT.
25647 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
25648 (vectorizable_load): Handle polynomial vf.
25649 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
25651 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
25653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25654 Alan Hayward <alan.hayward@arm.com>
25655 David Sherwood <david.sherwood@arm.com>
25657 * match.pd: Handle bit operations involving three constants
25658 and try to fold one pair.
25660 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25662 * tree-vect-loop-manip.c: Include gimple-fold.h.
25663 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
25664 niters_maybe_zero parameters. Handle other cases besides a step of 1.
25665 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
25666 Add a path that uses a step of VF instead of 1, but disable it
25668 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
25669 and niters_no_overflow parameters. Update calls to
25670 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
25671 Create a new SSA name if the latter choses to use a ste other
25672 than zero, and return it via niters_vector_mult_vf_var.
25673 * tree-vect-loop.c (vect_transform_loop): Update calls to
25674 vect_do_peeling, vect_gen_vector_loop_niters and
25675 slpeel_make_loop_iterate_ntimes.
25676 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
25677 (vect_gen_vector_loop_niters): Update declarations after above changes.
25679 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
25681 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
25682 128-bit round to integer instructions.
25683 (ceil<mode>2): Likewise.
25684 (btrunc<mode>2): Likewise.
25685 (round<mode>2): Likewise.
25687 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
25689 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
25690 unaligned VSX load/store on P8/P9.
25691 (expand_block_clear): Allow the use of unaligned VSX
25692 load/store on P8/P9.
25694 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
25696 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
25698 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
25699 swap associated with both a load and a store.
25701 2018-01-02 Andrew Waterman <andrew@sifive.com>
25703 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
25704 * config/riscv/riscv.md (clear_cache): Use it.
25706 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
25708 * web.c: Remove out-of-date comment.
25710 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25712 * expr.c (fixup_args_size_notes): Check that any existing
25713 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
25714 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
25715 (emit_single_push_insn): ...here.
25717 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25719 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
25720 (const_vector_encoded_nelts): New function.
25721 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
25722 (const_vector_int_elt, const_vector_elt): Declare.
25723 * emit-rtl.c (const_vector_int_elt_1): New function.
25724 (const_vector_elt): Likewise.
25725 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
25726 of CONST_VECTOR_ELT.
25728 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25730 * expr.c: Include rtx-vector-builder.h.
25731 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
25732 directly on the tree encoding.
25733 (const_vector_from_tree): Likewise.
25734 * optabs.c: Include rtx-vector-builder.h.
25735 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
25736 sequence of "u" values.
25737 * vec-perm-indices.c: Include rtx-vector-builder.h.
25738 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
25739 directly on the vec_perm_indices encoding.
25741 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25743 * doc/rtl.texi (const_vector): Describe new encoding scheme.
25744 * Makefile.in (OBJS): Add rtx-vector-builder.o.
25745 * rtx-vector-builder.h: New file.
25746 * rtx-vector-builder.c: Likewise.
25747 * rtl.h (rtx_def::u2): Add a const_vector field.
25748 (CONST_VECTOR_NPATTERNS): New macro.
25749 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
25750 (CONST_VECTOR_DUPLICATE_P): Likewise.
25751 (CONST_VECTOR_STEPPED_P): Likewise.
25752 (CONST_VECTOR_ENCODED_ELT): Likewise.
25753 (const_vec_duplicate_p): Check for a duplicated vector encoding.
25754 (unwrap_const_vec_duplicate): Likewise.
25755 (const_vec_series_p): Check for a non-duplicated vector encoding.
25756 Say that the function only returns true for integer vectors.
25757 * emit-rtl.c: Include rtx-vector-builder.h.
25758 (gen_const_vec_duplicate_1): Delete.
25759 (gen_const_vector): Call gen_const_vec_duplicate instead of
25760 gen_const_vec_duplicate_1.
25761 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
25762 (gen_const_vec_duplicate): Use rtx_vector_builder.
25763 (gen_const_vec_series): Likewise.
25764 (gen_rtx_CONST_VECTOR): Likewise.
25765 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
25766 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
25767 Build a new vector rather than modifying a CONST_VECTOR in-place.
25768 (handle_special_swappables): Update call accordingly.
