1 2018-09-21 Olivier Hainque <hainque@adacore.com>
3 * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
5 2018-09-21 Olivier Hainque <hainque@adacore.com>
7 * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
8 Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
9 (VXWORKS_LIBS_RTP): Minor reordering.
11 2018-09-21 Olivier Hainque <hainque@adacore.com>
13 * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
14 (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
16 2018-09-21 Olivier Hainque <hainque@adacore.com>
18 * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
19 (PTRDIFF_TYPE): Likewise.
21 2018-09-21 Olivier Hainque <hainque@adacore.com>
23 * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
24 triplet, similar to support for VxWorks7.
25 * config/vxworks-dummy.h: Provide a default definition
26 of TARGET_VXWORKS64 to 0.
28 2018-09-21 Olivier Hainque <hainque@adacore.com>
30 * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
31 * config/vxworks-dummy.h: here.
33 2018-09-21 Olivier Hainque <hainque@adacore.com>
35 * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
37 2018-09-21 Shaokun Zhang <zhangshaokun@hisilicon.com>
38 Bo Zhou <zbo.zhou@hisilicon.com>
40 * config/aarch64/aarch64-cores.def (tsv110): New CPU.
41 * config/aarch64/aarch64-tune.md: Regenerated.
42 * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
43 * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
44 * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
46 2018-09-21 Andrew Stubbs <ams@codesourcery.com>
47 Julian Brown <julian@codesourcery.com>
49 * builtins.c (get_builtin_sync_mem): Handle address spaces.
51 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
53 * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
54 if the call takes a static chain.
56 2018-09-21 Martin Liska <mliska@suse.cz>
58 * auto-profile.c (autofdo_source_profile::read): Do not
60 (read_profile): Do not add working sets.
61 (read_autofdo_file): Remove sum_all.
62 (afdo_callsite_hot_enough_for_early_inline): Remove const
64 * coverage.c (struct counts_entry): Remove gcov_summary.
65 (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
66 do not support GCOV_TAG_PROGRAM_SUMMARY.
67 (get_coverage_counts): Remove summary and expected
69 * coverage.h (get_coverage_counts): Likewise.
70 * doc/gcov-dump.texi: Remove -w option.
71 * gcov-dump.c (dump_working_sets): Remove.
72 (main): Do not support '-w' option.
73 (print_usage): Likewise.
74 (tag_summary): Likewise.
75 * gcov-io.c (gcov_write_summary): Do not dump
77 (gcov_read_summary): Likewise.
78 (gcov_histo_index): Remove.
79 (gcov_histogram_merge): Likewise.
80 (compute_working_sets): Likewise.
81 * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
83 (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
84 (GCOV_TAG_SUMMARY_LENGTH): Adjust.
85 (GCOV_HISTOGRAM_SIZE): Remove.
86 (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
87 (struct gcov_summary): Simplify rapidly just
88 to runs and sum_max fields.
89 (gcov_histo_index): Remove.
90 (NUM_GCOV_WORKING_SETS): Likewise.
91 (compute_working_sets): Likewise.
92 * gcov-tool.c (print_overlap_usage_message): Remove
94 * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
95 (output_lines): Remove program related line.
96 * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
97 * lto-cgraph.c (output_profile_summary): Do not stream GCOV
99 (input_profile_summary): Do not read it.
100 (merge_profile_summaries): And do not merge it.
101 (input_symtab): Do not call removed function.
102 * modulo-sched.c (sms_schedule): Do not print sum_max.
103 * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
104 removed when histogram method was invented.
105 (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
107 * postreload-gcse.c (eliminate_partially_redundant_load): Fix
109 * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
110 and dump selected value.
111 * profile.c (add_working_set): Remove.
112 (get_working_sets): Likewise.
113 (find_working_set): Likewise.
114 (get_exec_counts): Do not work with working sets.
115 (read_profile_edge_counts): Do not inform as sum_max is removed.
116 (compute_branch_probabilities): Likewise.
117 (compute_value_histograms): Remove argument for call of
119 * profile.h: Do not make gcov_summary const.
121 2018-09-21 Monk Chiang <sh.chiang04@gmail.com>
123 * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
125 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
127 PR tree-optimization/86990
128 * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
129 Check that the entire merged store group is made of constants only for
132 2018-09-20 Allan Sandfeld Jensen <allan.jensen@qt.io>
134 * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
135 (VTABLE_VERIFICATION_SPEC): Likewise.
136 (SANITIZER_EARLY_SPEC): Likewise.
137 (SANITIZER_SPEC): Likewise.
138 * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
139 * doc/invoke.texi (Link Options): Document -r.
141 2018-09-20 Richard Biener <rguenther@suse.de>
144 * gimplify.c (gimplify_expr): Retain alignment of
145 addressable lvalue in dereference.
147 2018-09-20 Alexandre Oliva <aoliva@redhat.com>
150 * configure.ac: Check for .loc is_stmt support.
151 * configure, config.in: Rebuilt.
152 * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
155 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
157 * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
158 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
161 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
163 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
165 * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
168 2018-09-20 Marek Polacek <polacek@redhat.com>
170 * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
172 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
174 PR tree-optimization/87288
175 * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
176 into account when determining PEELING_FOR_NITERS.
178 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
180 PR tree-optimization/86877
181 * tree-vect-loop.c (vect_analyze_loop_2): Call
182 vect_verify_datarefs_alignment.
184 2018-09-19 Marek Polacek <polacek@redhat.com>
186 * doc/invoke.texi: Document -Wclass-conversion.
188 2018-09-19 John David Anglin <danglin@gcc.gnu.org>
190 * config/pa/pa.c (pa_adjust_priority): Delete.
191 (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
193 * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
194 (atomic_storehi): Likewise.
195 (atomic_storesi): Likewise.
196 (atomic_loaddi): Restore compare and swap exchange loop code.
198 2018-09-19 Segher Boessenkool <segher@kernel.crashing.org>
200 PR rtl-optimization/86902
201 * combine.c (try_combine): When changing the CC mode used, don't change
202 an unrelated mode in other_insn to that new CC mode.
204 2018-09-19 David Malcolm <dmalcolm@redhat.com>
206 * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
207 with %T in place of calls to dump_generic_expr.
208 (prune_runtime_alias_test_list): Likewise.
209 (create_runtime_alias_checks): Likewise.
210 * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
211 (vect_analyze_data_ref_dependence): Likewise.
212 (vect_slp_analyze_data_ref_dependence): Likewise.
213 (vect_record_base_alignment): Likewise. Use %G in place of call
215 (vect_compute_data_ref_alignment): Likewise.
216 (verify_data_ref_alignment): Likewise.
217 (vect_find_same_alignment_drs): Likewise.
218 (vect_analyze_group_access_1): Likewise.
219 (vect_analyze_data_ref_accesses): Likewise.
220 (dependence_distance_ge_vf): Likewise.
221 (dump_lower_bound): Likewise.
222 (vect_prune_runtime_alias_test_list): Likewise.
223 (vect_find_stmt_data_reference): Likewise.
224 (vect_analyze_data_refs): Likewise.
225 (vect_create_addr_base_for_vector_ref): Likewise.
226 (vect_create_data_ref_ptr): Likewise.
227 * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
228 (vect_can_advance_ivs_p): Likewise.
229 (vect_update_ivs_after_vectorizer): Likewise.
230 (vect_gen_prolog_loop_niters): Likewise.
231 (vect_prepare_for_masked_peels): Likewise.
232 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
233 (vect_determine_vectorization_factor): Likewise.
234 (vect_is_simple_iv_evolution): Likewise.
235 (vect_analyze_scalar_cycles_1): Likewise.
236 (vect_analyze_loop_operations): Likewise.
237 (report_vect_op): Likewise.
238 (vect_is_slp_reduction): Likewise.
239 (check_reduction_path): Likewise.
240 (vect_is_simple_reduction): Likewise.
241 (vect_create_epilog_for_reduction): Likewise.
242 (vect_finalize_reduction:): Likewise.
243 (vectorizable_induction): Likewise.
244 (vect_transform_loop_stmt): Likewise.
245 (vect_transform_loop): Likewise.
246 (optimize_mask_stores): Likewise.
247 * tree-vect-patterns.c (vect_pattern_detected): Likewise.
248 (vect_split_statement): Likewise.
249 (vect_recog_over_widening_pattern): Likewise.
250 (vect_recog_average_pattern): Likewise.
251 (vect_determine_min_output_precision_1): Likewise.
252 (vect_determine_precisions_from_range): Likewise.
253 (vect_determine_precisions_from_users): Likewise.
254 (vect_mark_pattern_stmts): Likewise.
255 (vect_pattern_recog_1): Likewise.
256 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
257 (vect_record_max_nunits): Likewise.
258 (vect_build_slp_tree_1): Likewise.
259 (vect_build_slp_tree_2): Likewise.
260 (vect_print_slp_tree): Likewise.
261 (vect_analyze_slp_instance): Likewise.
262 (vect_detect_hybrid_slp_stmts): Likewise.
263 (vect_detect_hybrid_slp_1): Likewise.
264 (vect_slp_analyze_operations): Likewise.
265 (vect_slp_analyze_bb_1): Likewise.
266 (vect_transform_slp_perm_load): Likewise.
267 (vect_schedule_slp_instance): Likewise.
268 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
269 (vect_mark_stmts_to_be_vectorized): Likewise.
270 (vect_init_vector_1): Likewise.
271 (vect_get_vec_def_for_operand): Likewise.
272 (vect_finish_stmt_generation_1): Likewise.
273 (vect_check_load_store_mask): Likewise.
274 (vectorizable_call): Likewise.
275 (vectorizable_conversion): Likewise.
276 (vectorizable_operation): Likewise.
277 (vectorizable_load): Likewise.
278 (vect_analyze_stmt): Likewise.
279 (vect_is_simple_use): Likewise.
280 (vect_get_vector_types_for_stmt): Likewise.
281 (vect_get_mask_type_for_stmt): Likewise.
282 * tree-vectorizer.c (increase_alignment): Likewise.
284 2018-09-19 Andrew Stubbs <ams@codesourcery.com>
286 * doc/rtl.texi: Adjust vec_select description.
287 * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
288 non-constant selectors.
290 2018-09-19 Matthew Malcomson <matthew.malcomson@arm.com>
292 * config/aarch64/aarch64-protos.h
293 (aarch64_offset_9bit_signed_unscaled_p): New declaration.
294 * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
295 (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
296 * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
297 (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
298 (AARCH64_FL_PROFILE): Move index so flags are ordered.
299 (AARCH64_ISA_RCPC8_4): New flag.
300 * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
301 to aarch64_offset_9bit_signed_unscaled_p.
302 * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
304 * config/aarch64/constraints.md (Ust): New constraint.
305 * config/aarch64/predicates.md.
306 (aarch64_9bit_offset_memory_operand): New predicate.
307 (aarch64_rcpc_memory_operand): New predicate.
309 2018-09-19 Eric Botcazou <ebotcazou@adacore.com>
311 PR rtl-optimization/87361
312 * rtlanal.c (nonzero_bits1): Revert accidental change.
314 2018-09-19 Richard Biener <rguenther@suse.de>
316 PR tree-optimization/87349
317 PR tree-optimization/87342
318 * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
320 2018-09-18 Marek Polacek <polacek@redhat.com>
322 P1064R0 - Allowing Virtual Function Calls in Constant Expressions
323 * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
325 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
327 * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
329 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
331 PR rtl-optimization/86882
332 * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
334 2018-09-18 Uros Bizjak <ubizjak@gmail.com>
336 * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
337 *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
339 2018-09-18 Jonathan Wakely <jwakely@redhat.com>
342 * doc/invoke.texi (Link Options): Fix formatting and grammar.
344 2018-09-18 Richard Biener <rguenther@suse.de>
347 * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
348 (compute_samebase_partition_bases): Likewise.
349 (coalesce_ssa_name): Always use compute_optimized_partition_bases.
350 (gimple_can_coalesce_p): Simplify.
352 2018-09-18 Hans-Peter Nilsson <hp@bitrange.com>
354 Handle a library implementation of ffs calling __builtin_ffs.
355 * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
356 (mmix_init_libfuncs): New function: make __builtin_ffs expand
359 2018-09-17 David Malcolm <dmalcolm@redhat.com>
361 * diagnostic-show-locus.c (class layout_range): Add field
363 (layout_range::layout_range): Add "original_idx" param and use it
364 to initialize new field.
365 (make_range): Use 0 for original_idx.
366 (layout::layout): Pass in index to calls to
367 maybe_add_location_range.
368 (layout::maybe_add_location_range): Add param "original_idx" and
369 pass it on to layout_range.
370 (layout::print_any_labels): Pass on range->m_original_idx to
372 (gcc_rich_location::add_location_if_nearby): Use 0 for
374 * gcc-rich-location.h (text_range_label::get_text): Update for new
376 (range_label_for_type_mismatch::get_text): Likewise.
378 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
380 * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
382 2018-09-17 David Malcolm <dmalcolm@redhat.com>
384 * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
385 format_string_diagnostic_t.
386 (fmtwarn_n): Likewise.
387 * substring-locations.c
388 (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
389 (format_warning_n_va): Convert to...
390 (format_string_diagnostic_t::emit_warning_n_va): ...this.
391 (format_warning_va): Convert to...
392 (format_string_diagnostic_t::emit_warning_va): ...this.
393 (format_warning_at_substring): Convert to...
394 (format_string_diagnostic_t::emit_warning): ...this.
395 (format_warning_at_substring_n): Convert to...
396 (format_string_diagnostic_t::emit_warning_n): ...this.
397 * substring-locations.h (class format_string_diagnostic_t): New
399 (format_warning_va): Convert to
400 format_string_diagnostic_t::emit_warning_va.
401 (format_warning_n_va): Convert to
402 format_string_diagnostic_t::emit_warning_n_va.
403 (format_warning_at_substring): Convert to
404 format_string_diagnostic_t::emit_warning.
405 (format_warning_at_substring_n): Convert to
406 format_string_diagnostic_t::emit_warning_n.
408 2018-09-17 Cesar Philippidis <cesar@codesourcery.com>
409 Bernd Schmidt <bernds_cb1@t-online.de>
411 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
414 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
416 * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
417 operand 0 predicate to nonimmediate operand.
418 (rint<mode>2_frndint): Remove insn pattern.
419 (rint<mode>2): Change operand 1 predicate to general_operand.
420 Extend operand 1 to XFmode and generate rintxf2 insn.
421 (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
422 Do not use X87MODEF mode macro.
423 (frndintxf2_<rounding>_i387): Rename from
424 frndint<mode>2_<rounding>_i387. Do not use X87MODEF mode macro.
425 (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
426 to XFmode and generate significandxf3 insn.
428 2018-09-17 Richard Biener <rguenther@suse.de>
430 PR tree-optimization/87328
431 * tree-ssa-sccvn.c (process_bb): Remove assertion about not
432 visiting unexecutable backedges when not iterating.
433 (do_rpo_vn): Mark all edges not executable even when not
436 2018-09-17 Martin Jambor <mjambor@suse.cz>
439 * doc/invoke.texi (Warning Options): Likewise.
441 2018-09-17 Richard Biener <rguenther@suse.de>
443 PR tree-optimization/87301
444 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
445 clean EH info from leftover copy assignments.
447 2018-09-17 Martin Liska <mliska@suse.cz>
449 PR gcov-profile/85871
450 * gcov.c (output_intermediate_file): Fix out of bounds
453 2018-09-17 Vineet Gupta <vgupta@synopsys.com>
455 * config/arc/arc.c: Object attributes for core4 not reflected
457 * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
460 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
462 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
464 2018-09-17 Martin Liska <mliska@suse.cz>
466 * doc/gcov.texi: Document new option --use-hotness-colors.
467 * gcov.c (struct source_info): Declare new field.
468 (source_info::source_info): Set default for maximum_count.
469 (print_usage): Add new -q option.
470 (process_args): Process it.
471 (accumulate_line_info): Save src->maximum_count.
472 (output_line_beginning): Make color line number if
473 flag_use_hotness_colors is set.
474 (output_line_details): Pass default argument value.
475 (output_lines): Pass src->maximum_count.
477 2018-09-17 Martin Liska <mliska@suse.cz>
479 * common/config/i386/i386-common.c (ix86_get_valid_option_values):
480 Use processor_names table.
481 * config/i386/i386.c (ix86_default_align): Use
482 processor_cost_table for alignment values.
483 (ix86_option_override_internal): Use processor_names.
484 (ix86_function_specific_print): Likewise.
485 * config/i386/i386.h (struct processor_costs):
486 Add alignment values.
487 (struct ptt): Remove and replace with const char *.
488 * config/i386/x86-tune-costs.h (struct processor_costs):
489 Declare default alignments for all costs.
491 2018-09-17 Aldy Hernandez <aldyh@redhat.com>
493 * tree-vrp.c (extract_range_from_unary_expr): Do not special case
494 symbolics or VR_VARYING ranges for ABS_EXPR.
495 * wide-int-range.cc (wide_int_range_abs): Return positive numbers
496 when range will wrap.
498 2018-09-15 Eric Botcazou <ebotcazou@adacore.com>
501 * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
502 before and after a JUMP_TABLE_DATA.
504 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
507 * dojump.c (do_compare_and_jump): Canonicalize function pointers
508 when one operand is a function pointer. Use POINTER_TYPE_P and
509 FUNC_OR_METHOD_TYPE_P.
510 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
511 * fold-const.c (build_range_check): Likewise.
512 * match.pd (simple_comparison): Likewise.
514 2018-09-14 David Malcolm <dmalcolm@redhat.com>
517 * spellcheck.c (get_edit_distance_cutoff): New function.
518 (selftest::test_edit_distance_unit_test_oneway): Rename to...
519 (selftest::test_get_edit_distance_one_way): ...this.
520 (selftest::test_get_edit_distance_unit): Rename to...
521 (selftest::test_get_edit_distance_both_ways): ...this.
522 (selftest::test_edit_distances): Move tests to this new function,
523 and test some more pairs of strings. Update for above renaming.
524 (selftest::get_old_cutoff): New function.
525 (selftest::test_get_edit_distance_cutoff): New function.
526 (selftest::assert_suggested_for): New function.
527 (ASSERT_SUGGESTED_FOR): New macro.
528 (selftest::assert_not_suggested_for): New function.
529 (ASSERT_NOT_SUGGESTED_FOR): New macro.
530 (selftest::test_suggestions): New function.
531 (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
532 tests to selftest::test_edit_distances and call it. Add calls to
533 selftest::test_get_edit_distance_cutoff and
534 selftest::test_suggestions.
535 * spellcheck.h (get_edit_distance_cutoff): New function declaration.
536 (best_match::consider): Replace hard-coded cutoff calculation with
538 (best_match::get_cutoff): New declaration.
539 (best_match::get_best_meaningful_candidate): Likewise.
541 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
543 * builtins.c (fold_builtin_strlen): Remove TODO comment.
545 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
548 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
550 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
551 terminated string literal.
553 2018-09-14 Martin Sebor <msebor@redhat.com>
555 * builtins.c (unterminated_array): Handle ARRAY_REF.
556 (expand_builtin_stpcpy_1): Detect unterminated char arrays.
557 * builtins.h (unterminated_array): Declare extern.
558 * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
560 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
563 2018-09-14 Martin Sebor <msebor@redhat.com>
564 Jeff Law <law@redhat.com>
566 * builtins.c (unterminated_array): New.
567 (expand_builtin_strcpy): Adjust.
568 (expand_builtin_strcpy_args): Detect unterminated arrays.
569 * gimple-fold.c (get_maxval_strlen): Add argument. Detect
571 * gimple-fold.h (get_maxval_strlen): Add argument.
572 (gimple_fold_builtin_strcpy): Detec unterminated arrays.
574 * gimple-fold.c (get_range_strlen): Add argument.
575 (get_maxval_strlen): Adjust.
576 * gimple-fold.h (get_range_strlen): Add argument.
578 2018-09-14 Wei Xiao <wei3.xiao@intel.com>
580 * config/i386/movdirintrin.h: Fix copyright year.
582 2018-09-14 Uros Bizjak <ubizjak@gmail.com>
584 * reg-stack.c: Include regs.h.
585 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
586 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
587 FIRST_STACK_REG, not DFmode.
588 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
589 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX.
590 (change stack): Default register mode to the reg_raw_mode of
591 FIRST_STACK_REG, not DFmode.
592 * config/i386/i386.md (*swap<mode>): Remove insn pattern.
593 (*swapxf): Rename from swapxf.
595 2018-09-14 Carl Love <cel@us.ibm.com>
597 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
598 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
600 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
603 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
606 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
609 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
611 2018-09-14 Jason Merrill <jason@redhat.com>
613 Fix --enable-gather-detailed-mem-stats.
614 * hash-table.c (hash_table_usage): Change from variable to function.
615 * hash-table.h: Adjust.
616 * Makefile.in: Add missing dependencies on hash-table.h.
618 2018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
620 PR tree-optimization/87259
622 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
623 execute_cse_reciprocals_1 has tried transforming.
625 2018-09-14 Aldy Hernandez <aldyh@redhat.com>
627 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
628 VR_VARYING for PLUS/MINUS_EXPR.
630 2018-09-14 Ilya Leoshkevich <iii@linux.ibm.com>
632 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
635 2018-09-14 Richard Biener <rguenther@suse.de>
638 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
639 bits for the merged partition.
641 2018-09-13 Martin Sebor <msebor@redhat.com>
642 Bernd Edlinger <bernd.edlinger@hotmail.de>
644 * builtins.h (c_srlen): Add argument.
645 * builtins.c (warn_string_no_nul): New function.
646 (c_strlen): Add argument and use it. Update recursive calls.
647 Pass DECL argument to string_constant to get info on non
648 terminated strings. Update *NONSTR as needed.
649 (fold_builtin_strlen): Add argument to calls to c_strlen.
650 Warn for unterminated arrays.
651 (warn_string_no_null): Add prototype.
652 * expr.c (string_constant): Update arguments. Update recursive
653 calls appropriately. Detect missing NUL terminator and outermost
654 declaration its missing in.
655 Improve checks for arrays with nonzero lower bound or elements
656 that are not a single byte. Simplify offset computation.
657 Simplify checks for non-NUL terminated strings.
658 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
659 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
661 2018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
663 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
665 * fold-const.c (c_getstr): Fix function comment. Remove unused third
666 argument. Fix range checks.
667 * fold-const.h (c_getstr): Adjust protoype.
668 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
669 string is constant but contains no NUL byte.
671 * expr.c (string_constant): Adjust function comment.
672 Remove bogus check for zero termination.
674 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
676 * varasm.c (compare_constant): Compare type size of STRING_CSTs.
677 (get_constant_size): Don't make STRING_CSTs larger than they are.
678 (check_string_literal): New check function for STRING_CSTs.
679 (output_constant): Use it.
681 2018-09-13 Eric Botcazou <ebotcazou@adacore.com>
684 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
686 2018-09-13 Richard Biener <rguenther@suse.de>
688 PR tree-optimization/87263
689 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
690 (struct unwind_state): Add max_rpo field.
691 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
692 Compute max_rpo, the max RPO number a block can be backwards reached
693 from. Re-write non-iterating mode to a RPO ordered worklist approach,
694 separating it from the iterating mode.
696 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
698 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
699 (rfs_decision): New scheduling decision.
701 2018-09-13 Richard Biener <rguenther@suse.de>
704 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
705 (vn_nary_op_insert_pieces_predicated): Do not write useless
706 valid_dominated_by_p entry outside of the allocated storage.
708 2018-09-13 Omar Sandoval <osandov@osandov.com>
709 Tom de Vries <tdevries@suse.de>
712 * dwarf2out.c (is_c): New function.
713 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
715 2018-09-13 Sam Tebbs <sam.tebbs@arm.com>
718 * config/aarch64/aarch64.md (*aarch64_bfxil):
720 * config/aarch64/constraints.md (Ulc): Define.
721 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
723 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
726 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
728 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
729 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
730 aarch64_layout_frame call.
731 (aarch64_expand_epilogue): Likewise.
732 (aarch64_initial_elimination_offset): Likewise.
733 (aarch64_get_separate_components): Likewise.
734 (aarch64_use_return_insn_p): Likewise.
735 (aarch64_layout_frame): Remove unneeded check.
737 2018-09-13 Jakub Jelinek <jakub@redhat.com>
739 * configure.ac: Only append
740 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
741 gcc_config_arguments if it was never reconfigured or last reconfigure
742 was with different arguments.
743 * configure: Regenerated.