25769 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
25770 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
25771 Build a new vector rather than modifying a CONST_VECTOR in-place.
25772 (handle_special_swappables): Update call accordingly.
25774 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25776 * simplify-rtx.c (simplify_const_binary_operation): Use
25777 CONST_VECTOR_ELT instead of XVECEXP.
25779 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25781 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
25782 the selector elements to be different from the data elements
25783 if the selector is a VECTOR_CST.
25784 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
25785 ssizetype for the selector.
25787 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25789 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
25790 before testing each element individually.
25791 * tree-vect-generic.c (lower_vec_perm): Likewise.
25793 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25795 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
25796 * selftest-run-tests.c (selftest::run_tests): Call it.
25797 * vector-builder.h (vector_builder::operator ==): New function.
25798 (vector_builder::operator !=): Likewise.
25799 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
25800 (vec_perm_indices::all_from_input_p): New function.
25801 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
25802 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
25803 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
25804 instead of reading the VECTOR_CST directly. Detect whether both
25805 vector inputs are the same before constructing the vec_perm_indices,
25806 and update the number of inputs argument accordingly. Use the
25807 utility functions added above. Only construct sel2 if we need to.
25809 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25811 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
25812 the broadcast of the low byte.
25813 (expand_mult_highpart): Use an explicit encoding for the permutes.
25814 * optabs-query.c (can_mult_highpart_p): Likewise.
25815 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
25816 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25817 (vectorizable_bswap): Likewise.
25818 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
25819 explicit encoding for the power-of-2 permutes.
25820 (vect_permute_store_chain): Likewise.
25821 (vect_grouped_load_supported): Likewise.
25822 (vect_permute_load_chain): Likewise.
25824 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25826 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
25827 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
25828 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
25829 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
25830 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
25831 (vect_gen_perm_mask_any): Likewise.
25833 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25835 * int-vector-builder.h: New file.
25836 * vec-perm-indices.h: Include int-vector-builder.h.
25837 (vec_perm_indices): Redefine as an int_vector_builder.
25838 (auto_vec_perm_indices): Delete.
25839 (vec_perm_builder): Redefine as a stand-alone class.
25840 (vec_perm_indices::vec_perm_indices): New function.
25841 (vec_perm_indices::clamp): Likewise.
25842 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
25843 (vec_perm_indices::new_vector): New function.
25844 (vec_perm_indices::new_expanded_vector): Update for new
25845 vec_perm_indices class.
25846 (vec_perm_indices::rotate_inputs): New function.
25847 (vec_perm_indices::all_in_range_p): Operate directly on the
25848 encoded form, without computing elided elements.
25849 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
25850 encoding. Update for new vec_perm_indices class.
25851 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
25852 the given vec_perm_builder.
25853 (expand_vec_perm_var): Update vec_perm_builder constructor.
25854 (expand_mult_highpart): Use vec_perm_builder instead of
25855 auto_vec_perm_indices.
25856 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
25857 vec_perm_indices instead of auto_vec_perm_indices. Use a single
25858 or double series encoding as appropriate.
25859 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
25860 vec_perm_indices instead of auto_vec_perm_indices.
25861 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
25862 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
25863 (vect_permute_store_chain): Likewise.
25864 (vect_grouped_load_supported): Likewise.
25865 (vect_permute_load_chain): Likewise.
25866 (vect_shift_permute_load_chain): Likewise.
25867 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
25868 (vect_transform_slp_perm_load): Likewise.
25869 (vect_schedule_slp_instance): Likewise.
25870 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25871 (vectorizable_mask_load_store): Likewise.
25872 (vectorizable_bswap): Likewise.
25873 (vectorizable_store): Likewise.
25874 (vectorizable_load): Likewise.
25875 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
25876 vec_perm_indices instead of auto_vec_perm_indices. Use
25877 tree_to_vec_perm_builder to read the vector from a tree.
25878 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
25879 vec_perm_builder instead of a vec_perm_indices.
25880 (have_whole_vector_shift): Use vec_perm_builder and
25881 vec_perm_indices instead of auto_vec_perm_indices. Leave the
25882 truncation to calc_vec_perm_mask_for_shift.
25883 (vect_create_epilog_for_reduction): Likewise.
25884 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
25885 from auto_vec_perm_indices to vec_perm_indices.
25886 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
25887 instead of changing individual elements.