745 2018-09-13 Jakub Jelinek <jakub@redhat.com>
746 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
749 * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
750 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
752 2018-09-13 Jakub Jelinek <jakub@redhat.com>
754 PR tree-optimization/87287
755 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
756 X % C == 0 to X % (unsigned) C == 0 optimization to ...
757 * match.pd (X % C == 0): ... here. New optimization.
759 2018-09-12 Jakub Jelinek <jakub@redhat.com>
762 * expr.h (maybe_optimize_mod_cmp): Declare.
763 * expr.c (mod_inv): New function.
764 (maybe_optimize_mod_cmp): New function.
765 (do_store_flag): Use it.
766 * cfgexpand.c (expand_gimple_cond): Likewise.
768 2018-09-09 Cesar Philippidis <cesar@codesourcery.com>
769 Julian Brown <julian@codesourcery.com>
772 * gimplify.c (gimplify_scan_omp_clauses): Set
773 target_firstprivatize_array_bases in OpenACC parallel and kernels
774 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
775 OpenACC data regions.
777 2018-09-12 Uros Bizjak <ubizjak@gmail.com>
779 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
780 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
782 2018-09-12 Richard Biener <rguenther@suse.de>
784 PR tree-optimization/87280
785 * tree-ssa-sccvn.c (process_bb): Handle the case of executable
786 edge but unreachable target.
787 (do_rpo_vn): For conservatively handling a PHI only mark
788 the backedge executable but not the block reachable.
790 2018-09-12 Richard Biener <rguenther@suse.de>
792 PR tree-optimization/87266
793 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
796 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
798 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
800 ("trunc<BFP:mode><DFP_ALL:mode>2")
801 ("trunc<DFP_ALL:mode><BFP:mode>2")
802 ("extend<BFP:mode><DFP_ALL:mode>2")
803 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
804 according to the target operand type.
806 2018-09-12 Jakub Jelinek <jakub@redhat.com>
807 Andreas Krebbel <krebbel@linux.ibm.com>
809 PR tree-optimization/86844
810 * gimple-ssa-store-merging.c
811 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
812 there are any overlapping stores in between them, make sure they are
813 also coalesced or we give up completely.
815 2018-09-12 Jakub Jelinek <jakub@redhat.com>
818 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
819 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
821 2018-09-12 Tom de Vries <tdevries@suse.de>
823 * common.opt (gdescribe-dies): Add option.
824 * dwarf2out.c (add_name_and_src_coords_attributes): Add description
825 attribute for artifical and nameless decls.
826 (dwarf2out_register_external_die): Add description attribute to
827 external reference die.
828 (add_desc_attribute): New functions.
829 (gen_subprogram_die): Add description attribute to
830 DW_TAG_call_site_parameter.
831 * tree-pretty-print.c (print_generic_expr_to_str): New function.
832 * tree-pretty-print.h (print_generic_expr_to_str): Declare.
833 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
835 (@item -gdescribe-dies): Add.
837 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
839 * tree-vrp.c (vrp_shift_undefined_p): Remove.
840 (extract_range_from_binary_expr_1: Call
841 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
842 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
845 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
847 * gimple-ssa-warn-alloca.c
848 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
849 field for ALLOCA_BOUND_*_LARGE.
851 2018-09-11 Nathan Sidwell <nathan@acm.org>
853 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
855 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
857 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
858 for clobbers. Remove obsolete comment.
860 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
862 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
863 mpxchk, mpxld and mpxst types.
864 (define_attr length_immediate): Remove all processing of mpx types.
865 (define_attr prefix_0f): Ditto.
866 (define_attr memory): Ditto.
868 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
870 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
871 (log<mode>2): Change operand 1 predicate to general_operand.
872 Extend operand 1 to XFmode and generate logxf3 insn.
873 (log10<mode>2): Change operand 1 predicate to general_operand.
874 Extend operand 1 to XFmode and generate log10xf3 insn.
875 (log2<mode>2): Change operand 1 predicate to general_operand.
876 Extend operand 1 to XFmode and generate log2xf3 insn.
877 (fyl2xp1_extend<mode>xf3_i387): Remove.
878 (log1p<mode>2): Change operand 1 predicate to general_operand.
879 Extend operand 1 to XFmode and generate log1pxf3 insn.
880 (fxtract_extend<mode>xf3_i387): Remove.
881 (logb<mode>2): Change operand 1 predicate to general_operand.
882 Extend operand 1 to XFmode and generate logbxf3 insn.
883 (ilogb<mode>2): Change operand 1 predicate to general_operand.
884 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
885 (significand<mode>2): Change operand 1 predicate to general_operand.
886 Extend operand 1 to XFmode and generate significandxf3 insn.
888 2018-09-11 Nathan Sidwell <nathan@acm.org>
890 * gcc.c (perror_with_name, pfatal_with_name): Delete.
891 (load_specs): Use fatal_error.
892 (DELETE_IF_ORDINARY, process_command): Use error.
893 (execute, run_attempt): Use fatal_error.
895 2018-09-11 Andrew Stubbs <ams@codesourcery.com>
897 * diagnostic-core.h (sorry_at): New prototype.
898 * diagnostic.c (sorry_at): New function.
900 2018-09-11 Aldy Hernandez <aldyh@redhat.com>
902 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
903 by zero as VR_UNDEFINED.
905 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
907 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
908 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
909 (<sincos>mode2): New expander.
910 (sincos_extend<mode>xf3_i387): Remove insn pattern.
911 (sincos -> sin, cos splitters): Remove splitter patterns.
912 (sincos<mode>3): Change operand 2 predicate to general_operand.
913 Extend operand 2 to XFmode and generate sincosxf3 insn.
914 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
915 Change operand 3 predicate to const1_operand.
916 (fptan_extend<mode>xf4_i387): Remove insn pattern.
917 (tanxf2): Update operands in the call to fptanxf4_i387.
918 (tan<mode>2): Change operand 1 predicate to general_operand.
919 Extend operand 1 to XFmode and generate tanxf3 insn.
920 (atan2xf3): Rename from *fpatanxf3_i387.
921 (fpatan_extend<mode>xf3_i387): Remove insn pattern.
922 (atan2xf3): Remove expander.
923 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand.
924 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
925 (atan<mode>2): Change operand 1 predicate to general_operand.
926 Extend operand 1 to XFmode and generate atanxf3 insn.
928 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
930 * config/i386/i386.md (x87/SSE constant load splitter): Use
931 memory_operand instead of nonimmediate_operand for input operand
934 2018-09-09 Uros Bizjak <ubizjak@gmail.com>
936 * config/i386/i386.md (float partial SSE register stall splitter): Move
937 splitter near its instruction pattern.
938 (float_extend partial SSE register stall splitter): Ditto.
939 (float_truncate partial SSE register stall splitter): Ditto.
941 2018-09-09 Hans-Peter Nilsson <hp@bitrange.com>
944 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
945 to speculation_safe_value_not_needed.
948 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
949 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
950 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
951 leaf_function_p, instead use has_hard_reg_initial_val.
953 2018-09-09 Nathan Sidwell <nathan@acm.org>
955 * gcc.h (pfatal_with_name): Don't declare here.
956 * gcc.c (pfatal_with_name): Make static.
958 2018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com>
960 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
963 2018-09-08 John David Anglin <danglin@gcc.gnu.org>
965 PR rtl-optimization/85458
966 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
967 priority hook to reduce the priority of EXPR.
969 2018-09-07 Uros Bizjak <ubizjak@gmail.com>
971 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
972 DImode for x87 on 32bit targets. Conditionally disable x87 modes
973 with X87_ENABLE_FLOAT. Remove preparation code.
974 (*float<SWI48:mode><MODEF:mode>2): Rename from
975 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed
976 math using "enabled" attribute.
977 (*floatdi<MODEF:mode>2_i387): Rename from
978 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and
979 enable for 32bit targets only.
980 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
982 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
983 as operand 1 predicate. Rewrite as define_insn_and_split.
984 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
986 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
988 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
989 to fallthru to FLOAT case.
991 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
994 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
995 around folding of vec_sl to handle out of range shift values.
997 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
999 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
1000 Update callers to gen_fix_trunc<mode>_i387_fisttp
1001 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
1002 nonimmediate_operand.
1003 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
1004 and corresponding splitters.
1005 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
1006 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
1007 (fix_truncdi_i387_with_temp): Remove insn pattern
1008 and corresponding splitters.
1009 (fix_trunc<mode>_i387): Change operand 0 predicate to
1010 nonimmediate_operand.
1011 (fix_trunc<mode>_i387_with_temp): Remove insn pattern
1012 and corresponding splitters.
1013 (*fistdi2_1): Remove.
1015 (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
1016 (lrintxfdi2): Remove expander. Reimplement as define_insn.
1017 (*fist<mode>2_1): Remove.
1018 (fist<mode>2): Ditto.
1019 (fist<mode>2_with_temp): Remove insn pattern and corresponding
1021 (lrintxf<mode>2): Remove expander. Reimplement as define_insn.
1022 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
1023 (fistdi2_<rounding>): Change operand 0 predicate to
1024 nonimmediate_operand.
1025 (fistdi2_<rounding>_with_temp): Remove insn pattern
1026 and corresponding splitters.
1027 (fist<mode>2_<rounding>): Change operand 0 predicate to
1028 nonimmediate_operand.
1029 (fist<mode>2_<rounding>_with_temp): Remove insn pattern
1030 and corresponding splitters.
1032 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
1034 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
1036 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
1039 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
1041 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
1042 early gimple folding of vec_splat().
1043 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
1044 * gimple-fold.h: Add an extern define for tree_vec_extract().
1046 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
1048 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
1049 wrappers around TREE_TYPE comparisons.
1051 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
1054 * config/s390/predicates.md: Add nonsym_memory_operand.
1055 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
1056 contains a SYMBOL_REF, load it into an intermediate pseudo.
1057 (s390_emit_compare_and_swap): Legitimize operand.
1058 * config/s390/s390.md: Use the new nonsym_memory_operand
1059 with UNSPECV_CAS patterns.
1061 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
1064 * config/s390/s390-passes.def: New file.
1065 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
1067 (make_pass_s390_early_mach): Add declaration.
1068 * config/s390/s390.c (make_pass_s390_early_mach):
1069 (s390_option_override): Remove dynamic registration.
1070 * config/s390/t-s390: Add s390-passes.def.
1072 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
1074 * config/s390/s390.c (s390_decompose_constant_pool_ref):
1075 Remove UNSPEC_LTREL_BASE check.
1076 (annotate_constant_pool_refs): Likewise.
1077 (find_constant_pool_ref): Likewise.
1078 (find_ltrel_base): Removed.
1079 (replace_ltrel_base): Removed.
1080 (s390_mainpool_finish): Remove replace_ltrel_base call.
1081 (s390_chunkify_start): Remove pending LTREL_BASE logic.
1082 (s390_chunkify_finish): Remove replace_ltrel_base call.
1083 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
1085 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
1088 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
1089 to speculation_safe_value_not_needed.
1091 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
1092 Bernd Schmidt <bernds_cb1@t-online.de>
1094 * config/nvptx/nvptx-opts.h: New file.
1095 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
1096 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
1098 (TARGET_SM35): New macro.
1099 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
1101 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
1103 (misa=): New option.
1104 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
1106 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
1108 * config/i386/i386.md (truncdfsf2): Remove expander.
1109 (truncdfsf2_with_temp): Ditto.
1110 (truncxf<mode>2): Ditto.
1111 (*truncdfsf_fast_mixed): Remove insn pattern.
1112 (*truncdfsf_fast_i387): Ditto.
1113 (*truncdfsf_mixed): Ditto.
1114 (*truncdfsf_i387): Ditto.
1115 (*truncdfsf2_i387_1): Ditto.
1116 (*truncxfsf2_mixed): Ditto.
1117 (*truncxfdf2_mixed): Ditto.
1118 (*truncxf<mode>2_i387_noop): Ditto. Update callers
1119 to call gen_truncxf<mode>2 instead.
1120 (*truncxf<mode>2_i387): Remove.
1121 (reg->reg splitters): Remove splitter pattern.
1122 (reg->mem splitters): Ditto.
1124 (truncdfsf2): New insn pattern.
1125 (truncxf<mode>2): Ditto.
1127 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1129 * tree-ssa-math-opts.c (is_mult_by): New function.
1130 (is_square_of): Use the above.
1131 (optimize_recip_sqrt): New function.
1132 (pass_cse_reciprocals::execute): Use the above.
1134 2018-09-05 Richard Biener <rguenther@suse.de>
1137 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
1138 to zero-init the emplaced vec.
1140 2018-09-05 Martin Liska <mliska@suse.cz>
1142 PR tree-optimization/87205
1143 * tree-switch-conversion.c (pass_lower_switch::execute):
1144 Group cases for switch statements.
1146 2018-09-05 Richard Biener <rguenther@suse.de>
1148 PR tree-optimization/87217
1149 * tree-ssa-sccvn.c (vuse_valueize): New.
1150 (vn_reference_lookup_pieces): Use it.
1151 (vn_reference_lookup): Likewise.
1153 2018-09-05 Nathan Sidwell <nathan@acm.org>
1156 * stor-layout.c (place_field): Scan forwards to check last
1157 bitfield when ms_bitfield_placement is in effect.
1159 2018-09-05 Richard Biener <rguenther@suse.de>
1162 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
1165 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
1166 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
1168 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
1169 * config.gcc (extra_objs): Build it.
1170 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
1172 * config/aarch64/aarch64-passes.def
1173 (pass_tag_collision_avoidance): New pass.
1174 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
1175 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
1176 (aarch64_classify_address): Remove static qualifier.
1177 (aarch64_address_info, aarch64_address_type): Move to...
1178 * config/aarch64/aarch64-protos.h: ... here.
1179 (make_pass_tag_collision_avoidance): New function.
1180 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
1183 2018-09-05 Martin Liska <mliska@suse.cz>
1185 * doc/gcov.texi: Update documentation of humar
1187 * gcov.c (format_count): Print one decimal place, it provides
1188 more fine number of situations like '1G' vs. '1.4G'.
1190 2018-09-05 Martin Liska <mliska@suse.cz>
1193 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
1194 * optc-gen.awk: Allow 'Var' for Deprecated options in order
1195 to generate a MASK value.
1197 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
1200 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
1201 if frame pointer isn't used.
1202 (compute_frame_pointer_to_fb_displacement): Likewise.
1204 2018-09-04 Jakub Jelinek <jakub@redhat.com>
1207 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
1208 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
1209 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
1210 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
1211 and OPTION_MASK_ISA_XSAVEC_UNSET.
1213 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
1215 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
1216 XOR operations in NAND case.
1218 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
1220 * wide-int-range.cc (wide_int_range_convert): New.
1221 * wide-int-range.h (wide_int_range_convert): New.
1222 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
1223 code into wide_int_range_convert.
1224 (extract_range_into_wide_ints): Do not munge anti range constants
1225 into the entire domain. Just return the range back.
1227 2018-09-04 Martin Liska <mliska@suse.cz>
1229 * genmatch.c (output_line_directive): Add new argument
1231 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
1233 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
1235 * doc/invoke.texi (Option Summary): Add whitespace.
1237 * doc/invoke.texi (Option Summary): Add -Waligned-new.
1239 2018-09-04 Richard Biener <rguenther@suse.de>
1241 PR tree-optimization/87211
1242 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
1243 backedge value we're supposed to treat as VARYING also number
1244 the PHI to VARYING in case it got a different value-number already.
1246 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
1248 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
1249 (extract_range_from_binary_expr_1): Do not call
1250 vrp_can_optimize_bit_op.
1251 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
1253 (wide_int_range_get_mask_and_bounds): New.
1254 (wide_int_range_optimize_bit_op): New.
1255 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
1256 (wide_int_range_bit_and): Same.
1257 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
1258 (wide_int_range_optimize_bit_op): New.
1259 (wide_int_range_get_mask_and_bounds): New.
1261 2018-09-04 Richard Biener <rguenther@suse.de>
1263 PR tree-optimization/87176
1264 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
1265 variable. When value-numbering a virtual PHI node make sure
1266 to not value-number to the backedge value.
1268 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
1270 * doc/extend.texi (Long Long, Hex Floats): Document support for
1271 long long and hex floats in more recent versions of ISO C++.
1273 2018-09-03 Richard Biener <rguenther@suse.de>
1275 PR tree-optimization/87177
1276 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
1279 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
1281 * bb-reorder.c (edge_order): Convert to C-qsort-style
1282 tri-state comparator.
1283 (reorder_basic_blocks_simple): Change std::stable_sort to gcc_stablesort.
1285 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
1287 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
1288 tri-state comparator.
1289 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
1291 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
1293 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
1294 (mergesort): ... here as maximum count for using netsort.
1295 (gcc_qsort): Set nlim to 3 if stable sort is requested.
1296 (gcc_stablesort): New.
1297 * system.h (gcc_stablesort): Declare.
1299 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
1301 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
1302 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
1303 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
1305 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
1307 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
1308 lxsdx and stxsdx alternatives.
1309 (*mov<mode>_hardfloat64): Ditto.
1310 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
1312 2018-09-03 Richard Biener <rguenther@suse.de>
1314 PR tree-optimization/87200
1315 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
1318 2018-09-03 Martin Liska <mliska@suse.cz>
1320 PR tree-optimization/87201
1321 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
1322 Fix parenthesis in an expression.
1324 2018-09-03 Richard Biener <rguenther@suse.de>
1326 PR tree-optimization/87197
1327 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
1328 visited. CSE the VN_INFO hashtable lookup.
1330 PR tree-optimization/87169
1331 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
1332 iterating make sure there's no extra backedges from irreducible
1333 regions feeding the header. Mark the destination block
1336 2018-09-03 Martin Liska <mliska@suse.cz>
1339 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
1340 * common/common-targhooks.c (default_get_valid_option_values):
1342 * common/common-targhooks.h (default_get_valid_option_values):
1344 * common/config/i386/i386-common.c: Move processor_target_table
1346 (ix86_get_valid_option_values): New function.
1347 (TARGET_GET_VALID_OPTION_VALUES): New macro.
1348 * config/i386/i386.c (struct ptt): Move to i386-common.c.
1349 (PTA_*): Move all defined masks into i386-common.c.
1350 (ix86_function_specific_restore): Use new processor_cost_table.
1351 * config/i386/i386.h (struct ptt): Moved from i386.c.
1352 (struct pta): Likewise.
1353 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
1354 * doc/tm.texi.in: Likewise.
1355 * opt-suggestions.c (option_proposer::suggest_option):
1356 Pass prefix to build_option_suggestions.
1357 (option_proposer::get_completions): Likewise.
1358 (option_proposer::build_option_suggestions): Use the new target
1360 * opts.c (struct option_help_tuple): New struct.
1361 (print_filtered_help): Use the new target hook.
1363 2018-09-03 Martin Liska <mliska@suse.cz>
1366 * predict.c (set_even_probabilities): Add likely_edges
1367 argument and handle cases where we have precisely one
1369 (combine_predictions_for_bb): Catch also likely_edges.
1370 (tree_predict_by_opcode): Handle gswitch statements.
1371 * tree-cfg.h (find_case_label_for_value): New declaration.
1372 (find_taken_edge_switch_expr): Likewise.
1373 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
1374 Find pivot in decision tree based on probabily, not by number of
1377 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
1379 * doc/standards.texi (Standards): Update Objective-C reference.
1381 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
1383 * doc/install.texi (Prerequisites): Update link for MPC.
1385 2018-09-01 Michael Matz <matz@suse.de>
1387 PR tree-optimization/87074
1388 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
1389 PHIs for outer-loop uses.
1391 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
1393 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
1394 * doc/invoke.texi (C Dialect Options): Ditto.
1396 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
1398 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
1400 2018-08-31 Richard Biener <rguenther@suse.de>
1402 PR tree-optimization/87168
1403 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
1404 (rpo_elim::eliminate_avail): When OP was not visited it must
1407 2018-08-31 David Malcolm <dmalcolm@redhat.com>
1409 * tree-vrp.c (copy_value_range): Convert param "from" from
1410 "value_range *" to "const value_range *".
1411 (range_is_null): Likewise for param "vr".
1412 (range_int_cst_p): Likewise.
1413 (range_int_cst_singleton_p): Likewise.
1414 (symbolic_range_p): Likewise.
1415 (value_ranges_intersect_p): Likewise for both params.
1416 (value_range_nonnegative_p): Likewise for param "vr".
1417 (value_range_constant_singleton): Likewise.
1418 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
1419 (extract_range_into_wide_ints): Likewise for param "vr".
1420 (extract_range_from_multiplicative_op): Likewise for params "vr0"
1422 (vrp_can_optimize_bit_op): Likewise.
1423 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
1425 (extract_range_from_unary_expr): Likewise.
1426 (debug_value_range): Likewise for param "vr".
1427 (value_range::dump): Add "const" qualifier.
1428 (vrp_prop::check_array_ref): Convert local "vr" from
1429 "value_range *" to "const value_range *".
1430 (vrp_prop::check_mem_ref): Likewise.
1431 (vrp_prop::visit_stmt): Likewise for local "old_vr".
1432 (vrp_intersect_ranges_1): Likewise for param "vr_1".
1433 (vrp_intersect_ranges): Likewise.
1434 (simplify_stmt_for_jump_threading): Likewise for local "vr".
1435 (vrp_prop::vrp_finalize): Likewise.
1436 * tree-vrp.h (value_range::dump): Add "const" qualifier.
1437 (vrp_intersect_ranges): Add "const" qualifier to params as above.
1438 (extract_range_from_unary_expr): Likewise.
1439 (value_range_constant_singleton): Likewise.
1440 (symbolic_range_p): Likewise.
1441 (copy_value_range): Likewise.
1442 (extract_range_from_binary_expr_1): Likewise.
1443 (range_int_cst_p): Likewise.
1444 (vrp_set_zero_nonzero_bits): Likewise.
1445 (range_int_cst_singleton_p): Likewise.
1447 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
1449 * config/aarch64/arm_neon.h (vabsd_s64): New.
1450 (vnegd_s64): Likewise.
1452 2018-08-31 Martin Jambor <mjambor@suse.cz>
1454 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
1456 2018-08-31 Martin Liska <mliska@suse.cz>
1458 * ipa-icf.c (sem_item::add_type): Use
1459 sem_item::m_type_hash_cache.
1460 * ipa-icf.h: Move the cache from sem_item_optimizer
1463 2018-08-31 Nathan Sidwell <nathan@acm.org>
1465 * doc/extend.texi (Backwards Compatibility): Remove implicit
1466 extern C leeway of () being (...).
1468 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1470 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
1472 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
1476 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
1478 2018-08-31 Jakub Jelinek <jakub@redhat.com>
1481 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
1482 gen_int_mode. Formatting fixes.
1484 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
1486 * target.def (custom_function_descriptors): Improve documentation.
1487 * doc/tm.texi.in (Trampolines): Expand discussion of function
1488 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
1489 beginning of the section.
1490 * doc/tm.texi: Regenerated.
1492 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
1494 * cfg.h (class auto_edge_flag): Spell out the template-id of the
1495 base class in the initializer list. This is a workaround for
1496 building with older GCC.
1497 (class auto_bb_flag): Likewise.
1499 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
1501 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
1502 (altivec_vcmpequ<VI_char>_p): Remove star.
1503 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
1505 (expand_strncmp_vec_sequence): New function.
1506 (emit_final_str_compare_vec): New function.
1507 (expand_strn_compare): Add support for vector strncmp.
1508 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
1509 length specification to bytes.
1510 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
1511 (vcmpnezb_p): New pattern.
1512 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
1513 for option -mstring-compare-inline-limit.
1515 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
1517 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
1518 (PTA_SKYLAKE): Add PTA_AES.
1519 (PTA_GOLDMONT): Likewise.
1521 2018-08-29 Jan Hubicka <jh@suse.cz>
1524 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
1525 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
1527 2018-08-29 Jan Hubicka <jh@suse.cz>
1529 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
1531 (hash_tree): Do not visit TYPE_STUB_DECL.
1532 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
1533 stream TYPE_STUB_DECL.
1534 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
1535 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
1536 after free_lang_data.
1537 (type_in_anonymous_namespace_p): Likewise.
1539 2018-08-29 Jan Hubicka <jh@suse.cz>
1541 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
1542 comment that it has to be even number.
1543 (class sreal): Change m_sig type to int32_t.
1544 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
1545 int64_t for temporary calculations.
1546 (sreal_verify_basics): Drop one bit from minimum and maximum.
1548 2018-08-30 Richard Biener <rguenther@suse.de>
1550 PR tree-optimization/87147
1551 * tree-ssa-sccvn.c (SSA_VISITED): New function.
1552 (visit_phi): When the degenerate result is from the backedge and
1553 we didn't visit its definition yet drop to VARYING.
1554 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
1556 2018-08-29 Jan Hubicka <jh@suse.cz>
1558 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
1560 (hash_tree): Likewise.
1562 2018-08-29 Jan Hubicka <jh@suse.cz>
1564 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
1565 and TYPE_NEXT_REF_TO.
1567 2018-08-29 Jan Hubicka <jh@suse.cz>
1569 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
1570 comment that it has to be even number.
1571 (class sreal): Change m_sig type to int32_t.
1572 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
1573 int64_t for temporary calculations.
1574 (sreal_verify_basics): Drop one bit from minimum and maximum.
1576 2018-08-30 Tamar Christina <tamar.christina@arm.com>
1578 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
1580 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
1583 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
1584 if to_add is negative.
1586 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
1589 * builtins.c (c_strlen): Improve range checks.
1591 2018-08-29 Martin Sebor <msebor@redhat.com>
1592 Jeff Law <law@redhat.com>
1594 PR tree-optimization/86714
1595 PR tree-optimization/86711
1596 * builtins.c (c_strlen): Add arguments to call to string_constant.
1597 * expr.c (string_constant): Add argument. Detect missing nul
1598 terminator and outermost declaration it's missing in.
1599 * expr.h (string_constant): Add argument.
1600 * fold-const.c (read_from_constant_string): Add arguments to call to
1602 (c_getstr): Likewise.
1603 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
1605 * tree-ssa-strlen.c (get_stridx): Likewise.
1607 2018-08-29 Jan Hubicka <jh@suse.cz>
1609 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
1610 Do not stream DECL_VINDEX.
1611 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
1612 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
1613 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
1615 2018-08-29 Richard Biener <rguenther@suse.de>
1617 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
1618 virtual operands that are not default defs to honor region
1620 (rpo_vn_valueize): Remove ineffective code here.
1622 2018-08-29 Richard Biener <rguenther@suse.de>
1624 PR tree-optimization/87132
1625 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
1626 when skipping defs reachable over backedges.
1628 2018-08-29 Richard Biener <rguenther@suse.de>
1630 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
1631 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
1632 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
1633 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
1634 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
1635 (vn_reference_lookup_3): Remove use of const_parms.
1636 (free_rpo_vn): Do not free const_parms.
1637 (do_rpo_vn): Do not call init_const_parms.
1638 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
1639 SSA_NAME_POINTS_TO_READONLY_MEMORY.
1640 (call_may_clobber_ref_p_1): Likewise.
1642 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
1645 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
1647 (-ftree-scev-cprop): Document.
1649 2018-08-29 Jan Hubicka <jh@suse.cz>
1651 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
1653 (sreal constructor): Update.
1654 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
1655 sreal:operator/): Update.
1657 2018-08-29 Martin Liska <mliska@suse.cz>
1659 * tree-switch-conversion.c (switch_conversion::expand):
1660 Strenghten assumption about gswitch statements.
1662 2018-08-29 Richard Biener <rguenther@suse.de>
1664 PR tree-optimization/87117
1665 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
1666 re-value-number released SSA VDEFs.
1668 2018-08-29 Richard Biener <rguenther@suse.de>
1670 PR tree-optimization/87126
1671 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
1673 2018-08-28 Jim Wilson <jimw@sifive.com>
1675 * config/riscv/pic.md: Rewrite.
1676 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
1678 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
1679 (SOFTF, default_load, softload, softstore): New.
1681 2018-08-28 Jeff Law <law@redhat.com>
1683 * fold-const.c (fold_binary_loc): Remove recently added assert.
1685 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
1687 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
1688 to OP parmeter of generated function.
1690 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
1692 PR tree-optimization/87009
1693 * match.pd: Add boolean optimizations.
1695 2018-08-28 Martin Sebor <msebor@redhat.com>
1698 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
1699 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
1700 (pass_walloca::gate): Use it.
1701 (alloca_call_type): Same.
1702 (pass_walloca::execute): Same.
1703 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
1705 2018-08-28 David Malcolm <dmalcolm@redhat.com>
1707 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
1708 GCC_VERSION for usage of "__gcc_dump_printf__" format from
1711 2018-08-28 Richard Biener <rguenther@suse.de>
1713 PR tree-optimization/87124
1714 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
1715 constants before looking up avail.
1717 2018-08-28 Jakub Jelinek <jakub@redhat.com>
1720 * calls.c (maybe_warn_nonstring_arg): Punt early if
1721 warn_stringop_overflow is zero. Don't call get_range_strlen
1722 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
1723 Swap comparison operands to have constants on rhs. Only use
1724 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
1725 increment lenrng[0].
1727 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
1729 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
1730 use of tree_to_shwi. Remove duplicated test for the size being
1731 a whole number of bytes.
1733 2018-08-28 Richard Biener <rguenther@suse.de>
1735 PR tree-optimization/87117
1736 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
1737 Handle removed stmt without LHS (GIMPLE_NOP).
1739 2018-08-28 Richard Biener <rguenther@suse.de>
1741 PR tree-optimization/87117
1742 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
1743 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
1745 2018-08-28 Richard Biener <rguenther@suse.de>
1747 PR tree-optimization/87117
1748 * tree-ssa-pre.c (compute_avail): Do not make expressions
1749 with predicated values available.
1750 (get_expr_value_id): Assert we do not run into predicated value
1753 2018-08-28 Richard Biener <rguenther@suse.de>
1755 PR tree-optimization/87117
1756 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
1757 get virtual operands.
1758 (get_expr_operands): Handle STRING_CST like other decls.
1760 2018-08-28 Martin Liska <mliska@suse.cz>
1762 * tree.h: Update documentation of fndecl_built_in_p
1766 2018-08-27 Jeff Law <law@redhat.com>
1767 PR tree-optimization/87110
1768 * tree-ssa-dse.c (compute_trims): Handle non-constant
1771 2018-08-27 Martin Sebor <msebor@redhat.com>
1773 PR tree-optimization/86914
1774 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
1776 2018-08-27 Martin Sebor <msebor@redhat.com>
1778 PR tree-optimization/87112
1779 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
1780 the type of the bound argument.
1782 2018-08-27 Jeff Law <law@redhat.com>
1784 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
1785 type does not have a TYPE_SIZE_UNIT.
1787 2018-08-27 Steve Ellcey <sellcey@cavium.com>
1789 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
1790 with include of backend.h.
1792 2018-08-27 Richard Biener <rguenther@suse.de>
1794 PR tree-optimization/86927
1795 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
1796 use const cond reduction code.
1798 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
1800 PR tree-optimization/85758
1801 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
1803 2018-08-27 David Malcolm <dmalcolm@redhat.com>
1806 * diagnostic-show-locus.c (class layout_range): Update for
1807 conversion of show_caret_p to a tri-state.
1808 (layout_range::layout_range): Likewise.
1809 (make_range): Likewise.
1810 (layout::maybe_add_location_range): Likewise.
1811 (layout::should_print_annotation_line_p): Don't show annotation
1812 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
1813 (layout::get_state_at_point): Update for conversion of
1814 show_caret_p to a tri-state. Bail out early for
1815 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
1816 underlining or source colorization.
1817 (gcc_rich_location::add_location_if_nearby): Update for conversion
1818 of show_caret_p to a tri-state.
1819 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
1820 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
1822 (selftest::test_one_liner_labels): Likewise.
1823 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
1824 conversion of show_caret_p to a tri-state.
1825 * pretty-print.c (text_info::set_location): Likewise.
1826 * pretty-print.h (text_info::set_location): Likewise.
1827 * substring-locations.c (format_warning_n_va): Likewise.
1828 * tree-diagnostic.c (default_tree_printer): Likewise.
1829 * tree-pretty-print.c (newline_and_indent): Likewise.
1831 2018-08-27 David Malcolm <dmalcolm@redhat.com>
1834 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
1835 line above for line-insertion fix-it hints.
1836 (selftest::test_fixit_insert_containing_newline): Update the
1837 expected results, and add a test with line-numbering enabled.
1839 2018-08-27 Martin Liska <mliska@suse.cz>
1842 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
1843 params with DECL_HAS_VALUE_EXPR_P.
1845 2018-08-27 Martin Liska <mliska@suse.cz>
1847 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
1848 selected expansion strategy.
1850 2018-08-27 Martin Liska <mliska@suse.cz>
1852 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
1853 * builtins.c (is_builtin_fn): Likewise.
1854 * attribs.c (diag_attr_exclusions): Use new function
1855 fndecl_built_in_p and remove check for FUNCTION_DECL if
1857 (builtin_mathfn_code): Likewise.
1858 (fold_builtin_expect): Likewise.
1859 (fold_call_expr): Likewise.
1860 (fold_builtin_call_array): Likewise.
1861 (fold_call_stmt): Likewise.
1862 (set_builtin_user_assembler_name): Likewise.
1863 (is_simple_builtin): Likewise.
1864 * calls.c (gimple_alloca_call_p): Likewise.
1865 (maybe_warn_nonstring_arg): Likewise.
1866 * cfgexpand.c (expand_call_stmt): Likewise.
1867 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
1868 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
1869 (cgraph_node::verify_node): Likewise.
1870 * cgraphclones.c (build_function_decl_skip_args): Likewise.
1871 (cgraph_node::create_clone): Likewise.
1872 * config/arm/arm.c (arm_insert_attributes): Likewise.
1873 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
1874 * dse.c (scan_insn): Likewise.
1875 * expr.c (expand_expr_real_1): Likewise.
1876 * fold-const.c (operand_equal_p): Likewise.
1877 (fold_binary_loc): Likewise.
1878 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
1879 * gimple-low.c (lower_stmt): Likewise.
1880 * gimple-pretty-print.c (dump_gimple_call): Likewise.
1881 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Likewise.
1882 * gimple.c (gimple_build_call_from_tree): Likewise.
1883 (gimple_call_builtin_p): Likewise.
1884 (gimple_call_combined_fn): Likewise.
1885 * gimplify.c (gimplify_call_expr): Likewise.
1886 (gimple_boolify): Likewise.
1887 (gimplify_modify_expr): Likewise.
1888 (gimplify_addr_expr): Likewise.
1889 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
1890 * ipa-cp.c (determine_versionability): Likewise.
1891 * ipa-fnsummary.c (compute_fn_summary): Likewise.
1892 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
1893 * ipa-split.c (visit_bb): Likewise.
1894 (split_function): Likewise.
1895 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1896 * lto-cgraph.c (input_node): Likewise.
1897 * lto-streamer-out.c (write_symbol): Likewise.
1898 * omp-low.c (setjmp_or_longjmp_p): Likewise.
1899 (lower_omp_1): Likewise.
1900 * predict.c (strip_predict_hints): Likewise.
1901 * print-tree.c (print_node): Likewise.
1902 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
1903 * trans-mem.c (is_tm_irrevocable): Likewise.
1904 (is_tm_load): Likewise.
1905 (is_tm_simple_load): Likewise.
1906 (is_tm_store): Likewise.
1907 (is_tm_simple_store): Likewise.
1908 (is_tm_abort): Likewise.
1909 (tm_region_init_1): Likewise.
1910 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
1911 * tree-cfg.c (verify_gimple_call): Likewise.
1912 (move_stmt_r): Likewise.
1913 (stmt_can_terminate_bb_p): Likewise.
1914 * tree-eh.c (lower_eh_constructs_2): Likewise.
1915 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
1916 * tree-inline.c (remap_gimple_stmt): Likewise.
1917 (copy_bb): Likewise.
1918 (estimate_num_insns): Likewise.
1919 (fold_marked_statements): Likewise.
1920 * tree-sra.c (scan_function): Likewise.
1921 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
1922 (optimize_stack_restore): Likewise.
1923 (pass_fold_builtins::execute): Likewise.
1924 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
1925 (mark_all_reaching_defs_necessary_1): Likewise.
1926 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
1927 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
1928 (pass_forwprop::execute): Likewise.
1929 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1930 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
1931 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
1932 * tree-ssa-strlen.c (get_string_length): Likewise.
1933 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
1934 (find_func_aliases_for_call): Likewise.
1935 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
1936 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
1937 * tree-tailcall.c (find_tail_calls): Likewise.
1938 * tree.c (need_assembler_name_p): Likewise.
1939 (free_lang_data_in_decl): Likewise.
1940 (get_call_combined_fn): Likewise.
1941 * ubsan.c (is_ubsan_builtin_p): Likewise.
1942 * varasm.c (incorporeal_function_p): Likewise.
1943 * tree.h (DECL_BUILT_IN): Remove and replace with
1945 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
1946 (fndecl_built_in_p): New.
1948 2018-08-27 Martin Liska <mliska@suse.cz>
1950 PR tree-optimization/86847
1951 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
1952 Dump also subtree probability.
1953 (switch_decision_tree::do_jump_if_equal): New function.
1954 (switch_decision_tree::emit_case_nodes): Handle special
1955 situations in balanced tree that can be emitted much simpler.
1956 Fix calculation of probabilities that happen in tree expansion.
1957 * tree-switch-conversion.h (struct cluster): Add
1959 (struct simple_cluster): Likewise.
1960 (struct case_tree_node): Add new function has_child.
1961 (do_jump_if_equal): New.
1963 2018-08-27 Martin Liska <mliska@suse.cz>
1965 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
1966 Add new argument to bit_test_cluster constructor.
1967 (bit_test_cluster::emit): Set bits really number of values
1969 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
1970 probability argument.
1971 * tree-switch-conversion.h (struct bit_test_cluster):
1972 Add m_handles_entire_switch.
1974 2018-08-27 Martin Liska <mliska@suse.cz>
1976 PR tree-optimization/86702
1977 * tree-switch-conversion.c (jump_table_cluster::emit):
1978 Make probabilities even for values in jump table
1979 according to number of cases handled.
1980 (switch_decision_tree::compute_cases_per_edge): Pass
1981 argument to reset_out_edges_aux function.
1982 (switch_decision_tree::analyze_switch_statement): Likewise.
1983 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
1986 2018-08-27 Martin Liska <mliska@suse.cz>
1988 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
1989 cfun argument explicitly.
1990 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
1991 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
1992 function gimple_switch_default_bb.
1993 (convert_switch_statements):
1995 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
1996 * stmt.c (label_to_block_fn): Use label_to_block and pass
1997 cfun argument explicitly and use gimple_switch_label_bb.
1998 (expand_case): Likewise.
1999 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
2000 cfun argument explicitly. Likewise.
2001 (make_edges_bb): Likewise.
2002 (make_cond_expr_edges): Likewise.
2003 (get_cases_for_edge): Likewise.
2004 (make_gimple_switch_edges): Likewise.
2005 (label_to_block_fn): Likewise.
2006 (label_to_block): Likewise.
2007 (make_goto_expr_edges): Likewise.
2008 (make_gimple_asm_edges): Likewise.
2009 (main_block_label): Likewise.
2010 (group_case_labels_stmt): Likewise.
2011 (find_taken_edge_computed_goto): Likewise.
2012 (find_taken_edge_switch_expr): Likewise.
2013 (gimple_verify_flow_info): Likewise.
2014 (gimple_redirect_edge_and_branch): Likewise.
2015 (gimple_switch_label_bb): New function.
2016 (gimple_switch_default_bb): Likewise.
2017 (gimple_switch_edge): Likewise.
2018 (gimple_switch_default_edge): Likewise.
2019 * tree-cfg.h (label_to_block_fn): Remove and replace ...
2020 (label_to_block): ... with this.
2021 (gimple_switch_label_bb): New.
2022 (gimple_switch_default_bb): Likewise.
2023 (gimple_switch_edge): Likewise.
2024 (gimple_switch_default_edge): Likewise.
2025 * tree-cfgcleanup.c (convert_single_case_switch): Use
2026 new gimple functions and pass new argument to label_to_block.
2027 (cleanup_control_flow_bb):
2028 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
2029 cfun argument explicitly.
2030 (make_eh_edges): Likewise.
2031 (redirect_eh_dispatch_edge): Likewise.
2032 (lower_resx): Likewise.
2033 (lower_eh_dispatch): Likewise.
2034 (maybe_remove_unreachable_handlers): Likewise.
2035 (unsplit_eh): Likewise.
2036 (cleanup_empty_eh): Likewise.
2037 (verify_eh_edges): Likewise.
2038 (verify_eh_dispatch_edge): Likewise.
2039 * tree-ssa-dom.c (record_edge_info): Likewise.
2040 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
2041 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
2042 (thread_through_normal_block): Likewise.
2043 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
2044 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
2045 * tree-switch-conversion.c (switch_conversion::collect): Use new
2047 (switch_conversion::check_final_bb): Likewise.
2048 (switch_conversion::gather_default_values): Pass new argument
2050 (switch_conversion::build_constructors): Likewise.
2051 (switch_decision_tree::compute_cases_per_edge): Use new
2052 gimple_switch_edge function.
2053 (switch_decision_tree::analyze_switch_statement): Pass new argument
2055 (switch_decision_tree::try_switch_expansion): Use
2056 gimple_switch_default_edge.
2057 * tree-vrp.c (find_switch_asserts): Pass new argument
2059 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
2060 (vr_values::simplify_switch_using_ranges): Likewise.
2062 2018-08-27 Richard Biener <rguenther@suse.de>
2064 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
2065 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
2067 * tree-ssa-sccvn.h (struct vn_pval): New structure.
2068 (struct vn_nary_op_s): Add unwind_to member. Add
2069 predicated_values flag and put result into a union together
2070 with a linked list of vn_pval.
2071 (struct vn_ssa_aux): Add name member to make maintaining
2072 a map of SSA name to vn_ssa_aux possible. Remove no longer
2073 needed info, dfsnum, low, visited, on_sccstack, use_processed
2074 and range_info_anti_range_p members.
2075 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
2076 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
2078 (vn_valueize): New global.
2079 (vn_context_bb): Likewise.
2080 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
2081 VN_INFO_PTR_INFO): Remove.
2082 * tree-ssa-sccvn.c: ... (rewrite)
2083 (pass_fre::execute): For -O2+ initialize loops and run
2084 RPO VN in optimistic mode (iterating). For -O1 and -Og
2085 run RPO VN in non-optimistic mode.
2086 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
2087 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
2088 * doc/invoke.texi (sccvn-max-scc-size): Remove.
2089 (rpo-vn-max-loop-depth): Document.
2090 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
2091 when valuezing the VUSE signals we walked out of the region.
2092 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
2093 (phi_translate): Set VN context block to use for availability
2095 (compute_avail): Likewise.
2096 (pre_valueize): New function.
2097 (pass_pre::execute): Adjust to the RPO VN API.
2099 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
2100 (propagate_constants_for_unrolling): Remove.
2101 (tree_unroll_loops_completely): Perform value-numbering
2102 on the unrolled bodies loop parent.
2104 2018-08-27 Richard Biener <rguenther@suse.de>
2106 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
2107 for partial antic compute.
2109 2018-08-27 Jakub Jelinek <jakub@redhat.com>
2111 PR rtl-optimization/87065
2112 * combine.c (simplify_if_then_else): Formatting fix.
2113 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
2115 (known_cond): Don't return const_true_rtx for vector modes. Use
2116 CONST0_RTX instead of const0_rtx. Formatting fixes.
2118 2018-08-27 Martin Liska <mliska@suse.cz>
2120 PR gcov-profile/87069
2121 * gcov.c (process_file): Record files already processed
2122 and warn about a file being processed multiple times.
2124 2018-08-27 Martin Liska <mliska@suse.cz>
2127 * config/aarch64/aarch64.c (aarch64_override_options_internal):
2128 Set default values for x_aarch64_*_string strings.
2129 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
2130 prefix. For -mabi do not print '=ABI' in help and use
2131 <option_value> format for -msve-vector-bits and -moverride
2134 2018-08-26 Jeff Law <law@redhat.com>
2136 * config/mips/frame-header-opt.c: Include "backend.h" rather than
2139 2018-08-26 Marek Polacek <polacek@redhat.com>
2141 PR c++/87029, Implement -Wredundant-move.
2142 * doc/invoke.texi: Document -Wredundant-move.
2144 2018-08-25 Martin Sebor <msebor@redhat.com>
2146 PR tree-optimization/87059
2147 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
2148 to the same type as the other.
2149 * fold-const.c (fold_binary_loc): Assert expectation.
2151 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
2153 * config/darwin.c (machopic_legitimize_pic_address): Clean up
2154 extraneous parentheses, dead code section and formatting.
2156 2018-08-24 David Malcolm <dmalcolm@redhat.com>
2159 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
2160 wide enough for jumps in the line-numbering to be visible.
2161 (layout::print_gap_in_line_numbering): New member function.
2162 (layout::calculate_line_spans): When using line numbering, merge
2163 line spans that are only 1 line apart.
2164 (diagnostic_show_locus): When printing line numbers, show gaps in
2165 line numbering directly, rather than printing headers.
2166 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
2167 line-numbering with multiple line spans.
2168 (selftest::test_fixit_insert_containing_newline_2): Add test of
2169 line-numbering, in which the spans are close enough to be merged.
2171 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
2173 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
2174 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
2175 * tree-vrp.c (range_is_nonnull): Remove.
2176 (range_includes_zero_p): Accept value_range instead of min/max.
2177 (extract_range_from_binary_expr_1): Do not early bail on
2179 Use range_includes_zero_p instead of range_is_nonnull.
2180 (extract_range_from_unary_expr): Use range_includes_zero_p instead
2181 of range_is_nonnull.
2182 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
2183 special case VR_ANTI_RANGE.
2184 (vrp_finalize): Same.
2185 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
2187 (range_is_nonnull): Remove.
2188 * vr-values.c (vrp_stmt_computes_nonzero): Use
2189 range_includes_zero_p instead of range_is_nonnull.
2190 (extract_range_basic): Pass value_range to range_includes_zero_p
2191 instead of range_is_nonnull.
2193 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
2195 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
2196 * emit-rtl.h (rtl_data): Remove return_bnd.
2197 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
2198 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
2199 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
2200 (POINTER_BOUNDS_MODE): Remove definition.
2201 (make_pointer_bounds_mode): Remove.
2202 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
2203 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
2204 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
2205 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
2206 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
2207 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
2208 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
2210 * config/i386/i386-modes.def (BND32, BND64): Remove.
2211 * config/i386/i386.c (dbx_register_map): Remove bound registers.
2212 (dbx64_register_map): Ditto.
2213 (svr4_dbx_register_map): Ditto.
2214 (indirect_thunk_bnd_needed): Remove.
2215 (indirect_thunks_bnd_used): Ditto.
2216 (indirect_return_bnd_needed): Ditto.
2217 (indirect_return_via_cx_bnd): Ditto.
2218 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
2219 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
2220 (output_indirect_thunk): Ditto. Remove need_prefix argument.
2221 (output_indirect_thunk_function): Remove handling of
2222 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
2223 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
2224 (ix86_save_reg): Remove handling of crtl->return_bnd.
2225 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
2226 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
2227 and UNSPEC_BNDLX_ADDR.
2228 (ix86_output_indirect_branch_via_reg): Remove handling of
2229 indirect_thunk_prefix_bnd.
2230 (ix86_output_indirect_branch_via_push): Ditto.
2231 (ix86_output_function_return): Ditto.
2232 (ix86_output_indirect_function_return): Ditto.
2233 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
2234 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
2235 (CALL_USED_REGISTERS): Ditto.
2236 (REG_ALLOC_ORDER): Update for removal of bound registers.
2237 (HI_REGISTER_NAMES): Ditto.
2238 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
2239 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
2240 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
2241 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
2242 (FIRST_PSEUDO_REG): Update.
2243 (BND): Remove mode iterator.
2244 * config/i386/predicates.md (bnd_mem_operator): Remove.
2246 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
2248 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
2251 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
2253 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
2254 the case in which the permute needs only a single element and
2255 repeats for every vector of the result. Extend that case to
2256 handle variable-length vectors.
2257 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
2259 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
2262 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
2263 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
2265 2018-08-24 Richard Biener <rguenther@suse.de>
2267 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
2268 bb_flags_allocated members.
2269 (auto_flag): New RAII class for allocating flags.
2270 (auto_edge_flag): New RAII class for allocating edge flags.
2271 (auto_bb_flag): New RAII class for allocating bb flags.
2272 * cfgloop.c (verify_loop_structure): Allocate temporary edge
2274 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
2275 in favor of temporarily allocated BB flag.