25888 (aarch64_vectorize_vec_perm_const): Use new_vector to install
25889 the vector in d.perm.
25890 * config/arm/arm.c (expand_vec_perm_d::perm): Change
25891 from auto_vec_perm_indices to vec_perm_indices.
25892 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
25893 instead of changing individual elements.
25894 (arm_vectorize_vec_perm_const): Use new_vector to install
25895 the vector in d.perm.
25896 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
25897 Update vec_perm_builder constructor.
25898 (rs6000_expand_interleave): Likewise.
25899 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
25900 (rs6000_expand_interleave): Likewise.
25902 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25904 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
25905 to qimode could truncate the indices.
25906 * optabs.c (expand_vec_perm_var): Likewise.
25908 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25910 * Makefile.in (OBJS): Add vec-perm-indices.o.
25911 * vec-perm-indices.h: New file.
25912 * vec-perm-indices.c: Likewise.
25913 * target.h (vec_perm_indices): Replace with a forward class
25915 (auto_vec_perm_indices): Move to vec-perm-indices.h.
25916 * optabs.h: Include vec-perm-indices.h.
25917 (expand_vec_perm): Delete.
25918 (selector_fits_mode_p, expand_vec_perm_var): Declare.
25919 (expand_vec_perm_const): Declare.
25920 * target.def (vec_perm_const_ok): Replace with...
25921 (vec_perm_const): ...this new hook.
25922 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
25923 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
25924 * doc/tm.texi: Regenerate.
25925 * optabs.def (vec_perm_const): Delete.
25926 * doc/md.texi (vec_perm_const): Likewise.
25927 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
25928 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
25929 expand_vec_perm for constant permutation vectors. Assert that
25930 the mode of variable permutation vectors is the integer equivalent
25931 of the mode that is being permuted.
25932 * optabs-query.h (selector_fits_mode_p): Declare.
25933 * optabs-query.c: Include vec-perm-indices.h.
25934 (selector_fits_mode_p): New function.
25935 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
25936 is defined, instead of checking whether the vec_perm_const_optab
25937 exists. Use targetm.vectorize.vec_perm_const instead of
25938 targetm.vectorize.vec_perm_const_ok. Check whether the indices
25939 fit in the vector mode before using a variable permute.
25940 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
25941 vec_perm_indices instead of an rtx.
25942 (expand_vec_perm): Replace with...
25943 (expand_vec_perm_const): ...this new function. Take the selector
25944 as a vec_perm_indices rather than an rtx. Also take the mode of
25945 the selector. Update call to shift_amt_for_vec_perm_mask.
25946 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
25947 Use vec_perm_indices::new_expanded_vector to expand the original
25948 selector into bytes. Check whether the indices fit in the vector
25949 mode before using a variable permute.
25950 (expand_vec_perm_var): Make global.
25951 (expand_mult_highpart): Use expand_vec_perm_const.
25952 * fold-const.c: Includes vec-perm-indices.h.
25953 * tree-ssa-forwprop.c: Likewise.
25954 * tree-vect-data-refs.c: Likewise.
25955 * tree-vect-generic.c: Likewise.
25956 * tree-vect-loop.c: Likewise.
25957 * tree-vect-slp.c: Likewise.
25958 * tree-vect-stmts.c: Likewise.
25959 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
25961 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
25962 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
25963 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
25964 (aarch64_vectorize_vec_perm_const): ...this new function.
25965 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25966 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25967 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
25968 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
25969 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25970 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25971 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
25973 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
25974 check for NEON modes.
25975 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
25976 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
25977 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
25978 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
25980 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
25981 the old VEC_PERM_CONST conditions.
25982 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
25983 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
25984 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
25985 (ia64_vectorize_vec_perm_const_ok): Merge into...
25986 (ia64_vectorize_vec_perm_const): ...this new function.
25987 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
25988 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
25989 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
25990 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
25991 * config/mips/mips.c (mips_expand_vec_perm_const)
25992 (mips_vectorize_vec_perm_const_ok): Merge into...
25993 (mips_vectorize_vec_perm_const): ...this new function.
25994 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
25995 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
25996 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
25997 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
25998 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
25999 (rs6000_expand_vec_perm_const): Delete.