2276 * hsa-brig.c: Re-order includes.
2277 * hsa-dump.c: Likewise.
2278 * hsa-regalloc.c: Likewise.
2279 * print-rtl.c: Likewise.
2280 * profile-count.c: Likewise.
2282 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
2285 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
2288 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
2291 * wide-int-range.cc (wide_int_range_div): Do not ignore result
2292 from wide_int_range_multiplicative_op.
2294 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2296 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
2299 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
2301 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
2304 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
2306 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
2307 full GENERIC expression used for replacement.
2309 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
2311 * tree-vrp.c (abs_extent_range): Remove.
2312 (extract_range_into_wide_ints): Pass wide ints by reference.
2313 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
2314 Pass wide ints by reference in all calls to
2315 extract_range_into_wide_ints.
2316 * wide-int-range.cc (wide_int_range_div): New.
2317 * wide-int-range.h (wide_int_range_div): New.
2318 (wide_int_range_includes_zero_p): New.
2319 (wide_int_range_zero_p): New.
2321 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
2323 * config/aarch64/aarch64.md (arches): New enum.
2324 (arch): New enum attr.
2325 (arch_enabled): New attr.
2326 (enabled): Now uses arch_enabled only.
2327 (simd, sve, fp16): Removed attribute.
2328 (fp): Attr now defined in terms of 'arch'.
2329 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
2330 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
2331 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
2332 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
2333 attributes into 'arch'.
2334 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
2335 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
2336 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
2337 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
2338 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
2339 'simd' attribute into 'arch'.
2340 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
2341 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
2342 Convert use of 'fp' attribute to 'arch'.
2343 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
2344 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
2346 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
2347 (different modes) Merge 'fp' and 'simd' into 'arch'.
2348 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
2351 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
2353 PR rtl-optimization/87026
2354 * expmed.c (canonicalize_comparison): If we can no longer create
2355 pseudoregisters, don't.
2357 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
2360 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
2362 * config/arm/arm.c (speculation_barrier_libfunc): New static
2364 (arm_init_libfuncs): Initialize it.
2365 (arm_emit_speculation_barrier): New function.
2366 * config/arm/arm.md (speculation_barrier): Call
2367 arm_emit_speculation_barrier for architectures that do not have
2369 (speculation_barrier_insn): Only match on Armv7 or later.
2371 2018-08-23 Richard Biener <rguenther@suse.de>
2374 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
2377 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
2379 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
2380 of single-vector TBLs.
2381 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
2384 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
2387 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
2388 aarch64_evpc_tbl guard.
2390 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
2392 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
2395 2018-08-22 Martin Sebor <msebor@redhat.com>
2398 * tree-pretty-print.c (pretty_print_string): Add argument.
2399 (dump_generic_node): Call to pretty_print_string with string size.
2401 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
2403 PR rtl-optimization/86771
2404 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
2405 of two SETs into those two SETs, one to be placed at i2, if that SETs
2406 destination is modified between i2 and i3.
2408 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2410 PR tree-optimization/86725
2411 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
2413 (vect_analyze_scalar_cycles_1): Check it.
2415 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2417 PR tree-optimization/86725
2418 * tree-vect-loop.c (vect_is_simple_reduction): When treating
2419 an outer loop phi as a double reduction, make sure that the
2420 single user of the phi result is an inner loop phi.
2422 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2424 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
2425 grouped stores with gaps to a strided group.
2427 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2429 * tree-vect-stmts.c (get_group_load_store_type)
2430 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
2431 first statement in a group.
2433 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2435 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
2436 the sequence used in gcc/gcc.c.
2438 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2441 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
2444 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2446 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
2447 Darwin10-specific unwinder-shim.
2448 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
2449 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
2450 New to cater for Darwin10 Rosetta.
2452 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2454 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
2455 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
2457 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2462 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
2463 (dwarf2out_switch_text_section): Generate a local label for the
2464 second function sub-section and apply it as the second FDE start
2466 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
2467 second sub-section start.
2469 2018-08-22 Richard Biener <rguenther@suse.de>
2471 PR tree-optimization/86988
2472 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
2474 2018-08-22 Richard Biener <rguenther@suse.de>
2476 PR tree-optimization/86945
2477 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
2479 2018-08-22 Alexandre Oliva <oliva@adacore.com>
2481 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
2482 a comment about how uses of r2 for .sdata2 come about.
2484 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
2486 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
2488 2018-08-21 Marek Polacek <polacek@redhat.com>
2490 PR c++/86981, Implement -Wpessimizing-move.
2491 * doc/invoke.texi: Document -Wpessimizing-move.
2493 2018-08-21 Jan Hubicka <jh@suse.cz>
2495 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
2496 * tree.h (is_redundant_typedef): Remove.
2497 * dwarf2out.c (is_redundant_typedef): Turn into static function.
2499 2018-08-21 Jan Hubicka <jh@suse.cz>
2501 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
2504 2018-08-21 Tamar Christina <tamar.christina@arm.com>
2506 * expmed.c (extract_low_bits): Reject invalid subregs early.
2508 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
2511 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
2514 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2516 * config/vxworks.h: Guard vxworks_asm_out_constructor and
2517 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
2518 * config/vxworks.c: Likewise.
2520 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2522 * config/vxworks.c: Set targetm.have_ctors_dtors
2523 if HAVE_INITFINI_ARRAY_SUPPORT.
2524 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
2525 if HAVE_INITFINI_ARRAY_SUPPORT.
2527 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2529 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
2530 default search path for VxWorks < 7.
2532 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
2534 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
2535 (get_format_string): Refer to c_getstr.
2537 2018-08-21 Tom de Vries <tdevries@suse.de>
2539 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
2540 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
2541 (debuginfo_early_stop): Declare.
2542 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
2543 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
2544 (debuginfo_early_stop): New function.
2545 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
2546 and debuginfo_early_stop.
2547 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
2548 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
2549 (general_init): Call debuginfo_early_init.
2550 (finalize): Call debuginfo_fini.
2551 (do_compile): Call debuginfo_init.
2552 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
2554 (@item -fdump-debug, @item -fdump-earlydebug): Add.
2556 2018-08-21 Tom de Vries <tdevries@suse.de>
2558 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
2559 flag_dump_noaddr and flag_dump_unnumbered.
2561 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
2563 * wide-int-range.cc (wide_int_range_abs): New.
2564 (wide_int_range_order_set): Rename from wide_int_range_min_max.
2565 * wide-int-range.h (wide_int_range_abs): New.
2566 (wide_int_range_min_max): New.
2567 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
2568 case to call wide_int_range_abs.
2569 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
2570 (extract_range_from_abs_expr): Delete.
2572 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
2575 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
2576 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
2579 2018-08-20 Nathan Sidwell <nathan@acm.org>
2580 Jeff Law <law@redhat.com>
2582 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
2583 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
2585 2018-08-20 David Malcolm <dmalcolm@redhat.com>
2588 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
2589 (decl_attributes): Likewise.
2590 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
2592 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
2593 * diagnostic-core.h (class auto_diagnostic_group): New class.
2594 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
2595 (diagnostic_report_diagnostic): Handle the first diagnostics within
2597 (emit_diagnostic): Add auto_diagnostic_group instance.
2599 (inform_n): Likewise.
2600 (warning): Likewise.
2601 (warning_at): Likewise.
2602 (warning_n): Likewise.
2603 (pedwarn): Likewise.
2604 (permerror): Likewise.
2606 (error_n): Likewise.
2607 (error_at): Likewise.
2609 (fatal_error): Likewise.
2610 (internal_error): Likewise.
2611 (internal_error_no_backtrace): Likewise.
2612 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
2613 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
2614 * diagnostic.h (struct diagnostic_context): Add fields
2615 "diagnostic_group_nesting_depth",
2616 "diagnostic_group_emission_count", "begin_group_cb",
2618 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
2619 Add auto_diagnostic_group instance(s).
2620 (find_explicit_erroneous_behavior): Likewise.
2621 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
2622 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
2623 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
2624 (gimplify_va_arg_expr): Likewise.
2625 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
2626 (HSA_SORRY_AT): Likewise.
2627 * ipa-devirt.c (compare_virtual_tables): Likewise.
2628 (warn_odr): Likewise.
2629 * multiple_target.c (expand_target_clones): Likewise.
2630 * opts-common.c (cmdline_handle_error): Likewise.
2631 * reginfo.c (globalize_reg): Likewise.
2632 * substring-locations.c (format_warning_n_va): Likewise.
2633 * tree-inline.c (expand_call_inline): Likewise.
2634 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
2635 * tree-ssa-loop-niter.c
2636 (do_warn_aggressive_loop_optimizations): Likewise.
2637 * tree-ssa-uninit.c (warn_uninit): Likewise.
2638 * tree.c (warn_deprecated_use): Likewise.
2640 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
2643 * config/i386/i386.md (eh_return): Always update EH return
2644 address in word_mode.
2646 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
2648 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
2649 TARGET_SPLIT_COMPLEX_ARG is defined.
2651 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
2653 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
2655 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
2658 * expr.c (expand_assignment): Assert that bitpos is positive.
2659 (store_field): Likewise
2660 (expand_expr_real_1): Make sure that bitpos is positive.
2661 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
2664 2018-08-20 Nathan Sidwell <nathan@acm.org>
2666 * Makefile.in (CPP_ID_DATA_H): Delete.
2667 (CPP_INTERNAL_H): Don't add it.
2668 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
2669 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
2671 2018-08-20 Richard Biener <rguenther@suse.de>
2673 PR tree-optimization/78655
2674 * tree-vrp.c (extract_range_from_binary_expr_1): Make
2675 pointer + offset nonnull if either operand is nonnull work.
2677 2018-08-20 Tom de Vries <tdevries@suse.de>
2679 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
2680 unless the referenced die describes the added property using
2681 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
2682 Otherwise, add a DW_AT_location to the referenced die.
2684 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
2687 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
2688 register_operand when calling ix86_set_reg_reg_cost.
2689 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
2690 Set *total to 0 for operands that satisfy x86_64_immediate_operand
2691 predicate and to 1 otherwise.
2693 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
2695 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
2696 emit a diagnostic that it is not supported and reset the option.
2697 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
2698 supported and consume the option. (ASM_FINAL_SPEC): New.
2700 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
2702 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
2705 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
2707 C-SKY port: Documentation
2709 * doc/extend.texi (C-SKY Function Attributes): New section.
2710 * doc/invoke.texi (Option Summary): Add C-SKY options.
2711 (C-SKY Options): New section.
2712 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
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>
2719 C-SKY port: Backend implementation
2721 * config/csky/*: New.
2722 * common/config/csky/*: New.
2724 2018-08-17 Jojo <jijie_rong@c-sky.com>
2725 Huibin Wang <huibin_wang@c-sky.com>
2726 Sandra Loosemore <sandra@codesourcery.com>
2727 Chung-Lin Tang <cltang@codesourcery.com>
2728 Andrew Jenner <andrew@codesourcery.com>
2730 C-SKY port: Configury
2732 * config.gcc (csky-*-*): New.
2733 * configure.ac: Add csky to targets for dwarf2 debug_line support.
2734 * configure: Regenerated.
2736 2018-08-17 David Malcolm <dmalcolm@redhat.com>
2738 * dump-context.h: Include "dumpfile.h".
2739 (dump_context::dump_printf_va): Convert final param from va_list
2740 to va_list *. Convert from ATTRIBUTE_PRINTF to
2741 ATTRIBUTE_GCC_DUMP_PRINTF.
2742 (dump_context::dump_printf_loc_va): Likewise.
2743 * dumpfile.c: Include "stringpool.h".
2744 (make_item_for_dump_printf_va): Delete.
2745 (make_item_for_dump_printf): Delete.
2746 (class dump_pretty_printer): New class.
2747 (dump_pretty_printer::dump_pretty_printer): New ctor.
2748 (dump_pretty_printer::emit_items): New member function.
2749 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
2751 (dump_pretty_printer::emit_item): New member function.
2752 (dump_pretty_printer::stash_item): New member function.
2753 (dump_pretty_printer::format_decoder_cb): New member function.
2754 (dump_pretty_printer::decode_format): New member function.
2755 (dump_context::dump_printf_va): Reimplement in terms of
2756 dump_pretty_printer.
2757 (dump_context::dump_printf_loc_va): Convert final param from va_list
2759 (dump_context::begin_scope): Reimplement call to
2760 make_item_for_dump_printf.
2761 (dump_printf): Update for change to dump_printf_va.
2762 (dump_printf_loc): Likewise.
2763 (selftest::test_capture_of_dump_calls): Convert "stmt" from
2764 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
2765 with %T, %E, and %G.
2766 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
2767 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
2768 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
2769 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
2770 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
2771 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
2772 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
2773 within a dump_printf_loc call to "%wu".
2774 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
2775 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
2776 missing space after "=".
2777 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
2778 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
2779 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
2780 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
2781 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
2782 duplicate "vectorized" from message.
2784 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
2786 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
2787 polyNxK_t element's TYPE_STRING_FLAG.
2789 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
2791 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
2792 (they were unnamed before). Fix comments.
2794 2018-08-17 Nathan Sidwell <nathan@acm.org>
2796 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
2798 2018-08-17 Richard Biener <rguenther@suse.de>
2800 PR tree-optimization/86841
2801 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
2803 2018-08-17 Martin Liska <mliska@suse.cz>
2805 * common.opt: Remove Warn, Init and Report for options with
2806 Ignore/Deprecated flag. Warning is done automatically for
2808 * config/i386/i386.opt: Likewise.
2809 * config/ia64/ia64.opt: Likewise.
2810 * config/rs6000/rs6000.opt: Likewise.
2811 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
2812 Remove usage of flag_check_pointer_bounds.
2813 * lto-wrapper.c (merge_and_complain): Do not handle
2814 OPT_fcheck_pointer_bounds.
2815 (append_compiler_options): Likewise.
2816 * opt-functions.awk: Do not handle Deprecated.
2817 * optc-gen.awk: Check that Var, Report and Init are not
2818 used for an option with Ignore/Deprecated flag.
2819 * opts-common.c (decode_cmdline_option): Do not report
2821 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
2823 * opts.h (struct cl_option): Remove cl_deprecated flag.
2824 (CL_ERR_DEPRECATED): Remove error enum value.
2826 2018-08-17 Richard Biener <rguenther@suse.de>
2829 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
2830 across a va-arg-pack using call adjust its return value accordingly.
2832 2018-08-16 Martin Sebor <msebor@redhat.com>
2834 PR tree-optimization/86853
2835 * gimple-ssa-sprintf.c (struct format_result): Rename member.
2836 (struct fmtresult): Add member and initialize it in ctors.
2837 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
2838 (format_string): Handle %S the same as %ls. Set MAYFAIL.
2839 (format_directive): Set POSUNDER4K when MAYFAIL is set.
2840 (parse_directive): Handle %C same as %c.
2841 (sprintf_dom_walker::compute_format_length): Adjust.
2842 (is_call_safe): Adjust.
2844 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
2846 * builtins.c (c_strlen): Add new parameter eltsize. Use it
2847 for determining how to count the elements.
2848 * builtins.h (c_strlen): Adjust prototype.
2849 * expr.c (string_constant): Add new parameter mem_size.
2850 Set *mem_size appropriately.
2851 * expr.h (string_constant): Adjust protoype.
2852 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
2853 * gimple-fold.h (get_range_strlen): Adjust prototype.
2854 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
2855 (format_string): Call get_string_length with eltsize.
2857 2018-08-16 David Malcolm <dmalcolm@redhat.com>
2859 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
2860 to emit the span, rather than setting it as the prefix.
2862 2018-08-16 David Malcolm <dmalcolm@redhat.com>
2864 * diagnostic-show-locus.c (layout::start_annotation_line): Add
2865 "margin_char" parameter, defaulting to space. Use it in place
2866 of pp_space for the initial part of the margin.
2867 (layout::print_leading_fixits): Use '+' when filling the margin
2868 of line-insertion fix-it hints.
2870 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2872 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
2875 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2877 * config/rs6000/altivec.md: Don't set length attribute to the default
2879 * config/rs6000/darwin.md: Ditto.
2880 * config/rs6000/dfp.md: Ditto.
2881 * config/rs6000/htm.md: Ditto.
2882 * config/rs6000/rs6000.md: Ditto.
2883 * config/rs6000/sync.md: Ditto.
2884 * config/rs6000/vsx.md: Ditto.
2886 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2888 * config/rs6000/altivec.md: Don't set length attribute to the default
2889 value, for branch instructions.
2890 * config/rs6000/darwin.md: Ditto.
2891 * config/rs6000/rs6000.md: Ditto.
2893 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2895 * config/rs6000/rs6000.md (length): Always define as const_int 4.
2896 (unnamed conditional branch define_insn): Set length to 4 or 8
2897 depending on offset.
2898 (<bd>_<mode>): Similar, for alternative 0.
2899 (<bd>tf_<mode>): Ditto.
2901 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2903 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
2905 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
2907 * doc/rtl.texi: Replace old RTX class names with new names.
2910 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
2912 * expmed.h (canonicalize_comparison): New declaration.
2913 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
2914 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
2915 * optabs.c (prepare_cmp_insn): Likewise.
2916 * rtl.h (unsigned_condition_p): New function which checks if a
2917 comparison operator is unsigned.
2919 2018-08-16 Nathan Sidwell <nathan@acm.org>
2921 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
2922 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
2924 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2927 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
2928 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
2929 (mov<mov>): ..this and enable unconditionally.
2931 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2933 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
2935 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
2937 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
2938 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
2939 "Common" with "Target".
2941 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
2943 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
2944 * doc/invoke.texi (mmitigate-rop): Remove.
2945 * config/i386/i386.c: Do not include "regrename.h".
2946 (ix86_rop_should_change_byte_p, reg_encoded_number)
2947 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
2949 (ix86_reorg): Remove call to ix86_mitigate_rop.
2950 * config/i386/i386.md (attr "modrm_class"): Remove.
2951 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
2952 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
2953 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
2955 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
2957 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
2958 allow folding of mergeh() and mergel() for the float and double types.
2959 (fold_mergehl_helper): Rework to handle building a permute tree
2962 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
2964 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
2967 2018-08-15 David Malcolm <dmalcolm@redhat.com>
2969 * common.opt (fdiagnostics-show-labels): New option.
2970 * diagnostic-show-locus.c (class layout_range): Add field
2972 (class layout): Add field "m_show_labels_p".
2973 (layout_range::layout_range): Add param "label" and use it to
2975 (make_range): Pass in NULL for new "label" param of layout_range's
2977 (layout::layout): Initialize m_show_labels_p.
2978 (layout::maybe_add_location_range): Pass in loc_range->m_label
2979 when constructing layout_range instances.
2980 (struct line_label): New struct.
2981 (layout::print_any_labels): New member function.
2982 (layout::print_line): Call it if label-printing is enabled.
2983 (selftest::test_one_liner_labels): New test.
2984 (selftest::test_diagnostic_show_locus_one_liner): Call it.
2985 * diagnostic.c (diagnostic_initialize): Initialize
2986 context->show_labels_p.
2987 * diagnostic.h (struct diagnostic_context): Add field
2989 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
2990 -fno-diagnostics-show-labels.
2991 * dwarf2out.c (gen_producer_string): Add
2992 OPT_fdiagnostics_show_labels to the ignored options.
2993 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
2995 (gcc_rich_location::maybe_add_expr): Likewise.
2996 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
2997 label" param, defaulting to NULL.
2998 (gcc_rich_location::add_expr): Add "label" param.
2999 (gcc_rich_location::maybe_add_expr): Likewise.
3000 (class text_range_label): New class.
3001 (class range_label_for_type_mismatch): New class.
3002 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
3003 of format_warning_va.
3004 (fmtwarn_n): Likewise for new params of format_warning_n_va.
3005 * lto-wrapper.c (merge_and_complain): Add
3006 OPT_fdiagnostics_show_labels to the "pick one setting" options.
3007 (append_compiler_options): Likewise to the dropped options.
3008 (append_diag_options): Likewise to the passed-on options.
3009 * opts.c (common_handle_option): Handle the new option.
3010 * selftest-diagnostic.c
3011 (test_diagnostic_context::test_diagnostic_context): Enable
3013 * substring-locations.c: Include "gcc-rich-location.h".
3014 (format_warning_n_va): Add "fmt_label" and "param_label" params
3015 and use them as appropriate.
3016 (format_warning_va): Add "fmt_label" and "param_label" params,
3017 passing them on to format_warning_n_va.
3018 (format_warning_at_substring): Likewise.
3019 (format_warning_at_substring_n): Likewise.
3020 * substring-locations.h (format_warning_va): Add "fmt_label" and
3021 "param_label" params.
3022 (format_warning_n_va): Likewise.
3023 (format_warning_at_substring): Likewise.
3024 (format_warning_at_substring_n): Likewise.
3025 * toplev.c (general_init): Initialize global_dc->show_labels_p.
3027 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
3030 * builtins.c (expand_builtin_memcmp): Do not expand the call
3031 when overflow is detected.
3033 2018-08-15 Martin Sebor <msebor@redhat.com>
3035 PR tree-optimization/71625
3036 * config/aarch64/aarch64-builtins.c
3037 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
3039 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
3041 * config/s390/s390.c (s390_reorg): Remove loop.
3043 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
3046 (darwin_function_switched_text_sections): Delete.
3047 * gcc/config/darwin.h
3048 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
3050 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
3053 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
3054 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
3055 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
3057 2018-08-15 Martin Liska <mliska@suse.cz>
3059 PR tree-optimization/86925
3060 * predict.c (expr_expected_value_1): When taking
3061 later predictor, assign also probability.
3062 Use fold_build2_initializer_loc in order to fold
3063 the expression in -frounding-math.
3065 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
3067 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
3069 (expand_vec_perm_1): Try the new method.
3071 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
3074 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
3075 Check whether lra_live_max_point is 0 before dividing.
3077 2018-08-14 Martin Sebor <msebor@redhat.com>
3079 PR tree-optimization/86650
3080 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
3081 (vrp_prop::check_mem_ref): Same.
3083 2018-08-13 Liu Hao <lh_mouse@126.com>
3085 * pretty-print.c (eat_esc_sequence): Swap the foreground and
3086 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
3087 and clear it thereafter, as it only works for DBCS.
3089 2018-08-13 Liu Hao <lh_mouse@126.com>
3091 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
3092 handle returned by _get_osf_handle().
3094 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
3096 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
3097 for folding vec_perm.
3099 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
3101 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
3102 Add support for gimple-folding of vec_pack() and vec_unpack()
3105 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
3107 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
3108 vec_xst variants to the list.
3109 (rs6000_gimple_fold_builtin): Add support for folding unaligned
3110 vector loads and stores.
3112 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
3114 * config.gcc (rs6000-ibm-aix4.x): Delete.
3115 (rs6000-ibm-aix5.1): Delete.
3116 (rs6000-ibm-aix5.2): Delete.
3117 (rs6000-ibm-aix5.3): Delete.
3118 * config/rs6000/aix43.h: Delete.
3119 * config/rs6000/aix51.h: Delete.
3120 * config/rs6000/aix52.h: Delete.
3121 * config/rs6000/t-aix43: Delete.
3123 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
3125 * config/s390/s390.c (s390_decompose_constant_pool_ref):
3127 (s390_decompose_address): Factor out constant pool ref
3130 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
3132 * config/nds32/nds32-predicates.c
3133 (nds32_can_use_bclr_p): Change return type as bool.
3134 (nds32_can_use_bset_p): Ditto.
3135 (nds32_can_use_btgl_p): Ditto.
3136 (nds32_can_use_bitci_p): Ditto.
3137 * config/nds32/nds32-protos.h
3138 (nds32_can_use_bclr_p): Change declaration.
3139 (nds32_can_use_bset_p): Ditto.
3140 (nds32_can_use_btgl_p): Ditto.
3141 (nds32_can_use_bitci_p): Ditto.
3143 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
3145 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
3146 Support -msched-prolog-epilog option.
3147 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
3149 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
3151 * common/config/nds32/nds32-common.c
3152 (nds32_option_optimization_table): Enalbe -malways-align.
3154 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
3156 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
3158 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
3159 OPT_misr_secure_ case.
3160 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
3161 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
3162 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
3164 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
3165 (nds32_isr_info): New field security_level.
3166 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
3167 * config/nds32/nds32.md (return_internal): Consider critical attribute.
3168 * config/nds32/nds32.opt (misr-secure): New option.
3169 * config/nds32/nds32_init.inc: New file.
3170 * config/nds32/nds32_isr.h: New file.
3172 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
3174 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
3175 Update comment for atomic instructions.
3176 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
3177 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
3179 (atomic_loaddi): Revise fence expansion to only emit fence prior to
3180 load for __ATOMIC_SEQ_CST model.
3181 (atomic_loaddi_1): Remove float register target.
3182 (atomic_storedi): Handle CONST_INT values.
3183 (atomic_storedi_1): Remove float register source. Add special case
3185 (memory_barrier): New expander and insn.
3187 2018-08-11 Jakub Jelinek <jakub@redhat.com>
3189 PR tree-optimization/86835
3190 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
3191 new_stmt after def_gsi, make sure to insert new_square_stmt after
3192 that stmt, not 2 stmts before it.
3194 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
3197 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
3200 2018-08-10 Martin Liska <mliska@suse.cz>
3203 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
3205 * builtins.c (expand_builtin_expect_with_probability):
3207 (expand_builtin_expect_with_probability): New function.
3208 (build_builtin_expect_predicate): Add new argumnet probability
3209 for BUILT_IN_EXPECT_WITH_PROBABILITY.
3210 (fold_builtin_expect):
3213 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
3214 * builtins.h (fold_builtin_expect): Set new argument.
3215 * doc/extend.texi: Document __builtin_expect_with_probability.
3216 * doc/invoke.texi: Likewise.
3217 * gimple-fold.c (gimple_fold_call): Pass new argument.
3218 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
3219 also BUILT_IN_EXPECT_WITH_PROBABILITY.
3220 * predict.c (get_predictor_value): New function.
3221 (expr_expected_value): Add new argument probability. Assume
3222 that predictor and probability are always non-null.
3223 (expr_expected_value_1): Likewise. For __builtin_expect and
3224 __builtin_expect_with_probability set probability. Handle
3225 combination in binary expressions.
3226 (tree_predict_by_opcode): Simplify code by simply calling
3227 get_predictor_value.
3228 (pass_strip_predict_hints::execute): Add handling of
3229 BUILT_IN_EXPECT_WITH_PROBABILITY.
3230 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
3232 * tree.h (DECL_BUILT_IN_P): New function.
3234 2018-08-10 Martin Liska <mliska@suse.cz>
3236 PR tree-optimization/85799
3237 * passes.def: Add argument for pass_strip_predict_hints.
3238 * predict.c (class pass_strip_predict_hints): Add new argument
3240 (strip_predictor_early): New function.
3241 (pass_strip_predict_hints::execute): Call the function to
3243 (strip_predict_hints): New function.
3244 * predict.def: Fix comment.
3246 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
3248 * Makefile.in: Clarify which tm.texi to copy over to assert the
3249 right to grant a GFDL license for all.
3251 2018-08-09 Jeff Law <law@redhat.com>
3253 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
3256 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
3258 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
3261 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
3263 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
3264 reductions for variable-length vectors.
3266 2018-08-09 David Malcolm <dmalcolm@redhat.com>
3269 * common.opt (fdiagnostics-show-line-numbers): New option.
3270 * diagnostic-show-locus.c (class layout): Add fields
3271 "m_show_line_numbers_p" and "m_linenum_width";
3272 (num_digits): New function.
3273 (test_num_digits): New function.
3274 (layout::layout): Initialize new fields. Update m_x_offset
3275 logic to handle any left margin.
3276 (layout::print_source_line): Print line number when requested.
3277 (layout::start_annotation_line): New member function.
3278 (layout::print_annotation_line): Call it.
3279 (layout::print_leading_fixits): Likewise.
3280 (layout::print_trailing_fixits): Likewise. Update calls to
3281 move_to_column for new parameter.
3282 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
3283 it to potentially call start_annotation_line.
3284 (layout::show_ruler): Call start_annotation_line.
3285 (selftest::test_line_numbers_multiline_range): New selftest.
3286 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
3287 and selftest::test_line_numbers_multiline_range.
3288 * diagnostic.c (diagnostic_initialize): Initialize
3289 show_line_numbers_p.
3290 * diagnostic.h (struct diagnostic_context): Add field
3291 "show_line_numbers_p".
3292 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
3293 -fno-diagnostics-show-line-numbers.
3294 * dwarf2out.c (gen_producer_string): Add
3295 OPT_fdiagnostics_show_line_numbers to the ignored options.
3296 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
3297 one setting" options.
3298 (append_compiler_options): Likewise to the dropped options.
3299 (append_diag_options): Likewise to the passed-on options.
3300 * opts.c (common_handle_option): Handle the new option.
3301 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
3303 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
3305 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
3306 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
3307 third argument of type "const signed char" to descriptions of
3308 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
3309 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
3310 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
3311 __builtin_bcdsub_ov functions.
3313 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
3315 PR tree-optimization/86858
3316 * tree-vect-loop.c (vect_is_simple_reduction): Restore
3317 flow_bb_inside_loop_p calls.
3319 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
3321 PR tree-optimization/86871
3322 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
3323 instead of gimple_assign_lhs.
3325 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
3328 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
3329 register constraint to operand 0.
3330 (add<mode>3_carryinC): Likewise.
3331 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
3333 2018-08-09 Martin Liska <mliska@suse.cz>
3336 * common.opt: Remove extra line.
3338 2018-08-09 Martin Liska <mliska@suse.cz>
3340 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
3341 at the end of a line, make first letter capital and end up
3342 a sentence with a dot.
3343 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
3344 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
3345 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
3346 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
3347 (PARAM_MAX_ISL_OPERATIONS): Likewise.
3348 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
3349 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
3350 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
3351 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
3352 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
3353 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
3354 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
3355 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
3356 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
3357 (PARAM_TREE_REASSOC_WIDTH): Likewise.
3358 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
3359 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
3360 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
3362 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
3365 * config/s390/s390.c (s390_option_override_internal): Reduce the
3366 stack-clash-protection-probe-interval param if it would be too big
3369 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
3372 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
3373 * config/m68k/m68k.c (handle_move_double): Don't call
3374 m68k_final_prescan_insn.
3375 (m68k_adjust_decorated_operand): Renamed from
3376 m68k_final_prescan_insn, remove first and third operand and
3378 (print_operand): Call it.
3379 (print_operand_address): Call it.
3381 2018-08-08 Nathan Sidwell <nathan@acm.org>
3383 * diagnostic.c (diagnostic_report_current_module): Use
3384 linemap_included_from & linemap_included_from_linemap.
3386 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
3388 * config/aarch64/aarch64-cores.def: Add phecda core.
3389 * config/aarch64/aarch64-tune.md: Regenerate.
3390 * doc/invoke.texi: Add phecda core.
3392 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
3395 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
3397 * config/s390/s390.md ("movti"): Add more alternatives for
3398 constant to GPR copies.
3400 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
3402 * config/s390/s390.c: Fix whitespace damage throughout the file.
3403 * config/s390/s390.h: Likewise.
3404 * config/s390/tpf.h: Likewise.
3406 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
3408 * config/s390/s390.c (s390_loadrelative_operand_p):
3409 Remove TARGET_CPU_ZARCH usages.
3410 (s390_rtx_costs): Likewise.
3411 (s390_legitimate_constant_p): Likewise.
3412 (s390_cannot_force_const_mem): Likewise.
3413 (legitimate_reload_constant_p): Likewise.
3414 (s390_preferred_reload_class): Likewise.
3415 (legitimize_pic_address): Likewise.
3416 (legitimize_tls_address): Likewise.
3417 (s390_split_branches): Removed.
3418 (s390_add_execute): Removed.
3419 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
3420 (s390_mainpool_start): Likewise.
3421 (s390_mainpool_finish): Likewise.
3422 (s390_mainpool_cancel): Removed.
3423 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
3424 (s390_chunkify_cancel): Likewise.
3425 (s390_return_addr_rtx): Likewise.
3426 (s390_register_info): Remove split_branches_pending_p uages.
3427 (s390_optimize_register_info): Likewise.
3428 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
3429 split_branches_pending_p usages.
3430 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
3431 (s390_load_got): Likewise.
3432 (s390_expand_split_stack_prologue): Likewise.
3433 (output_asm_nops): Likewise.
3434 (s390_function_profiler): Likewise.
3435 (s390_emit_call): Likewise.
3436 (s390_conditional_register_usage): Likewise.
3437 (s390_optimize_prologue): Likewise.
3438 (s390_reorg): Remove TARGET_CPU_ZARCH and
3439 split_branches_pending_p usages.
3440 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
3442 (s390_output_indirect_thunk_function): Likewise.
3443 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
3444 (TARGET_CPU_ZARCH_P): Removed.
3445 (struct machine_function): Remove split_branches_pending_p.
3446 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
3448 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
3450 * common/config/s390/s390-common.c (processor_flags_table):
3452 * config.gcc: Remove with_arch/with_tune support.
3453 * config/s390/2064.md: Remove cpu attribute comparisons.
3454 * config/s390/driver-native.c (s390_host_detect_local_cpu):
3456 * config/s390/linux.h (ASM_SPEC):
3457 Remove -march support.
3458 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
3459 Use a table to get an arch level.
3460 * config/s390/s390-opts.h (enum processor_type):
3462 * config/s390/s390.c
3463 (processor_table): Remove entries, add arch_level values.
3464 (s390_issue_rate): Remove cases.
3465 (s390_option_override): Adjust
3466 s390_option_override_internal() call.
3467 (s390_option_override_internal): Remove deprecation warning.
3468 (s390_valid_target_attribute_tree): Adjust
3469 s390_option_override_internal() call.
3470 * config/s390/s390.h (struct s390_processor):
3471 Share with s390-c.c, add arch_level field.
3472 * config/s390/s390.md:
3473 Remove occurrences in cpu attribute.
3474 * config/s390/s390.opt: Remove -march/-mtune support.
3475 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
3476 * doc/invoke.texi: Remove deprecation warning.
3478 2018-08-08 Luis Machado <luis.machado@linaro.org>
3480 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
3482 (qdf24xx_tunings): Set vector cost structure to
3483 qdf24xx_vector_cost.
3485 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
3486 <register_sextend>: Set to 3.
3488 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
3491 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
3492 * config/aarch64/aarch64-simd.md
3493 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
3494 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
3495 (aarch64_frecpx<mode>): ...this new pattern.
3496 * config/aarch64/aarch64-simd-builtins.def: Remove comment
3497 about aarch64_frecp<FRECP:frecp_suffix><mode>.
3499 2018-08-07 Martin Liska <mliska@suse.cz>
3502 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
3503 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
3504 * predict.def (PRED_MALLOC_NONNULL): New predictor.
3505 * doc/extend.texi: Document that malloc attribute adds
3508 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
3511 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3512 Define to speculation_safe_value_not_needed.
3514 2018-08-06 Jeff Law <law@redhat.com>
3516 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
3517 the vr_values instance to cprop_into_stmt.
3518 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
3519 (cprop_operand): Also query EVRP to determine if OP is a constant.
3521 2018-08-06 Nathan Sidwell <nathan@acm.org>
3523 * diagnostic.c (diagnostic_report_current_module): Reroll
3524 included-at loop. Translate text.
3526 2018-08-06 David Malcolm <dmalcolm@redhat.com>
3528 * function-tests.c (selftest::test_expansion_to_rtl): Call
3529 free_after_compilation.
3531 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3533 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
3535 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
3537 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
3538 loops with memory block operations from getting unrolled.
3540 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
3543 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3544 Define to speculation_safe_value_not_needed.
3546 2018-08-06 Jeff Law <law@redhat.com>
3548 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
3551 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3554 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
3555 with all enabled __intN types.
3557 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
3559 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3561 * alias.c (record_set): Check for clobber high.
3562 * cfgexpand.c (expand_gimple_stmt): Likewise.
3563 * combine-stack-adj.c (single_set_for_csa): Likewise.
3564 * combine.c (find_single_use_1): Likewise.
3565 (set_nonzero_bits_and_sign_copies): Likewise.
3566 (get_combine_src_dest): Likewise.
3567 (is_parallel_of_n_reg_sets): Likewise.
3568 (try_combine): Likewise.
3569 (record_dead_and_set_regs_1): Likewise.
3570 (reg_dead_at_p_1): Likewise.
3571 (reg_dead_at_p): Likewise.
3572 * dce.c (deletable_insn_p): Likewise.
3573 (mark_nonreg_stores_1): Likewise.
3574 (mark_nonreg_stores_2): Likewise.
3575 * df-scan.c (df_find_hard_reg_defs): Likewise.
3576 (df_uses_record): Likewise.
3577 (df_get_call_refs): Likewise.
3578 * dwarf2out.c (mem_loc_descriptor): Likewise.
3579 * haifa-sched.c (haifa_classify_rtx): Likewise.
3580 * ira-build.c (create_insn_allocnos): Likewise.
3581 * ira-costs.c (scan_one_insn): Likewise.
3582 * ira.c (equiv_init_movable_p): Likewise.
3583 (rtx_moveable_p): Likewise.
3584 (interesting_dest_for_shprep): Likewise.
3585 * jump.c (mark_jump_label_1): Likewise.
3586 * postreload-gcse.c (record_opr_changes): Likewise.
3587 * postreload.c (reload_cse_simplify): Likewise.
3588 (struct reg_use): Add source expr.
3589 (reload_combine): Check for clobber high.
3590 (reload_combine_note_use): Likewise.
3591 (reload_cse_move2add): Likewise.
3592 (move2add_note_store): Likewise.
3593 * print-rtl.c (print_pattern): Likewise.
3594 * recog.c (decode_asm_operands): Likewise.
3595 (store_data_bypass_p): Likewise.
3596 (if_test_bypass_p): Likewise.
3597 * regcprop.c (kill_clobbered_value): Likewise.
3598 (kill_set_value): Likewise.
3599 * reginfo.c (reg_scan_mark_refs): Likewise.
3600 * reload1.c (maybe_fix_stack_asms): Likewise.
3601 (eliminate_regs_1): Likewise.
3602 (elimination_effects): Likewise.
3603 (mark_not_eliminable): Likewise.
3604 (scan_paradoxical_subregs): Likewise.
3605 (forget_old_reloads_1): Likewise.
3606 * reorg.c (find_end_label): Likewise.
3607 (try_merge_delay_insns): Likewise.
3608 (redundant_insn): Likewise.
3609 (own_thread_p): Likewise.
3610 (fill_simple_delay_slots): Likewise.
3611 (fill_slots_from_thread): Likewise.
3612 (dbr_schedule): Likewise.
3613 * resource.c (update_live_status): Likewise.
3614 (mark_referenced_resources): Likewise.
3615 (mark_set_resources): Likewise.
3616 * rtl.c (copy_rtx): Likewise.
3617 * rtlanal.c (reg_referenced_p): Likewise.
3618 (single_set_2): Likewise.
3619 (noop_move_p): Likewise.
3620 (note_stores): Likewise.
3621 * sched-deps.c (sched_analyze_reg): Likewise.
3622 (sched_analyze_insn): Likewise.
3624 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3626 * cse.c (invalidate_reg): New function extracted from...
3627 (invalidate): ...here.
3628 (canonicalize_insn): Check for clobber high.
3629 (invalidate_from_clobbers): invalidate clobber highs.
3630 (invalidate_from_sets_and_clobbers): Likewise.
3631 (count_reg_usage): Check for clobber high.
3632 (insn_live_p): Likewise.
3633 * cselib.c (cselib_expand_value_rtx_1):Likewise.
3634 (cselib_invalidate_regno): Check for clobber in setter.
3635 (cselib_invalidate_rtx): Pass through setter.
3636 (cselib_invalidate_rtx_note_stores):
3637 (cselib_process_insn): Check for clobber high.
3638 * cselib.h (cselib_invalidate_rtx): Add operand.
3640 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3642 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
3643 (mark_not_eliminable): Likewise.
3644 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
3645 * lra-lives.c (process_bb_lives): Check for clobber high.
3646 * lra.c (new_insn_reg): Remember clobber highs.
3647 (collect_non_operand_hard_regs): Check for clobber high.
3648 (lra_set_insn_recog_data): Likewise.
3649 (add_regs_to_insn_regno_info): Likewise.
3650 (lra_update_insn_regno_info): Likewise.
3652 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3654 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
3655 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
3657 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3659 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
3660 (copy_insn_1): Likewise.
3661 (gen_hard_reg_clobber_high): New gen function.
3662 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
3663 * genemit.c (gen_exp): Likewise.
3664 (gen_emit_seq): Pass through info.
3665 (gen_insn): Check for CLOBBER_HIGH.
3666 (gen_expand): Pass through info.
3667 (gen_split): Likewise.
3668 (output_add_clobbers): Likewise.
3669 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
3670 (remove_clobbers): Likewise.
3671 * rtl.h (gen_hard_reg_clobber_high): New declaration.
3673 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3675 * doc/rtl.texi (clobber_high): Add.
3676 (parallel): Add in clobber high
3677 * rtl.c (rtl_check_failed_code3): Add function.
3678 * rtl.def (CLOBBER_HIGH): Add expression.
3679 * rtl.h (RTL_CHECKC3): Add macro.
3680 (rtl_check_failed_code3): Add declaration.
3681 (XC3EXP): Add macro.
3683 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
3686 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
3687 cfun->machine->max_used_stack_alignment if needed.
3689 2018-08-04 Martin Sebor <msebor@redhat.com>
3691 PR tree-optimization/86571
3692 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
3695 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
3698 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3701 2018-08-03 Jeff Law <law@redhat.com>
3704 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3705 Define to speculation_safe_value_not_needed.
3707 2018-08-03 David Malcolm <dmalcolm@redhat.com>
3709 * doc/gcov.texi (-x): Remove duplicate "to".
3710 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
3711 (-Wif-not-aligned): Remove duplicate "is".
3712 (-flto): Remove duplicate "the".
3713 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
3714 duplicate "v5.00.b".
3715 (MSP430 Options): Remove duplicate "and" from the description
3716 of "-mgprel-sec=regexp".
3717 (x86 Options): Remove duplicate copies of "vmldLog102" and
3718 vmlsLog104 from description of "-mveclibabi=type".
3720 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
3722 * internal-fn.h (first_commutative_argument): Declare.
3723 * internal-fn.c (first_commutative_argument): New function.
3724 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
3725 restrictions for pattern statements. Use first_commutative_argument
3726 to look for commutative operands in calls to internal functions.
3728 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
3730 * Makefile.in (wide-int-range.o): New.
3731 * tree-vrp.c: Move all the wide_int_* functions to...
3732 * wide-int-range.cc: ...here.
3733 * tree-vrp.h: Move all the wide_int_* prototypes to...
3734 * wide-int-range.h: ...here.
3736 2018-08-03 Tom de Vries <tdevries@suse.de>
3738 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
3740 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
3741 * except.c (output_function_exception_table): Do early exit if
3742 targetm_common.except_unwind_info (&global_options) == UI_NONE.
3744 2018-08-03 Martin Liska <mliska@suse.cz>
3746 * predict.c (dump_prediction): Change to 2 digits
3749 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
3751 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
3753 (falkor_am_1_gtov_gtov): ... a new insn reservation.
3755 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
3757 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
3758 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
3759 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
3760 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
3761 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
3763 2018-08-02 David Malcolm <dmalcolm@redhat.com>
3765 * diagnostic-show-locus.c (diagnostic_show_locus): Use
3766 pp_take_prefix when saving the existing prefix.
3767 * diagnostic.c (diagnostic_append_note): Likewise.
3768 * langhooks.c (lhd_print_error_function): Likewise.
3769 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
3770 param's type. Free the existing prefix.
3771 (pp_take_prefix): New function.
3772 (pretty_printer::pretty_printer): Drop the prefix parameter.
3773 Rename the length parameter to match the comment.
3774 (pretty_printer::~pretty_printer): Free the prefix.
3775 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
3777 (struct pretty_printer): Drop the "const" from "prefix" field's
3778 type and clarify memory management.
3779 (pp_set_prefix): Drop the "const" from the 2nd param.
3780 (pp_take_prefix): New decl.
3782 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
3784 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
3785 for word_mode_ok here instead of passing as argument.
3786 (expand_block_compare): Change select_block_compare_mode() call.
3787 (expand_strncmp_gpr_sequence): New function.
3788 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
3790 2018-08-02 Jeff Law <law@redhat.com>
3793 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3794 Define to speculation_safe_value_not_needed.
3797 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3798 Define to speculation_safe_value_not_needed.
3800 2018-08-02 Tom de Vries <tdevries@suse.de>
3803 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
3804 function. Return UI_TARGET unconditionally.
3805 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
3806 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
3808 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
3810 * genemit.c (print_overload_test): Fix typo.
3812 2018-08-02 Richard Biener <rguenther@suse.de>
3814 PR tree-optimization/86816
3815 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
3816 which checks for value availability before querying it.
3817 (gvn_uses_equal): Use it.
3818 (same_succ_hash): Likewise.
3819 (gimple_equal_p): Likewise.
3821 2018-08-02 Nick Clifton <nickc@redhat.com>
3824 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3825 Define to speculation_safe_value_not_needed.
3828 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3829 Define to speculation_safe_value_not_needed.
3832 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3833 Define to speculation_safe_value_not_needed.
3836 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3837 Define to speculation_safe_value_not_needed.
3840 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3841 Define to speculation_safe_value_not_needed.
3844 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3845 Define to speculation_safe_value_not_needed.
3848 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3849 Define to speculation_safe_value_not_needed.
3852 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3853 Define to speculation_safe_value_not_needed.
3856 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
3857 speculation_safe_value_not_needed.
3860 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
3861 to speculation_safe_value_not_needed.
3863 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
3865 * doc/md.texi: Expand the documentation of instruction names
3866 to mention port-local uses. Document '@' in pattern names.
3867 * read-md.h (overloaded_instance, overloaded_name): New structs.
3869 (md_reader::handle_overloaded_name): New member function.
3870 (md_reader::get_overloads): Likewise.
3871 (md_reader::m_first_overload): New member variable.
3872 (md_reader::m_next_overload_ptr): Likewise.
3873 (md_reader::m_overloads_htab): Likewise.
3874 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
3875 m_next_overload_ptr and m_overloads_htab.
3876 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
3877 (get_mode_token, get_code_token, get_int_token): New functions.
3878 (map_attr_string): Add an optional argument that passes back
3879 the associated iterator.
3880 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
3881 (md_reader::handle_overloaded_name, add_overload_instance): New
3883 (apply_iterators): Handle '@' names. Report an error if '@'
3884 is used without iterators.
3885 (initialize_iterators): Initialize the new iterator_group fields.
3886 * genopinit.c (handle_overloaded_code_for)
3887 (handle_overloaded_gen): New functions.
3888 (main): Use them to print declarations of maybe_code_for_* and
3889 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
3890 * genemit.c (print_overload_arguments, print_overload_test)
3891 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
3892 (main): Use it to print definitions of maybe_code_for_* and
3893 maybe_gen_* functions.
3894 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
3895 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
3896 instead of explicit mode checks.
3897 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
3898 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
3899 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
3900 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
3901 (aarch64_expand_compare_and_swap): Likewise
3902 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
3903 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
3904 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
3905 (aarch64_constant_pool_reload_icode): Delete.
3906 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
3907 instead of aarch64_constant_pool_reload_icode. Use
3908 code_for_aarch64_reload_mov instead of explicit mode checks.
3909 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
3910 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
3911 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
3912 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
3913 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
3914 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
3915 (aarch64_atomic_load_op_code): Delete.
3916 (aarch64_emit_atomic_load_op): Likewise.
3917 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
3918 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
3919 instead of aarch64_emit_atomic_load_op.
3920 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
3921 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
3922 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
3923 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
3924 character before the pattern name.
3925 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
3926 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
3927 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
3928 (aarch64_frecps<mode>): Likewise.
3929 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
3930 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
3931 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
3932 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
3933 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
3935 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
3937 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
3938 Allow HFmode constants if TARGET_FP_F16INST.
3940 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
3943 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
3944 No longer check last store for clobber of address register.
3946 2018-08-02 Martin Liska <mliska@suse.cz>
3948 PR gcov-profile/86817
3949 * gcov.c (process_all_functions): New function.
3951 (process_file): Move functions processing to
3952 process_all_functions.
3954 2018-08-02 David Malcolm <dmalcolm@redhat.com>
3956 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
3957 "const" to the "gimple *" and "rtx_insn *" parameters.
3958 * dumpfile.h (dump_user_location_t::dump_user_location_t):
3960 (dump_location_t::dump_location_t): Likewise.