26000 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
26002 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26003 (altivec_expand_vec_perm_const_le): Take each operand individually.
26004 Operate on constant selectors rather than rtxes.
26005 (altivec_expand_vec_perm_const): Likewise. Update call to
26006 altivec_expand_vec_perm_const_le.
26007 (rs6000_expand_vec_perm_const): Delete.
26008 (rs6000_vectorize_vec_perm_const_ok): Delete.
26009 (rs6000_vectorize_vec_perm_const): New function.
26010 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
26011 an element count and rtx array.
26012 (rs6000_expand_extract_even): Update call accordingly.
26013 (rs6000_expand_interleave): Likewise.
26014 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
26015 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
26016 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
26017 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
26018 (rs6000_expand_vec_perm_const): Delete.
26019 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
26020 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26021 (altivec_expand_vec_perm_const_le): Take each operand individually.
26022 Operate on constant selectors rather than rtxes.
26023 (altivec_expand_vec_perm_const): Likewise. Update call to
26024 altivec_expand_vec_perm_const_le.
26025 (rs6000_expand_vec_perm_const): Delete.
26026 (rs6000_vectorize_vec_perm_const_ok): Delete.
26027 (rs6000_vectorize_vec_perm_const): New function. Remove stray
26028 reference to the SPE evmerge intructions.
26029 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
26030 an element count and rtx array.
26031 (rs6000_expand_extract_even): Update call accordingly.
26032 (rs6000_expand_interleave): Likewise.
26033 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
26034 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
26036 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26038 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26040 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
26041 vector mode and that that mode matches the mode of the data
26043 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
26044 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
26045 directly using expand_vec_perm_1 when forcing selectors into
26047 (expand_vec_perm_var): New function, split out from expand_vec_perm.
26049 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26051 * optabs-query.h (can_vec_perm_p): Delete.
26052 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
26053 * optabs-query.c (can_vec_perm_p): Split into...
26054 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
26055 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
26056 particular selector is valid.
26057 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26058 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
26059 (vect_grouped_load_supported): Likewise.
26060 (vect_shift_permute_load_chain): Likewise.
26061 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
26062 (vect_transform_slp_perm_load): Likewise.
26063 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26064 (vectorizable_bswap): Likewise.
26065 (vect_gen_perm_mask_checked): Likewise.
26066 * fold-const.c (fold_ternary_loc): Likewise. Don't take
26067 implementations of variable permutation vectors into account
26068 when deciding which selector to use.
26069 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
26070 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
26071 with a false third argument.
26072 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
26073 to test whether the constant selector is valid and can_vec_perm_var_p
26074 to test whether a variable selector is valid.
26076 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26078 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
26079 * optabs-query.c (can_vec_perm_p): Likewise.
26080 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
26081 instead of vec_perm_indices.
26082 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
26083 (vect_gen_perm_mask_checked): Likewise,
26084 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
26085 (vect_gen_perm_mask_checked): Likewise,
26087 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26089 * optabs-query.h (qimode_for_vec_perm): Declare.
26090 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
26091 (qimode_for_vec_perm): ...this new function.
26092 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
26094 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
26096 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
26097 does not have a conditional at the top.
26099 2018-01-02 Richard Biener <rguenther@suse.de>
26101 * ipa-inline.c (big_speedup_p): Fix expression.
26103 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
26106 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
26109 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
26113 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
26114 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
26115 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
26116 cond_taken_branch_cost 3->4.
26118 2018-01-01 Jakub Jelinek <jakub@redhat.com>
26120 PR tree-optimization/83581
26121 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
26122 TODO_cleanup_cfg if any changes have been made.
26124 PR middle-end/83608
26125 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
26126 convert_modes if target mode has the right side, but different mode
26129 PR middle-end/83609
26130 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
26131 last argument when extracting from CONCAT. If either from_real or
26132 from_imag is NULL, use expansion through memory. If result is not
26133 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
26134 the parts directly to inner mode, if even that fails, use expansion
26137 PR middle-end/83623
26138 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
26139 check for bswap in mode rather than HImode and use that in expand_unop
26142 Copyright (C) 2018 Free Software Foundation, Inc.
26144 Copying and distribution of this file, with or without modification,
26145 are permitted in any medium without royalty provided the copyright
26146 notice and this notice are preserved.