3962 2018-08-01 Martin Sebor <msebor@redhat.com>
3964 PR tree-optimization/86650
3965 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
3966 rather than a "gcall *". Directly pass the data of interest
3967 to percent_K_format, rather than building a temporary CALL_EXPR
3969 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
3970 (gimple_fold_builtin_strncat): Adjust.
3971 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
3972 gcall* argument with gimple*.
3973 * gimple-ssa-warn-restrict.c (check_call): Same.
3974 (wrestrict_dom_walker::before_dom_children): Same.
3975 (builtin_access::builtin_access): Same.
3976 (check_bounds_or_overlap): Same
3977 (maybe_diag_overlap): Same.
3978 (maybe_diag_offset_bounds): Same.
3979 * tree-diagnostic.c (default_tree_printer): Move usage of
3980 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
3982 * tree-pretty-print.c (percent_K_format): Add argument.
3983 * tree-pretty-print.h: Add argument.
3984 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
3985 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
3986 (maybe_diag_stxncpy_trunc): Same.
3987 (handle_builtin_stxncpy): Same.
3988 (handle_builtin_strcat): Same.
3990 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3992 * match.pd: Optimise pointer range checks.
3994 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3996 PR tree-optimization/86758
3997 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
3998 to remove pattern statements.
4000 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4002 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
4003 result of dfs_enumerate_from when constructing stmt_vec_infos,
4004 instead of additionally calling get_loop_body.
4006 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4008 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
4010 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
4011 When creating an iv, assert that the step is not known to be zero.
4012 (vect_setup_realignment): Update call accordingly.
4013 * tree-vect-stmts.c (vectorizable_store): Likewise.
4014 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
4016 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4018 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
4019 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
4020 (vectorizable_reduction): Likewise.
4021 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4022 (vect_detect_hybrid_slp_stmts): Likewise.
4023 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
4025 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
4027 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
4028 (wide_int_set_zero_nonzero_bits): ...this.
4029 (zero_nonzero_bits_from_vr): Rename to...
4030 (vrp_set_zero_nonzero_bits): ...this.
4031 (extract_range_from_multiplicative_op_1): Abstract wide int
4033 (wide_int_range_multiplicative_op): ...here.
4034 (extract_range_from_binary_expr_1): Extract wide int binary
4035 operations into their own functions.
4036 (wide_int_range_lshift): New.
4037 (wide_int_range_can_optimize_bit_op): New.
4038 (wide_int_range_shift_undefined_p): New.
4039 (wide_int_range_bit_xor): New.
4040 (wide_int_range_bit_ior): New.
4041 (wide_int_range_bit_and): New.
4042 (wide_int_range_trunc_mod): New.
4043 (extract_range_into_wide_ints): New.
4044 (vrp_shift_undefined_p): New.
4045 (extract_range_from_multiplicative_op): New.
4046 (vrp_can_optimize_bit_op): New.
4047 * tree-vrp.h (value_range::dump): New.
4048 (wide_int_range_multiplicative_op): New.
4049 (wide_int_range_lshift):New.
4050 (wide_int_range_shift_undefined_p): New.
4051 (wide_int_range_bit_xor): New.
4052 (wide_int_range_bit_ior): New.
4053 (wide_int_range_bit_and): New.
4054 (wide_int_range_trunc_mod): New.
4055 (zero_nonzero_bits_from_bounds): Rename to...
4056 (wide_int_set_zero_nonzero_bits): ...this.
4057 (zero_nonzero_bits_from_vr): Rename to...
4058 (vrp_set_zero_nonzero_bits): ...this.
4059 (range_easy_mask_min_max): Rename to...
4060 (wide_int_range_can_optimize_bit_op): this.
4061 * vr-values.c (simplify_bit_ops_using_ranges): Rename
4062 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
4064 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4066 * tree-vectorizer.h (vect_orig_stmt): New function.
4067 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
4068 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
4069 (vect_create_epilog_for_reduction): Likewise.
4070 (vectorizable_live_operation): Likewise.
4071 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
4072 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
4073 * tree-vect-stmts.c (vectorizable_call): Likewise.
4074 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
4076 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4078 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
4080 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
4081 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
4082 (vect_transform_loop): Likewise.
4083 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
4085 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4087 * tree-vectorizer.h (vect_schedule_slp): Return void.
4088 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
4089 (vect_schedule_slp): Likewise.
4091 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4093 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
4095 (vect_transform_loop): Update calls accordingly. Schedule SLP
4096 instances before the main loop, if any exist.
4098 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4100 PR tree-optimization/86749
4101 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
4102 If the lhs is used in a COND_EXPR, check that it is being used
4103 as the "then" or "else" value.
4105 2018-08-01 Tom de Vries <tdevries@suse.de>
4108 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
4109 speculation_safe_value_not_needed.
4111 2018-08-01 Richard Biener <rguenther@suse.de>
4113 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
4116 2018-08-01 Martin Liska <mliska@suse.cz>
4118 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
4119 format how successful transformation is dumped.
4120 (gimple_mod_pow2_value_transform): Likewise.
4121 (gimple_mod_subtract_transform): Likewise.
4122 (gimple_stringops_transform): Likewise.
4124 2018-08-01 Martin Liska <mliska@suse.cz>
4127 * value-prof.c (interesting_stringop_to_profile_p):
4128 Simplify the code and add BUILT_IN_MEMMOVE.
4129 (gimple_stringops_transform): Likewise.
4131 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
4133 * config/aarch64/aarch64-simd.md
4134 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
4135 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
4136 use GPI iterator instead of SI mode.
4138 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
4140 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
4141 rs6000_speculation_barrier.
4142 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
4143 new barrier pattern name.
4145 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
4147 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
4148 (speculation_barrier): New insn.
4150 2018-08-01 Richard Biener <rguenther@suse.de>
4153 * graphite.h: Include isl/id.h and isl/space.h to allow build
4156 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
4159 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4160 mode for COFF targets.
4161 * defaults.h (TARGET_COFF): Define.
4162 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
4163 TARGET_COFF): Define.
4164 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
4165 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
4166 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
4168 2018-07-31 Alexandre Oliva <oliva@adacore.com>
4170 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
4171 * gimple-streamer-out.c (output_bb): Save it.
4172 * lto-streamer-in.c (input_struct_function_base): Restore
4173 instance discriminator if available. Create map on demand.
4174 * lto-streamer-out.c (output_struct_function_base): Save it if
4176 * final.c (decl_to_instance_map): Document LTO strategy.
4178 2018-07-31 Alexandre Oliva <oliva@adacore.com>
4179 Olivier Hainque <hainque@adacore.com>
4181 * debug.h (decl_to_instance_map_t): New type.
4182 (decl_to_instance_map): Declare.
4183 (maybe_create_decl_to_instance_map): New inline function.
4184 * final.c (bb_discriminator, last_bb_discriminator): New statics,
4185 to track basic block discriminators.
4186 (final_start_function_1): Initialize them.
4187 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
4189 (decl_to_instance_map): New variable.
4190 (map_decl_to_instance, maybe_set_discriminator): New functions.
4191 (notice_source_line): Set discriminator.
4193 2018-07-31 Ian Lance Taylor <iant@golang.org>
4195 * targhooks.c (default_have_speculation_safe_value): Add
4198 2018-07-31 David Malcolm <dmalcolm@redhat.com>
4200 * dump-context.h: Include "pretty-print.h".
4201 (dump_context::refresh_dumps_are_enabled): New decl.
4202 (dump_context::emit_item): New decl.
4203 (class dump_context): Add fields "m_test_pp" and
4205 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
4206 (temp_dump_context::get_dumped_text): New decl.
4207 (class temp_dump_context): Add field "m_pp".
4208 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
4209 (dump_context::refresh_dumps_are_enabled): ...and add a test for
4211 (set_dump_file): Update for above change.
4212 (set_alt_dump_file): Likewise.
4213 (dump_loc): New overload, taking a pretty_printer *.
4214 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
4215 to any test pretty-printer.
4216 (make_item_for_dump_gimple_stmt): New function, adapted from
4217 optinfo::add_gimple_stmt.
4218 (dump_context::dump_gimple_stmt): 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_gimple_expr): New function, adapted from
4222 optinfo::add_gimple_expr.
4223 (dump_context::dump_gimple_expr): Call it, and use the result,
4224 eliminating the direct usage of dump_file and alt_dump_file in
4225 favor of indirectly using them via emit_item.
4226 (make_item_for_dump_generic_expr): New function, adapted from
4228 (dump_context::dump_generic_expr): 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_printf_va): New function, adapted from
4232 optinfo::add_printf_va.
4233 (make_item_for_dump_printf): New function.
4234 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
4235 and use the result, eliminating the direct usage of dump_file and
4236 alt_dump_file in favor of indirectly using them via emit_item.
4237 (make_item_for_dump_dec): New function.
4238 (dump_context::dump_dec): Call it, and use the result,
4239 eliminating the direct usage of dump_file and alt_dump_file in
4240 favor of indirectly using them via emit_item.
4241 (make_item_for_dump_symtab_node): New function, adapted from
4242 optinfo::add_symtab_node.
4243 (dump_context::dump_symtab_node): Call it, and use the result,
4244 eliminating the direct usage of dump_file and alt_dump_file in
4245 favor of indirectly using them via emit_item.
4246 (dump_context::begin_scope): Reimplement, avoiding direct usage
4247 of dump_file and alt_dump_file in favor of indirectly using them
4249 (dump_context::emit_item): New member function.
4250 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
4251 Set up test pretty-printer on the underlying context. Call
4252 refresh_dumps_are_enabled.
4253 (temp_dump_context::~temp_dump_context): Call
4254 refresh_dumps_are_enabled.
4255 (temp_dump_context::get_dumped_text): New member function.
4256 (selftest::verify_dumped_text): New function.
4257 (ASSERT_DUMPED_TEXT_EQ): New macro.
4258 (selftest::test_capture_of_dump_calls): Run all tests twice, with
4259 and then without optinfo enabled. Add uses of
4260 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
4261 * dumpfile.h: Update comment for the dump_* API.
4262 * optinfo-emit-json.cc
4263 (selftest::test_building_json_from_dump_calls): Update for new
4264 param for temp_dump_context ctor.
4265 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
4266 and "m_owned" field.
4267 (optinfo_item::~optinfo_item): Likewise.
4268 (optinfo::add_item): New member function.
4269 (optinfo::emit): Update comment.
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.h (class dump_context): New forward decl.
4279 (optinfo::add_item): New decl.
4280 (optinfo::add_string): Delete.
4281 (optinfo::add_printf): Delete.
4282 (optinfo::add_printf_va): Delete.
4283 (optinfo::add_gimple_stmt): Delete.
4284 (optinfo::add_gimple_expr): Delete.
4285 (optinfo::add_tree): Delete.
4286 (optinfo::add_symtab_node): Delete.
4287 (optinfo::add_dec): Delete.
4288 (optinfo::add_poly_int): Delete.
4289 (optinfo_item::optinfo_item): Remove "owned" param.
4290 (class optinfo_item): Remove field "m_owned".
4292 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4295 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
4296 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
4297 requested variable alignment.
4298 (expand_one_ssa_partition): Likewise.
4299 (expand_one_var): Likewise.
4301 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4303 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
4304 to speculation_safe_value_not_needed.
4306 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4308 * targhooks.h (speculation_safe_value_not_needed): New prototype.
4309 * targhooks.c (speculation_safe_value_not_needed): New function.
4310 * target.def (have_speculation_safe_value): Update documentation.
4311 * doc/tm.texi: Regenerated.
4313 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4315 * config/aarch64/iterators.md (ALLI_TI): New iterator.
4316 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
4318 (despeculate_copy<ALLI:mode>_insn): New insn.
4319 (despeculate_copyti_insn): New insn.
4320 (despeculate_simple<ALLI:mode>): New insn
4321 (despeculate_simpleti): New insn.
4322 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
4324 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
4325 aarch64_speculation_safe_value.
4326 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
4328 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4330 * config/aarch64/aarch64-speculation.cc: New file.
4331 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
4332 before pass_reorder_blocks.
4333 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
4335 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
4336 X14 and X15 when tracking speculation.
4337 * config/aarch64/aarch64.md (register name constants): Add
4338 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
4339 (unspec): Add UNSPEC_SPECULATION_TRACKER.
4340 (speculation_barrier): New insn attribute.
4341 (cmp<mode>): Allow SP in comparisons.
4342 (speculation_tracker): New insn.
4343 (speculation_barrier): Add speculation_barrier attribute.
4344 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
4345 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
4346 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
4348 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4350 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
4351 aarch64_track_speculation is true.
4352 (tb<optab><mode>1): Likewise.
4353 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
4354 generate CB[N]Z when tracking speculation.
4355 (aarch64_split_compare_and_swap): Likewise.
4356 (aarch64_split_atomic_op): Likewise.
4358 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4360 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
4362 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4364 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
4365 (speculation_barrier): New insn.
4367 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4369 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
4370 * config/arm/arm.md (speculation_barrier): New expand.
4371 (speculation_barrier_insn): New pattern.
4373 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4375 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
4376 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
4377 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
4378 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
4380 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
4381 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
4382 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
4383 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
4384 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
4385 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
4386 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
4387 * builtins.c (expand_speculation_safe_value): New function.
4388 (expand_builtin): Call it.
4389 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
4390 * doc/extend.texi: Document __builtin_speculation_safe_value.
4391 * doc/md.texi: Document "speculation_barrier" pattern.
4392 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
4393 TARGET_HAVE_SPECULATION_SAFE_VALUE.
4394 * doc/tm.texi: Regenerated.
4395 * target.def (have_speculation_safe_value, speculation_safe_value): New
4397 * targhooks.c (default_have_speculation_safe_value): New function.
4398 (default_speculation_safe_value): New function.
4399 * targhooks.h (default_have_speculation_safe_value): Add prototype.
4400 (default_speculation_safe_value): Add prototype.
4402 2018-07-31 David Malcolm <dmalcolm@redhat.com>
4404 * dump-context.h (dump_context::dump_loc): New decl.
4405 * dumpfile.c (dump_context::dump_loc): New member function.
4406 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
4407 and dump_gimple_stmt.
4408 (dump_context::dump_gimple_expr_loc): Likewise, using
4410 (dump_context::dump_generic_expr_loc): Likewise, using
4412 (dump_context::dump_printf_loc_va): Likewise, using
4414 (dump_context::begin_scope): Explicitly using the global function
4415 "dump_loc", rather than the member function.
4417 2018-07-31 Martin Sebor <msebor@redhat.com>
4419 PR tree-optimization/86741
4420 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
4422 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
4424 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
4425 depend on whether prefetch instructions will be emitted or not.
4426 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
4427 will be emitted or not.
4428 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
4429 (TARGET_SETMEM_PFD): New macros.
4431 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4433 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
4434 (NULL_STMT_VEC_INFO): Delete.
4435 (stmt_vec_info::operator*): Likewise.
4436 (stmt_vec_info::operator gimple *): Likewise.
4437 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
4438 of NULL_STMT_VEC_INFO.
4439 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4440 (vect_reassociating_reduction_p): Likewise.
4441 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
4442 (vectorizable_store): Likewise.
4443 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
4444 (vec_info::free_stmt_vec_infos): Likewise.
4446 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4448 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
4449 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
4451 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4453 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
4454 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
4455 (vec_info::free_stmt_vec_info): New private member functions.
4456 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
4457 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
4458 * tree-parloops.c (gather_scalar_reductions): Remove calls to
4459 set_stmt_vec_info_vec and free_stmt_vec_infos.
4460 * tree-vect-loop.c (_loop_vec_info): Remove call to
4461 set_stmt_vec_info_vec.
4462 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
4463 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
4464 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
4465 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
4466 (vec_info::free_stmt_vec_info): ...these new functions. Remove
4467 assignments in {vec_info::,}new_stmt_vec_info that are redundant
4468 with the clearing in the xcalloc.
4469 (stmt_vec_info_vec): Delete.
4470 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
4471 (vectorize_loops): Likewise.
4472 (vec_info::~vec_info): Remove argument from call to
4473 free_stmt_vec_infos.
4474 (vec_info::add_stmt): Remove vinfo argument from call to
4477 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4479 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
4480 rather than a gimple stmt.
4481 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
4482 information for pattern statements when passed the original
4483 statement; instead wait to be passed the pattern statement itself.
4484 Don't call set_vinfo_for_stmt here.
4485 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
4486 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
4487 stmt_vec_infos here.
4488 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
4489 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
4490 stmt_vec_infos entry.
4492 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4494 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
4495 * tree-vectorizer.c (vec_info::replace_stmt): New function.
4496 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
4497 * tree-vect-stmts.c (vectorizable_call): Likewise.
4498 (vectorizable_simd_clone_call): Likewise.
4500 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4502 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
4503 * tree-vectorizer.c (vec_info::remove_stmt): New function.
4504 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
4505 * tree-vect-loop.c (vect_transform_loop): Likewise.
4506 * tree-vect-slp.c (vect_schedule_slp): Likewise.
4507 * tree-vect-stmts.c (vect_remove_stores): Likewise.
4509 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4511 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
4512 (vect_dr_stmt): Delete.
4513 * tree-vectorizer.c (vec_info::lookup_dr): New function.
4514 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
4516 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
4517 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
4518 (vect_verify_datarefs_alignment, vect_peeling_supportable)
4519 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
4520 (vect_analyze_data_refs): Likewise.
4521 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
4523 (vect_find_same_alignment_drs): Likewise.
4524 (vect_slp_analyze_node_dependences): Update calls accordingly.
4525 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
4526 instead of DR_VECT_AUX.
4527 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
4528 of a vector data references. Use vec_info::lookup_dr instead of
4530 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
4531 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
4532 instead of DR_VECT_AUX.
4534 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4536 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
4538 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
4540 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4541 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4542 (vect_gen_prolog_loop_niters): Likewise.
4544 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4546 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
4547 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
4548 (vect_known_alignment_in_bytes, vect_dr_behavior)
4549 (vect_get_scalar_dr_size): Take references as dr_vec_infos
4550 instead of data_references. Update calls to other routines for
4551 which the same change has been made.
4552 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
4553 dr_vec_infos instead of stmt_vec_infos.
4554 (vect_analyze_data_ref_dependence): Update call accordingly.
4555 (vect_slp_analyze_data_ref_dependence)
4556 (vect_record_base_alignments): Use DR_VECT_AUX.
4557 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
4558 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
4559 (vector_alignment_reachable_p, vect_get_data_access_cost)
4560 (vect_peeling_supportable, vect_analyze_group_access_1)
4561 (vect_analyze_group_access, vect_analyze_data_ref_access)
4562 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
4563 (vect_compile_time_alias, vect_small_gap_p)
4564 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
4565 (vect_supportable_dr_alignment): Take references as dr_vec_infos
4566 instead of data_references. Update calls to other routines for
4567 which the same change has been made.
4568 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
4569 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
4570 (vect_slp_analyze_and_verify_node_alignment)
4571 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
4572 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
4573 (vect_setup_realignment): Use dr_vec_infos. Update calls after
4575 (_vect_peel_info::dr): Replace with...
4576 (_vect_peel_info::dr_info): ...this new field.
4577 (vect_peeling_hash_get_most_frequent)
4578 (vect_peeling_hash_choose_best_peeling): Update accordingly.
4579 (vect_peeling_hash_get_lowest_cost):
4580 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
4581 routines for which the same change has been made.
4582 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
4584 * tree-vect-loop-manip.c (get_misalign_in_elems)
4585 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
4587 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4588 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
4589 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
4590 (get_group_load_store_type, get_negative_load_store_type)
4591 (vect_get_data_ptr_increment, vectorizable_store)
4592 (vectorizable_load): Likewise.
4593 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
4594 Update calls to other routines for which the same change has been made.
4596 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4598 * tree-vectorizer.h (vec_info::move_dr): New member function.
4599 (dataref_aux): Rename to...
4600 (dr_vec_info): ...this and add "dr" and "stmt" fields.
4601 (_stmt_vec_info::dr_aux): Update accordingly.
4602 (_stmt_vec_info::data_ref_info): Delete.
4603 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
4604 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
4605 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
4606 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
4608 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
4609 (STMT_VINFO_DR_INFO): New macro.
4610 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
4611 (set_dr_misalignment): Update after rename of dataref_aux.
4612 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
4613 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
4614 initialization of STMT_VINFO_DATA_REF.
4615 * tree-vectorizer.c (vec_info::move_dr): New function.
4616 * tree-vect-patterns.c (vect_recog_bool_pattern)
4617 (vect_recog_mask_conversion_pattern)
4618 (vect_recog_gather_scatter_pattern): Use it.
4619 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
4620 the "dr" and "stmt" fields of dr_vec_info instead of
4621 STMT_VINFO_DATA_REF.
4623 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4625 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
4626 (is_pattern_stmt_p): Use it.
4627 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
4628 on pattern statements.
4630 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4632 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
4633 original stmt as a stmt_vec_info rather than a gimple stmt.
4634 (vect_pattern_recog_1): Take the statement directly as a
4635 stmt_vec_info, rather than via a gimple_stmt_iterator.
4636 Update call to vect_mark_pattern_stmts.
4637 (vect_pattern_recog): Update calls accordingly.
4639 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4641 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
4642 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
4643 a vect_def_type for the first argument.
4644 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
4645 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
4646 operand if it isn't defined by a vectorized statement.
4647 (vect_build_gather_load_calls): Remove the mask_dt argument and
4648 update calls to vect_get_vec_def_for_stmt_copy.
4649 (vectorizable_bswap): Likewise the dt argument.
4650 (vectorizable_call): Update calls to vectorizable_bswap and
4651 vect_get_vec_def_for_stmt_copy.
4652 (vectorizable_simd_clone_call, vectorizable_assignment)
4653 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
4654 (vectorizable_comparison): Update calls to
4655 vect_get_vec_def_for_stmt_copy.
4656 (vectorizable_store): Likewise. Remove now-unnecessary calls to
4658 (vect_get_loop_based_defs): Remove dt argument and update call
4659 to vect_get_vec_def_for_stmt_copy.
4660 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
4661 and vect_get_vec_def_for_stmt_copy.
4662 (vectorizable_load): Update calls to vect_build_gather_load_calls
4663 and vect_get_vec_def_for_stmt_copy.
4664 * tree-vect-loop.c (vect_create_epilog_for_reduction)
4665 (vectorizable_reduction, vectorizable_live_operation): Update calls
4666 to vect_get_vec_def_for_stmt_copy.
4668 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4670 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
4671 and gimple stmt arguments with a stmt_vec_info.
4672 (vect_record_base_alignments): Update calls accordingly.
4673 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
4674 and gimple stmt arguments with a stmt_vec_info.
4675 (vect_build_slp_tree_1): Remove vinfo argument and update call
4676 to vect_record_max_nunits.
4677 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
4678 and vect_record_max_nunits.
4680 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4682 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
4683 file and take a stmt_vec_info instead of a gimple stmt.
4684 (supportable_widening_operation, vect_finish_replace_stmt)
4685 (vect_finish_stmt_generation, vect_get_store_rhs)
4686 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
4687 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
4688 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
4689 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
4690 (vect_create_data_ref_ptr, bump_vector_ptr)
4691 (vect_permute_store_chain, vect_setup_realignment)
4692 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
4693 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
4694 (vectorizable_reduction, vectorizable_induction)
4695 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
4696 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
4697 than gimple stmts as arguments.
4698 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
4699 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
4700 (can_group_stmts_p, vect_check_gather_scatter)
4701 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
4702 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
4703 (vect_permute_load_chain, vect_shift_permute_load_chain)
4704 (vect_transform_grouped_load)
4705 (vect_record_grouped_load_vectors): Likewise.
4706 * tree-vect-loop.c (vect_fixup_reduc_chain)
4707 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
4708 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
4709 (vectorizable_reduction, vectorizable_induction)
4710 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
4711 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
4712 (vect_get_load_store_mask): Likewise.
4713 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
4714 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
4715 * tree-vect-stmts.c (vect_mark_relevant)
4716 (is_simple_and_all_uses_invariant)
4717 (exist_non_indexing_operands_for_use_p, process_use)
4718 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
4719 (vect_get_vec_def_for_operand, vect_get_vec_defs)
4720 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
4721 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
4722 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
4723 (get_negative_load_store_type, get_load_store_type)
4724 (vect_check_load_store_mask, vect_check_store_rhs)
4725 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
4726 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4727 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4728 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4729 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
4730 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
4731 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
4732 (supportable_widening_operation): Likewise.
4734 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4736 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
4737 a stmt_vec_info instead of a gcall.
4738 (vect_check_gather_scatter): Update call accordingly.
4739 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
4741 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
4742 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
4743 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
4744 instead of a gimple stmt.
4745 (vect_transform_loop): Update calls accordingly.
4746 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
4747 stmt_vec_infos instead of gimple stmts.
4748 (vect_analyze_slp_instance): Update use accordingly.
4749 * tree-vect-stmts.c (read_vector_array, write_vector_array)
4750 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
4751 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
4752 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
4753 (vect_gen_widened_results_half, vect_get_loop_based_defs)
4754 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
4755 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
4756 down to subroutines.
4758 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4760 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
4761 of the worklist from a vector of gimple stmts to a vector of
4763 * tree-vect-stmts.c (vect_mark_relevant, process_use)
4764 (vect_mark_stmts_to_be_vectorized): Likewise
4766 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4768 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
4769 statement before passing it to vect_analyze_stmt.
4770 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
4771 the chain of phi vector definitions. Track the exit phi via its
4773 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
4774 STMT_VINFO_REDUC_DEF.
4775 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
4776 stmt_vec_infos to handle the statement chains.
4777 (vect_get_slp_defs): Record the first statement in the node
4778 using a stmt_vec_info.
4779 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
4780 statements here and pass their stmt_vec_info down to subroutines.
4781 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
4782 down to vect_finish_stmt_generation.
4783 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
4784 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
4785 stmt_vec_infos to subroutines.
4786 (vect_remove_stores): Use stmt_vec_infos to handle the statement
4789 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4791 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
4792 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
4793 (vect_permute_store_chain, vect_setup_realignment)
4794 (vect_permute_load_chain, vect_shift_permute_load_chain)
4795 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
4796 stmts internally, and when passing values to other vectorizer routines.
4797 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
4798 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
4799 (vect_analyze_loop_operations, get_initial_def_for_reduction)
4800 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
4801 (vectorizable_reduction, vectorizable_induction)
4802 (vectorizable_live_operation, vect_transform_loop_stmt)
4803 (vect_transform_loop): Likewise.
4804 * tree-vect-patterns.c (vect_reassociating_reduction_p)
4805 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
4806 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
4807 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4808 (vect_slp_analyze_node_operations_1): Likewise.
4809 * tree-vect-stmts.c (vect_mark_relevant, process_use)
4810 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
4811 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
4812 (vect_finish_stmt_generation_1, get_group_load_store_type)
4813 (get_load_store_type, vect_build_gather_load_calls)
4814 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4815 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4816 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4817 (vectorizable_store, vectorizable_load, vectorizable_condition)
4818 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
4819 (supportable_widening_operation): Likewise.
4820 (vect_get_vector_types_for_stmt): Likewise.
4821 * tree-vectorizer.h (vect_dr_behavior): Likewise.
4823 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4825 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4826 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
4827 (vect_permute_store_chain, vect_permute_load_chain)
4828 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
4829 repeated stmt_vec_info lookups.
4830 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
4831 (vect_update_ivs_after_vectorizer): Likewise.
4832 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
4833 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
4834 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
4835 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4836 (vect_bb_slp_scalar_cost): Likewise.
4837 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
4839 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4841 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
4842 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
4843 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
4844 to get gassigns and gcalls, rather than operating on generc gimple
4846 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
4847 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
4848 (vectorizable_load, vect_analyze_stmt): Likewise.
4849 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
4851 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4853 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
4854 return stmt_vec_infos rather than gimple stmts. Do not accept
4856 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
4858 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
4859 Update use of get_later_stmt.
4860 (vect_get_constant_vectors): Update call accordingly.
4861 (vect_schedule_slp_instance): Likewise
4862 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
4863 (vect_slp_analyze_instance_dependence): Likewise.
4864 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
4866 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4868 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
4869 (stmt_info_for_cost::stmt_info): ...this new field.
4870 (add_stmt_costs): Update accordingly.
4871 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
4872 (vect_get_known_peeling_cost): Likewise.
4873 (vect_estimate_min_profitable_iters): Likewise.
4874 * tree-vect-stmts.c (record_stmt_cost): Likewise.
4876 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4878 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
4879 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4880 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
4882 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
4884 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4886 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
4887 a gimple stmt to a stmt_vec_info.
4888 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
4890 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4892 * tree-vectorizer.h (vec_info::grouped_stores): Change from
4893 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4894 (_loop_vec_info::reduction_chains): Likewise.
4895 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
4897 * tree-vect-slp.c (vect_analyze_slp): Likewise.
4899 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4901 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
4902 a gimple stmt to a stmt_vec_info.
4903 (_stmt_vec_info::next_element): Likewise.
4904 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
4905 (vect_slp_analyze_and_verify_node_alignment)
4906 (vect_analyze_group_access_1, vect_analyze_group_access)
4907 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
4908 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
4909 (vect_supportable_dr_alignment): Update accordingly.
4910 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
4911 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
4912 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
4913 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
4914 * tree-vect-slp.c (vect_build_slp_tree_1)
4915 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
4916 (vect_split_slp_store_group, vect_analyze_slp_instance)
4917 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
4918 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
4919 (get_group_load_store_type, get_load_store_type)
4920 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
4921 (vect_transform_stmt, vect_remove_stores): Likewise.
4923 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4925 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
4927 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4928 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
4929 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
4930 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
4931 (vector_alignment_reachable_p, vect_get_data_access_cost)
4932 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
4933 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
4934 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
4935 (vect_analyze_group_access_1, vect_analyze_group_access)
4936 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
4937 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
4938 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
4939 result of vect_dr_stmt and use the stmt_vec_info instead of
4940 the associated gimple stmt.
4941 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4942 (vect_gen_prolog_loop_niters): Likewise.
4943 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4945 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4947 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
4948 to a vec<stmt_vec_info>.
4949 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
4950 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
4952 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
4953 to a vec<stmt_vec_info>.
4954 (bst_traits::value_type, bst_traits::value_type): Likewise.
4955 (bst_traits::hash): Update accordingly.
4956 (vect_get_and_check_slp_defs): Change the stmts parameter from
4957 a vec<gimple *> to a vec<stmt_vec_info>.
4958 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
4959 (vect_build_slp_tree): Likewise.
4960 (vect_build_slp_tree_2): Likewise. Update uses of
4961 SLP_TREE_SCALAR_STMTS.
4962 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
4963 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
4964 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
4965 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
4966 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
4967 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
4968 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
4969 (vect_get_constant_vectors, vect_get_slp_defs)
4970 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
4971 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
4972 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
4973 instead of gimple stmts.
4974 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
4975 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
4976 (vect_slp_analyze_instance_dependence): Update uses of
4977 SLP_TREE_SCALAR_STMTS.
4978 (vect_slp_analyze_and_verify_node_alignment): Likewise.
4979 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
4980 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
4981 (get_initial_defs_for_reduction): Likewise.
4982 (vect_create_epilog_for_reduction): Likewise.
4983 (vectorize_fold_left_reduction): Likewise.
4984 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
4985 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
4986 (can_vectorize_live_stmts): Likewise.
4988 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4990 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
4991 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4992 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
4994 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
4996 (gather_scalar_reductions): Update after above interface changes.
4997 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
4998 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
5000 (vect_force_simple_reduction): Likewise.
5001 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
5002 LOOP_VINFO_REDUCTIONS.
5003 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5005 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5007 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
5008 a gimple stmt to a stmt_vec_info.
5009 * tree-vect-loop.c (vect_active_double_reduction_p)
5010 (vect_force_simple_reduction, vectorizable_reduction): Update
5013 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5015 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
5016 vec<gimple *> to a vec<stmt_vec_info>.
5017 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
5018 the reduction_phis argument from a vec<gimple *> to a
5020 (vectorizable_reduction): Likewise the phis local variable that
5021 is passed to vect_create_epilog_for_reduction. Update for new type
5022 of SLP_TREE_VEC_STMTS.
5023 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
5024 (vectorizable_live_operation): Likewise.
5025 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
5026 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
5028 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5030 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
5031 a gimple stmt to a stmt_vec_info.
5032 (vectorizable_condition, vectorizable_live_operation)
5033 (vectorizable_reduction, vectorizable_induction): Pass back the
5034 vectorized statement as a stmt_vec_info.
5035 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
5036 use of STMT_VINFO_VEC_STMT.
5037 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
5038 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
5039 as stmt_vec_infos rather than gimple stmts.
5040 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
5042 (vectorizable_live_operation): Likewise.
5043 (vectorizable_reduction, vectorizable_induction): Likewise,
5044 updating use of STMT_VINFO_VEC_STMT.
5045 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
5046 of STMT_VINFO_VEC_STMT.
5047 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
5048 (vectorizable_simd_clone_call, vectorizable_conversion)
5049 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
5050 (vectorizable_store, vectorizable_load, vectorizable_condition)
5051 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
5052 from a gimple stmt to a stmt_vec_info.
5053 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
5054 pointer to a stmt_vec_info to the vectorizable_* routines.
5056 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5058 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
5059 a gimple stmt to a stmt_vec_info.
5060 (is_pattern_stmt_p): Update accordingly.
5061 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
5062 (vect_record_grouped_load_vectors): Likewise.
5063 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
5064 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
5065 (vect_model_reduction_cost): Likewise.
5066 (vect_create_epilog_for_reduction): Likewise.
5067 (vectorizable_reduction, vectorizable_induction): Likewise.
5068 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5069 Return the stmt_vec_info for the pattern statement.
5070 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
5071 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
5072 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
5073 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
5074 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
5075 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
5076 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
5077 (free_stmt_vec_info, vect_is_simple_use): Likewise.
5079 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5081 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
5082 (vect_finish_stmt_generation): Likewise.
5083 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
5084 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
5085 (vect_build_gather_load_calls): Use the return value of the above
5086 functions instead of a separate call to vinfo_for_stmt. Use narrow
5087 scopes for the input gimple stmt and wider scopes for the associated
5088 stmt_vec_info. Use vec_info::lookup_def when setting these
5089 stmt_vec_infos from an SSA_NAME definition.
5090 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
5091 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
5092 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
5093 (vectorizable_store, vectorizable_load, vectorizable_condition)
5094 (vectorizable_comparison): Likewise.
5095 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
5096 (vectorizable_reduction): Likewise.
5098 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5100 * tree-vectorizer.h (vect_is_simple_use): Add an optional
5101 stmt_vec_info * parameter before the optional gimple **.
5102 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
5103 (process_use, vect_get_vec_def_for_operand_1): Update callers.
5104 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
5105 * tree-vect-loop.c (vectorizable_reduction): Likewise.
5106 (vectorizable_live_operation): Likewise.
5107 * tree-vect-patterns.c (type_conversion_p): Likewise.
5108 (vect_look_through_possible_promotion): Likewise.
5109 (vect_recog_rotate_pattern): Likewise.
5110 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
5112 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5114 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
5115 a typedef to a wrapper class.
5116 (NULL_STMT_VEC_INFO): New macro.
5117 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
5118 (stmt_vec_info::operator*): New function.
5119 (stmt_vec_info::operator gimple *): Likewise.
5120 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
5121 (add_stmt_costs): Likewise.
5122 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
5123 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
5124 (vect_get_known_peeling_cost): Likewise.
5125 (vect_estimate_min_profitable_iters): Likewise.
5126 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5127 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
5128 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
5129 (vectorizable_store, free_stmt_vec_infos): Likewise.
5130 (new_stmt_vec_info): Change return type of xcalloc to
5133 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5135 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
5136 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
5137 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
5138 a single_imm_use-based sequence.
5139 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
5141 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5143 * tree-vectorizer.h (vec_info::lookup_def): Declare.
5144 * tree-vectorizer.c (vec_info::lookup_def): New function.
5145 * tree-vect-patterns.c (vect_get_internal_def): Use it.
5146 (vect_widened_op_tree): Likewise.
5147 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
5148 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
5149 (vectorizable_reduction): Likewise.
5150 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
5152 (vect_is_slp_reduction): Update calls accordingly. Use
5153 vec_info::lookup_def.
5154 (vect_is_simple_reduction): Likewise
5155 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
5157 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5159 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
5160 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
5161 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
5163 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
5164 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
5165 (vect_update_vf_for_slp, vect_analyze_loop_operations)
5166 (vect_is_slp_reduction, vectorizable_induction)
5167 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
5168 * tree-vect-patterns.c (vect_init_pattern_stmt):
5169 (vect_determine_min_output_precision_1, vect_determine_precisions)
5170 (vect_pattern_recog): Likewise.
5171 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
5172 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
5173 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
5174 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
5175 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
5176 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
5177 info field from a loop to a loop_vec_info.
5179 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5181 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
5182 (vec_info::add_stmt): Declare.
5183 * tree-vectorizer.c (vec_info::add_stmt): New function.
5184 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
5185 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
5186 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
5187 (vectorizable_induction): Likewise.
5188 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
5189 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
5190 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
5191 (vectorizable_load): Likewise.
5192 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5193 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
5194 (vect_recog_gather_scatter_pattern): Likewise.
5195 (append_pattern_def_seq): Likewise. Remove a check that is
5196 performed by add_stmt itself.
5198 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5200 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
5201 which make_ssa_name was called with new_stmt before new_stmt
5204 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5206 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
5208 (vect_is_slp_reduction): ...here...
5209 (vect_is_simple_reduction): ...and here. Remove repetition of tests
5210 that are already known to be false.
5212 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5214 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
5215 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
5216 STMT_VINFO_NUM_SLP_USES when it's true.
5217 (vect_free_slp_instance): Add a final_p parameter and pass it to
5219 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
5220 (vect_analyze_slp_instance): Likewise.
5221 (vect_slp_analyze_operations): Likewise.
5222 (vect_slp_analyze_bb_1): Likewise.
5223 * tree-vectorizer.c (vec_info): Likewise.
5224 * tree-vect-loop.c (vect_transform_loop): Likewise.
5226 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5228 * tree-vect-loop.c (vectorizable_reduction): Assert that the
5229 function is not called for second and subsequent members of
5232 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5234 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
5235 cases for nested loops from here to ...
5236 (vect_create_epilog_for_reduction): ...here. Only call
5237 vect_is_simple_use for inner-loop reductions.
5239 2018-07-31 Martin Liska <mliska@suse.cz>
5241 PR gcov-profile/85338
5242 PR gcov-profile/85350
5243 PR gcov-profile/85372
5244 * profile.c (struct location_triplet): New.
5245 (struct location_triplet_hash): Likewise.
5246 (output_location): Do not output a BB that
5247 is already recorded for a line.
5248 (branch_prob): Use streamed_locations.
5250 2018-07-31 Martin Liska <mliska@suse.cz>
5252 PR gcov-profile/85370
5253 * coverage.c (coverage_begin_function): Do not mark target
5254 clones as artificial functions.
5256 2018-07-31 Martin Liska <mliska@suse.cz>
5258 PR gcov-profile/83813
5259 PR gcov-profile/84758
5260 PR gcov-profile/85217
5261 PR gcov-profile/85332
5262 * profile.c (branch_prob): Do not record GOTO expressions
5263 for GIMPLE statements which locations are already streamed.
5265 2018-07-31 Olivier Hainque <hainque@adacore.com>
5267 * gcc.c (handle_spec_function): Accept a soft_matched_part
5268 argument, as do_spec_1. Pass it down to ...
5269 (eval_spec_function): Accept a soft_matched_part argument,
5270 and pass it down to ...
5271 (do_spec_2): Accept a soft_matched_part argument, and pass
5272 it down to do_spec_1.
5273 (do_spec_1): Pass soft_matched_part to handle_spec_function.
5274 (handle_braces): Update call to handle_spec_function.
5275 (driver::set_up_specs): Update calls to do_spec_2.
5276 (compare_debug_dump_opt_spec_function): Likewise.
5277 (compare_debug_self_opt_spec_function): Likewise.
5279 2018-07-31 Olivier Hainque <hainque@adacore.com>
5281 * common.opt (nolibc): New option.
5282 * doc/invoke.texi (Link Options): Document it.
5283 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
5284 * config/alpha/linux.h: Likewise.
5285 * config/arc/elf.h: Likewise.
5286 * config/arm/uclinux-elf.h: Likewise.
5287 * config/arm/unknown-elf.h: Likewise.
5288 * config/avr/avrlibc.h: Likewise.
5289 * config/bfin/bfin.h: Likewise.
5290 * config/bfin/linux.h: Likewise.
5291 * config/bfin/uclinux.h: Likewise.
5292 * config/darwin.h: Likewise.
5293 * config/darwin10.h: Likewise.
5294 * config/darwin12.h: Likewise.
5295 * config/gnu-user.h: Likewise.
5296 * config/lm32/uclinux-elf.h: Likewise.
5297 * config/pa/pa-hpux11.h: Likewise.
5298 * config/pa/pa64-hpux.h: Likewise.
5299 * config/sparc/sparc.h: Likewise.
5301 2018-07-31 Olivier Hainque <hainque@adacore.com>
5303 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
5304 undefined variables.
5306 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
5309 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
5312 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
5314 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
5315 terminated string literal.
5317 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
5319 PR rtl-optimization/85160
5320 * combine.c (is_just_move): New function.
5321 (try_combine): Allow combining two instructions into two if neither of
5322 the original instructions was a move.
5324 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
5327 * doc/extend.texi (Global Register Variables): Discourage use of type
5329 (Local Register Variables): Likewise.
5331 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
5333 PR tree-optimization/86506
5334 * hwint.h (ceil_log2): Resync with hwint.c implementation.
5336 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
5339 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
5340 hard_regno, make sure no insn between `from` and `to` clobbers it.
5342 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
5343 Tom de Vries <tdevries@suse.de>
5345 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
5346 (PTX_DEFAULT_RUNTIME_DIM): ... this.
5347 (nvptx_goacc_validate_dims): Set default worker and gang dims to
5348 PTX_DEFAULT_RUNTIME_DIM.
5349 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
5351 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
5353 * config/pa/pa.c (pa_output_addr_vec): Align address table.
5354 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
5355 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
5357 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
5359 * config/rs6000/constraints.md (wG constraint): Delete, no longer
5361 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
5362 predicate to reflect toc fusion has been deleted.
5363 (toc_fusion_mem_raw): Delete, no longer used.
5364 (toc_fusion_mem_wrapped): Likewise.
5365 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
5367 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
5368 Delete, no longer used.
5369 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
5370 meant to be used for toc fusion.
5371 (rs6000_debug_print_mode): Delete toc fusion debugging.
5372 (rs6000_debug_reg_global): Likewise.
5373 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
5374 fusion and secondary reload support that were never used.
5375 (rs6000_option_override_internal): Delete TOC fusion, that was only
5376 partially defined, and it did not work unless you also used the
5378 (rs6000_legitimate_address_p): Delete TOC fusion support.
5379 (rs6000_opt_masks): Likewise.
5380 (fusion_wrap_memory_address): Delete function, no longer used.
5381 (fusion_split_address); Delete TOC fusion support.
5382 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
5383 longer used with toc fusion being deleted.
5384 (TARGET_TOC_FUSION_FP): Likewise.
5385 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
5387 (toc fusion spliter): Delete TOC fusion support.
5388 (toc_fusionload_<mode>): Likewise.
5389 (toc_fusionload_di): Likewise.
5390 (fusion_gpr_load_<mode>): Delete generator function, this insn no
5391 longer needs to be named. Rename predicate to delete TOC fusion.
5392 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
5393 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
5394 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
5395 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
5396 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
5398 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
5400 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
5401 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
5402 __int128 in built-in function prototypes.
5403 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
5404 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
5406 2018-07-27 Martin Sebor <msebor@redhat.com>
5408 PR tree-optimization/86696
5409 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
5410 types, including enums.
5411 (handle_char_store): Be prepared for the above function to fail.
5413 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
5415 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
5416 when optimization level is lower than 2 or optimize for size.
5418 2018-07-26 Martin Sebor <msebor@redhat.com>
5420 PR tree-optimization/86043
5421 PR tree-optimization/86042
5422 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
5423 (get_string_cst_length): Rename...
5424 (get_min_string_length): ...to this. Add argument.
5425 (handle_char_store): Extend to handle multi-character stores by
5427 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
5428 * tree.h (initializer_zerop): Add argument.
5430 2018-07-26 Jakub Jelinek <jakub@redhat.com>
5433 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
5434 declare target to variables if they have always,{to,from,tofrom} map
5437 2018-07-26 Martin Liska <mliska@suse.cz>
5440 * lto-wrapper.c: Add linker_output as prefix
5441 for ltrans_output_file.
5443 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
5445 PR rtl-optimization/85805
5446 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
5447 value for hard registers if that was written in the same mode.
5449 2018-07-26 Martin Liska <mliska@suse.cz>
5451 PR gcov-profile/86536
5452 * gcov.c (format_gcov): Use printf format %.*f directly
5453 and do not handle special values.
5455 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5457 * common/config/arc/arc-common.c (arc_option_optimization_table):
5458 Update default optimizations for size.
5460 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5462 * config/arc/arc.md (movsf_insn): Add short instruction selection.
5463 * config/arc/constraints.md (CfZ): New constraint.
5464 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
5465 (subsf3_fpu): Likewise.
5466 (cmpsf_fpu): Likewise.
5467 (cmpsf_fpu_uneq): Likewise.
5469 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5471 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
5473 (arc_is_uncached_mem_p): uncached applies to both the variable and
5476 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5478 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
5481 2018-07-25 David Malcolm <dmalcolm@redhat.com>
5483 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
5484 field "m_scopes" from vec to auto_vec.
5486 2018-07-25 Martin Liska <mliska@suse.cz>
5488 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
5491 2018-07-25 Richard Biener <rguenther@suse.de>
5494 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
5495 special wrt context_die late.
5496 (gen_subprogram_die): Re-use DIEs in local scope.
5498 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
5500 PR tree-optimization/86644
5501 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
5503 2018-07-25 Martin Liska <mliska@suse.cz>
5506 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
5507 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
5509 2018-07-25 Martin Liska <mliska@suse.cz>
5512 * params.def: Explain ASan abbreviation and provide
5513 a documentation link.
5515 2018-07-24 Martin Sebor <msebor@redhat.com>
5517 PR tree-optimization/86622
5518 PR tree-optimization/86532
5519 * builtins.h (string_length): Declare.
5520 * builtins.c (c_strlen): Correct handling of non-constant offsets.
5521 (check_access): Be prepared for non-constant length ranges.
5522 (string_length): Make extern.
5523 * expr.c (string_constant): Only handle the minor non-constant
5524 array index. Use string_constant to compute the length of
5525 a generic string constant.
5527 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
5529 PR tree-optimization/86618
5530 * tree-vect-stmts.c (vectorizable_call): Don't take the address
5531 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
5533 2018-07-24 David Malcolm <dmalcolm@redhat.com>
5535 PR tree-optimization/86636
5536 * json.cc (json::object::set): Fix comment. Add assertions.
5537 (json::array::append): Move here from json.h. Add comment and an
5539 (json::string::string): Likewise.
5540 * json.h (json::array::append): Move to json.cc.
5541 (json::string::string): Likewise.
5542 * optinfo-emit-json.cc
5543 (optrecord_json_writer::impl_location_to_json): Assert that we
5544 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
5545 wrapper around it. Expand the location once, rather than three
5547 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
5548 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
5550 (optrecord_json_writer::optinfo_to_json): Likewise, in four
5551 places. Fix some overlong lines.
5553 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
5555 * config/aarch64/aarch64-simd.md
5556 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
5557 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
5558 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
5559 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
5560 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
5561 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
5562 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
5563 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
5564 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
5566 2018-07-24 Jakub Jelinek <jakub@redhat.com>
5569 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
5570 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
5571 and abs_d == d, do the power of two handling if profitable.
5573 2018-07-24 Richard Biener <rguenther@suse.de>
5575 * match.pd: Add BIT_FIELD_REF canonicalizations.
5577 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
5580 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
5582 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
5584 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
5585 terminated STRING_CST object.
5587 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
5589 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
5591 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
5593 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
5594 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
5595 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
5596 the elements into a register.
5597 (rs6000_split_v4si_init_di_reg): Delete.
5598 (rs6000_split_v4si_init): Delete.
5599 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
5600 (vsx_init_v4si): Rewrite as a define_expand.
5602 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
5604 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
5605 zero_extend argument from memory): New.
5607 2018-07-22 Martin Sebor <msebor@redhat.com>
5610 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
5611 diagnosing calls with unknown arguments unless -Walloca-larger-than
5612 is restricted to less than PTRDIFF_MAX bytes.
5614 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
5616 * doc/gcov.texi (Invoking Gcov): Editorial changes.
5618 2018-07-20 David Malcolm <dmalcolm@redhat.com>
5620 * pretty-print.c (text_info::set_location): Remove redundant
5621 "line_table" parameter from call to rich_location::set_range.
5623 2018-07-20 Martin Sebor <msebor@redhat.com>
5626 * builtins.c (expand_builtin_alloca): Adjust.
5627 * calls.c (alloc_max_size): Simplify.
5628 * cgraphunit.c (cgraph_node::expand): Adjust.
5629 * common.opt (larger_than_size, warn_frame_larger_than): Remove
5631 (frame_larger_than_size): Same.
5632 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
5633 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
5635 * doc/invoke.texi (GCC Command Options): Document option arguments.
5636 Explain byte-size arguments and suffixes.
5637 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
5638 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
5639 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
5640 * doc/options.texi (UInteger): Expand.
5641 (Host_Wide_Int, ByteSize): Document new properties.
5642 * final.c (final_start_function_1): Include sizes in an error message.
5643 * function.c (frame_offset_overflow): Same.
5644 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
5645 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
5646 Diagnose unbounded alloca calls only for limits of less than
5648 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
5649 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
5651 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
5652 only for limits of less than PTRDIFF_MAX.
5653 * langhooks-def.h (lhd_handle_option): Change function argument
5655 * langhooks.c (lhd_handle_option): Same.
5656 * langhooks.h (handle_option): Same.
5657 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
5659 (var_type, var_type_struct): Same.
5660 (var_set): Handle ByteSize flag.
5661 * optc-gen.awk: Add comments to output to ease debugging. Make
5662 use of HOST_WIDE_INT where appropriate.
5663 * opts-gen-save.awk: Use %lx to format unsigned long.
5664 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
5665 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
5666 arguments. Parse bytes-size suffixes.
5667 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
5668 (enum_value_to_arg): Same.
5669 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
5670 (handle_option): Adjust.
5671 (generate_option): Change function argument to HOST_WIDE_INT.
5672 (cmdline_handle_error): Adjust.
5673 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
5674 (set_option): Change function argument to HOST_WIDE_INT.
5675 (option_enabled): Handle cl_host_wide_int.
5676 (get_option_state): Handle CLVC_SIZE.
5677 (control_warning_option): Same.
5678 * opts.c (common_handle_option): Change function argument to
5679 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
5680 OPT_Wvla_larger_than_.
5681 * opts.h (enum cl_var_type): Add an enumerator.
5682 * stor-layout.c (layout_decl): Print a more meaningful warning.
5683 * toplev.c (output_stack_usage): Adjust.
5685 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
5687 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
5688 call to inline_expand_builtin_string_cmp.
5689 (expand_builtin_strcmp): Likewise.
5690 (expand_builtin_strncmp): Likewise.
5691 (inline_string_cmp): Delete the last parameter, change char_type_node
5692 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
5694 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
5695 the inlining expansion on target where the type of the call has same or
5696 narrower precision than unsigned char.
5698 2018-07-20 David Malcolm <dmalcolm@redhat.com>
5700 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
5701 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
5702 * common.opt (fsave-optimization-record): New option.
5703 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
5704 * doc/invoke.texi (-fsave-optimization-record): New option.
5705 * dumpfile.c: Include "optinfo-emit-json.h".
5706 (struct kv_pair): Move to coretypes.h.
5707 (optgroup_options): Make non-static.
5708 (dump_context::end_scope): Call
5709 optimization_records_maybe_pop_dump_scope.
5710 * dumpfile.h (optgroup_options): New decl.
5711 * json.cc: New file.
5713 * optinfo-emit-json.cc: New file.
5714 * optinfo-emit-json.h: New file.
5715 * optinfo.cc: Include "optinfo-emit-json.h".
5716 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
5717 (optinfo_enabled_p): Check optimization_records_enabled_p.
5718 (optinfo_wants_inlining_info_p): Likewise.
5719 * optinfo.h: Update comment.
5720 * profile-count.c (profile_quality_as_string): New function.
5721 * profile-count.h (profile_quality_as_string): New decl.
5722 (profile_count::quality): New accessor.
5723 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
5724 and optinfo_emit_json_cc_tests.
5725 * selftest.h (selftest::json_cc_tests): New decl.
5726 (selftest::optinfo_emit_json_cc_tests): New decl.
5727 * toplev.c: Include "optinfo-emit-json.h".
5728 (compile_file): Call optimization_records_finish.
5729 (do_compile): Call optimization_records_start.
5730 * tree-ssa-live.c: Include optinfo.h.
5731 (remove_unused_scope_block_p): Retain inlining information if
5732 optinfo_wants_inlining_info_p returns true.
5734 2018-07-20 Richard Biener <rguenther@suse.de>
5737 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
5738 to cover -flto-partition=none.
5740 2018-07-20 Martin Liska <mliska@suse.cz>
5742 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
5743 (get_decl_source_range): Remove unused function.
5745 2018-07-20 Richard Biener <rguenther@suse.de>
5747 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
5748 (struct vn_phi_s): Likewise.
5749 (struct vn_reference_s): Likewise.
5750 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
5751 for searching the slot of an entry known to be in the hash itself.
5752 (vn_phi_hasher::equal): Likewise.
5753 (vn_reference_hasher::equal): Likewise.
5754 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
5756 (optimistic_info, current_info): Remove, keeping only valid_info.
5757 (vn_reference_lookup_1): Remove fallback lookup.
5758 (vn_reference_lookup_2): Likewise.
5759 (vn_nary_op_lookup_1): Likewise.
5760 (vn_phi_lookup): Likewise.
5761 (vn_nary_build_or_lookup_1): Make sure to not chain the built
5763 (vn_reference_insert): Adjust, chain the inserted hash element
5764 at last_inserted_ref.
5765 (vn_reference_insert_pieces): Likewise.
5766 (visit_reference_op_call): Likewise.
5767 (vn_nary_op_insert_into): Chain the inserted hash element at
5769 (vn_nary_op_insert_pieces): Adjust.
5770 (vn_nary_op_insert): Likewise.
5771 (vn_nary_op_insert_stmt): Likewise.
5772 (vn_phi_insert): Adjust, chain the inserted hash element at
5774 (process_scc): Remove clearing and copying the optimistic
5775 table. Instead remove elements inserted during an optimistic
5776 iteration from the single table we maintain.
5777 (init_scc_vn): Adjust.
5778 (free_scc_vn): Likewise.
5779 (sccvn_dom_walker::record_cond): Likewise.
5780 (sccvn_dom_walker::after_dom_children): Likewise.
5782 2018-07-19 Martin Sebor <msebor@redhat.com>
5784 PR tree-optimization/84047
5785 PR tree-optimization/83776
5786 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
5787 (check_array_bounds): Call it.
5789 2018-07-19 Martin Sebor <msebor@redhat.com>
5791 * align.h (align_flags): Use member initialization.
5793 2018-07-19 David Malcolm <dmalcolm@redhat.com>
5795 * Makefile.in (OBJS): Add optinfo.o.
5796 * coretypes.h (class symtab_node): New forward decl.
5797 (struct cgraph_node): New forward decl.
5798 (class varpool_node): New forward decl.
5799 * dump-context.h: New file.
5800 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
5802 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
5803 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
5804 (set_alt_dump_file): Likewise.
5805 (dump_context::~dump_context): New dtor.
5806 (dump_gimple_stmt): Move implementation to...
5807 (dump_context::dump_gimple_stmt): ...this new member function.
5808 Add the stmt to any pending optinfo, creating one if need be.
5809 (dump_gimple_stmt_loc): Move implementation to...
5810 (dump_context::dump_gimple_stmt_loc): ...this new member function.
5811 Start a new optinfo and add the stmt to it.
5812 (dump_gimple_expr): Move implementation to...
5813 (dump_context::dump_gimple_expr): ...this new member function.
5814 Add the stmt to any pending optinfo, creating one if need be.
5815 (dump_gimple_expr_loc): Move implementation to...
5816 (dump_context::dump_gimple_expr_loc): ...this new member function.
5817 Start a new optinfo and add the stmt to it.
5818 (dump_generic_expr): Move implementation to...
5819 (dump_context::dump_generic_expr): ...this new member function.
5820 Add the tree to any pending optinfo, creating one if need be.
5821 (dump_generic_expr_loc): Move implementation to...
5822 (dump_context::dump_generic_expr_loc): ...this new member
5823 function. Add the tree to any pending optinfo, creating one if
5825 (dump_printf): Move implementation to...
5826 (dump_context::dump_printf_va): ...this new member function. Add
5827 the text to any pending optinfo, creating one if need be.
5828 (dump_printf_loc): Move implementation to...
5829 (dump_context::dump_printf_loc_va): ...this new member function.
5830 Start a new optinfo and add the stmt to it.
5831 (dump_dec): Move implementation to...
5832 (dump_context::dump_dec): ...this new member function. Add the
5833 value to any pending optinfo, creating one if need be.
5834 (dump_context::dump_symtab_node): New member function.
5835 (dump_context::get_scope_depth): New member function.
5836 (dump_context::begin_scope): New member function.
5837 (dump_context::end_scope): New member function.
5838 (dump_context::ensure_pending_optinfo): New member function.
5839 (dump_context::begin_next_optinfo): New member function.
5840 (dump_context::end_any_optinfo): New member function.
5841 (dump_context::s_current): New global.
5842 (dump_context::s_default): New global.
5843 (dump_scope_depth): Delete global.
5844 (dumpfile_ensure_any_optinfo_are_flushed): New function.
5845 (dump_symtab_node): New function.
5846 (get_dump_scope_depth): Reimplement in terms of dump_context.
5847 (dump_begin_scope): Likewise.
5848 (dump_end_scope): Likewise.
5849 (selftest::temp_dump_context::temp_dump_context): New ctor.
5850 (selftest::temp_dump_context::~temp_dump_context): New dtor.
5851 (selftest::verify_item): New function.
5852 (ASSERT_IS_TEXT): New macro.
5853 (ASSERT_IS_TREE): New macro.
5854 (ASSERT_IS_GIMPLE): New macro.
5855 (selftest::test_capture_of_dump_calls): New test.
5856 (selftest::dumpfile_c_tests): Call it.
5857 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
5858 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
5859 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
5860 descriptive comment.
5861 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
5862 (dump_node, dump_bb): Move these unrelated decls.
5863 (class dump_manager): Add leading comment.
5864 * optinfo.cc: New file.
5865 * optinfo.h: New file.
5867 2018-07-19 Michael Collison <michael.collison@arm.com>
5868 Richard Henderson <rth@redhat.com>
5870 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
5871 (subti): Handle op1 zero.
5872 (subvti4, usub4ti4): New.
5873 (*sub<GPI>3_compare1_imm): New.
5874 (sub<GPI>3_carryinCV): New.
5875 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
5876 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
5878 2018-07-19 Michael Collison <michael.collison@arm.com>
5879 Richard Henderson <rth@redhat.com>
5881 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
5882 (addti3): Create simpler code if low part is already known to be 0.
5883 (addvti4, uaddvti4): New.
5884 (*add<GPI>3_compareC_cconly_imm): New.
5885 (*add<GPI>3_compareC_cconly): New.
5886 (*add<GPI>3_compareC_imm): New.
5887 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
5888 handle constants within this pattern..
5889 (*add<GPI>3_compareV_cconly_imm): New.
5890 (*add<GPI>3_compareV_cconly): New.
5891 (*add<GPI>3_compareV_imm): New.
5892 (add<GPI>3_compareV): New.
5893 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
5894 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
5895 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
5896 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
5897 with 'comparison' operator.
5898 (*add<GPI>3_compareV_cconly_imm): Ditto.
5899 (*add<GPI>3_compareV_cconly): Ditto.
5900 (*add<GPI>3_compareV_imm): Ditto.
5901 (add<GPI>3_compareV): Ditto.
5902 (add<mode>3_carryinC): Ditto.
5903 (*add<mode>3_carryinC_zero): Ditto.
5904 (*add<mode>3_carryinC): Ditto.
5905 (add<mode>3_carryinV): Ditto.
5906 (*add<mode>3_carryinV_zero): Ditto.
5907 (*add<mode>3_carryinV): Ditto.
5909 2018-07-19 Michael Collison <michael.collison@arm.com>
5910 Richard Henderson <rth@redhat.com>
5912 * config/aarch64/aarch64-modes.def (CC_V): New.
5913 * config/aarch64/aarch64-protos.h
5914 (aarch64_addti_scratch_regs): Declare
5915 (aarch64_subvti_scratch_regs): Declare.
5916 (aarch64_expand_subvti): Declare.
5917 (aarch64_gen_unlikely_cbranch): Declare
5918 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
5919 for signed overflow using CC_Vmode.
5920 (aarch64_get_condition_code_1): Handle CC_Vmode.
5921 (aarch64_gen_unlikely_cbranch): New function.
5922 (aarch64_addti_scratch_regs): New function.
5923 (aarch64_subvti_scratch_regs): New function.
5924 (aarch64_expand_subvti): New function.
5926 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
5928 * config/aarch64/aarch64-option-extensions.def: New entry for profile
5930 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
5931 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
5934 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
5937 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5938 address check not strict.
5940 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
5942 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
5944 (store_pair_lanes<mode>): Likewise.
5945 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
5946 enum value 'ADDR_QUERY_LDP_STP_N'.
5947 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
5948 (aarch64_print_address_internal): Add declaration.
5949 (aarch64_print_ldpstp_address): Remove.
5950 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
5951 (aarch64_print_operand): Change printing of 'y'.
5952 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
5953 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
5954 'true' rather than '1'.
5955 * config/aarch64/constraints.md (Uml): Likewise.
5956 (Uml): Rename to Umn.
5959 2018-07-19 Richard Biener <rguenther@suse.de>
5961 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
5963 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
5964 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
5965 (vn_reference_hasher): Likewise.
5966 (struct vn_tables_s): Remove obstack and alloc-pool members.
5967 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
5968 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
5969 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
5970 (vn_reference_insert_pieces): Likewise.
5971 (alloc_vn_nary_op_noinit): Adjust.
5972 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
5973 (vn_phi_eq): Adjust.
5974 (shared_lookup_phiargs): Remove.
5975 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
5976 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
5977 (visit_reference_op_call): Likewise.
5978 (copy_nary, copy_phi, copy_reference): Remove.
5979 (process_scc): Rewind the obstack when iterating. Do not
5980 copy the elements to valid_info but just move them from one
5981 hashtable to the other.
5982 (allocate_vn_table): Adjust.
5983 (free_vn_table): Likewise.
5984 (init_scc_vn): Likewise.
5985 (free_scc_vn): Likewise.
5987 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
5990 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
5991 indirect_return as function type attribute.
5992 (ix86_attribute_table): Change indirect_return to function
5994 * doc/extend.texi: Update indirect_return attribute.
5996 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
5998 * wide-int.h (widest2_int): New.
5999 * gimple-fold.c (arith_overflowed_p): Use it.
6000 * tree.h (widest2_int_cst): New.
6001 * tree-vrp.c (wide_int_binop_overflow): Rename from
6002 vrp_int_const_binop.
6003 Rewrite to work on trees.
6004 (extract_range_from_multiplicative_op_1): Abstract code to...
6005 (wide_int_range_min_max): ...here.
6006 (wide_int_range_cross_product): ...and here.
6007 (extract_range_from_binary_expr_1): Abstract overflow code to...
6008 (wide_int_range_mult_wrapping): ...here.
6009 * tree-vrp.h (wide_int_range_cross_product): New.
6010 (wide_int_range_mult_wrapping): New.
6012 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
6013 Julia Koval <julia.koval@intel.com>
6015 * config/i386/x86-tune-costs.h (skylake_memcpy,
6016 skylake_memset): Replace rep_prefix with unrolling for size 512.
6018 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
6021 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
6022 comparision with EQ_EXPR in last stmt.
6024 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
6026 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
6027 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
6028 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
6029 previously known as "PowerPC AltiVec Built-in Functions". Move
6030 some material to new subsubsections "PowerPC AltiVec Built-in
6031 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
6033 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
6034 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
6035 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
6036 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
6038 2018-07-18 Richard Biener <rguenther@suse.de>
6040 PR tree-optimization/86557
6041 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
6044 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
6046 * config/s390/s390.c (s390_function_profiler): Generate CFI.
6048 2018-07-17 Jeff Law <law@redhat.com>
6050 * config/arm/arm.c (get_label_padding): Update for recent
6051 changes to label_to_alignment.
6053 PR tree-optimization/86010
6054 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
6056 * config/mips/mips.c (vr4130_align_insns): Update for recent
6057 changes to label_to_alignment.
6059 * config/frv/frv.c (frv_label_align): Update for recent changes
6060 to label_to_alignment.
6062 * config/nios2/nios2.c (nios2_label_align): Update for recent
6063 changes which dropped ALIGN_LABELS_LOG.
6065 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
6067 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
6068 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
6070 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
6072 * config/arc/arc.c (arc_label_align): Use align_labels instead of
6073 deprecated align_labels_log.
6075 2018-07-17 Richard Biener <rguenther@suse.de>
6078 * dwarf2out.c (init_sections_and_labels): Always generate
6079 a debug_line_str_section for early LTO debug.
6080 (dwarf2out_finish): Reset debug_line_str_hash output early.
6081 Bump counter for extra dwarf5 .debug_loc labels to not conflict
6083 (dwarf2out_early_finish): Output debug_line_str.
6085 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
6087 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
6088 index register on z196 or later.
6090 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
6092 * config/s390/s390.c (s390_default_align): Set default function
6094 (s390_override_options_after_change): Call s390_default align.
6095 (s390_option_override_internal): Call s390_default_align.
6096 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
6098 2018-07-17 Jakub Jelinek <jakub@redhat.com>
6101 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
6102 to _looptemp_ clauses, other than the first two.
6104 2018-07-17 Martin Liska <mliska@suse.cz>
6106 * opts.c: Do not enable OPT_falign_* for -Os.
6108 2018-07-17 Martin Liska <mliska@suse.cz>
6110 * align.h (MAX_CODE_ALIGN): New.
6111 (MAX_CODE_ALIGN_VALUE): New.
6112 * common/config/i386/i386-common.c (ix86_handle_option):
6113 (MAX_CODE_ALIGN): Moved to align.h.
6114 * final.c (MAX_CODE_ALIGN): Likewise.
6115 * opts.c (parse_and_check_align_values):
6116 (MAX_CODE_ALIGN): Likewise.
6117 (MAX_CODE_ALIGN_VALUE): Likewise.
6119 2018-07-17 Martin Liska <mliska@suse.cz>
6121 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
6122 in order to fulfil coding style.
6123 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
6124 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6125 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
6126 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
6127 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
6128 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
6129 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
6130 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
6131 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6133 2018-07-17 Martin Liska <mliska@suse.cz>
6135 * align.h: New file.
6136 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
6138 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
6139 align_flags of label_to_alignment.
6140 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
6142 * config/m68k/m68k.c: Do not use removed align_labels_value and
6144 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
6145 (LOOP_ALIGN): Likewise.
6146 (LABEL_ALIGN): Likewise.
6147 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
6148 Remove not used macro.
6149 (rs6000_loop_align): Change return type to align_flags.
6150 (rs6000_loop_align_max_skip): Remove.
6151 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
6152 Change return type to align_flags.
6153 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
6154 Remove not used macro.
6155 (rs6000_loop_align): Change return type to align_flags.
6156 (rs6000_loop_align_max_skip): Remove.
6157 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
6158 * config/rx/rx-protos.h (rx_align_for_label): Make it
6160 * config/rx/rx.c (rx_align_for_label): Change return type
6162 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
6164 into align_flags class.
6165 (LABEL_ALIGN): Likewise.
6166 (LOOP_ALIGN): Likewise.
6167 * config/s390/s390.c (s390_label_align): Use align_flags
6169 (s390_asm_output_function_label): Likewise.
6170 * config/sh/sh.c (sh_override_options_after_change):
6171 Use align_flags class directly without macros.
6172 (find_barrier): Likewise.
6173 (barrier_align): Likewise.
6174 (sh_loop_align): Likewise.
6175 * config/spu/spu.c (spu_option_override):
6176 Use align_flags_tuple::get_value instead of removed macros.
6177 (spu_sched_init): Likewise.
6178 * config/spu/spu.h (GTY): Likewise.
6179 * config/visium/visium.c (visium_option_override):
6180 Set "8" as default secondary alignment.
6181 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
6182 in order to guarantee secondary alignment of 8.
6183 * coretypes.h: Include align.h header file.
6184 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
6185 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
6186 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
6187 * doc/tm.texi.in: Likewise.
6188 * final.c (struct label_alignment): Remove not used structure.
6189 (LABEL_ALIGN): Change type to align_flags.
6190 (LOOP_ALIGN): Likewise.
6191 (JUMP_ALIGN): Likewise.
6192 (default_loop_align_max_skip): Remove.
6193 (default_label_align_max_skip): Likewise.
6194 (default_jump_align_max_skip): Likewise.
6195 (default_label_align_after_barrier_max_skip):
6196 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
6197 (LABEL_TO_MAX_SKIP): Remove.
6198 (label_to_alignment): Return align_flags type instead of integer.
6199 (label_to_max_skip): Remove.
6200 (align_fuzz): Use align_flags type.
6201 (compute_alignments): Use align_flags type and use align_flags::max
6202 to combine multiple alignments.
6203 (grow_label_align): Grow vec instead of C array.
6204 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
6205 (shorten_branches): Use align_flags type and use align_flags::max
6206 to combine multiple alignments.
6207 (final_scan_insn_1): Remove usage of secondary alignment that comes
6208 from label alignment, but instead use proper secondary alignment
6209 which is computed in grow_label_align.
6210 * flags.h (struct align_flags_tuple): Move to align.h.
6211 (struct align_flags): Likewise.
6212 (state_align_loops): Rename to align_loops.
6213 (state_align_jumps): Rename to align_jumps.
6214 (state_align_labels): Rename to align_labels.
6215 (state_align_functions): Rename to align_functions.
6216 (align_loops_log): Remove.
6217 (align_jumps_log): Remove.
6218 (align_labels_log): Remove.
6219 (align_functions_log): Remove.
6220 (align_loops_max_skip): Remove.
6221 (align_jumps_max_skip): Remove.
6222 (align_labels_max_skip): Remove.
6223 (align_functions_max_skip): Remove.
6224 (align_loops_value): Remove.
6225 (align_jumps_value): Remove.
6226 (align_labels_value): Remove.
6227 (align_functions_value): Remove.
6228 * output.h (label_to_alignment): Change return type to align_flags.
6229 (label_to_max_skip): Remove.
6230 * target.def: Remove loop_align_max_skip, label_align_max_skip,
6231 jump_align_max_skip macros.
6232 * targhooks.h (default_loop_align_max_skip): Remove.
6233 (default_label_align_max_skip): Likewise.
6234 (default_jump_align_max_skip): Likewise.
6235 (default_label_align_after_barrier_max_skip): Remove.
6236 * toplev.c (read_log_maxskip): Use ::normalize function.
6237 (parse_N_M): Remove not used argument and also call ::normalize.
6238 (parse_alignment_opts): Do not pass unused arguments.
6239 * varasm.c (assemble_start_function): Use directly align_functions
6240 instead of removed macros.
6241 * system.h: Do not poison removed macros.
6243 2018-07-17 Jakub Jelinek <jakub@redhat.com>
6246 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
6247 and cond temporaries don't have reference type if iterator has
6248 pointer type. For init use &for_pre_body instead of pre_p if
6249 for_pre_body is non-empty.
6251 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
6253 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
6254 double-double modes to SFmode directly directly.
6255 (trunc<mode>sf2_fprs): Delete.
6257 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
6259 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
6260 for conversions between IFmode and the decimal floating point modes.
6261 (init_float128_ieee): Use the correct names for conversions between
6262 KFmode and the decimal floating point modes.
6264 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
6266 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
6267 for the conversions between TDmode and IFmode.
6268 (init_float128_ieee): Use more correct names for the conversions
6269 between TDmode and KFmode.
6271 2018-07-16 Jakub Jelinek <jakub@redhat.com>
6273 PR tree-optimization/86526
6274 * builtins.c (expand_builtin_memcmp): Formatting fixes.
6275 (inline_expand_builtin_string_cmp): Likewise.
6276 (inline_string_cmp): Likewise. Use c_readstr instead of
6277 builtin_memcpy_read_str. Add unit_mode temporary.
6279 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
6282 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
6283 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
6285 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6287 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
6288 Alphabetize prototypes of built-in functions, separating out
6289 built-in functions that are listed in this section but should be
6290 described elsewhere.
6292 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
6295 * expmed.c (emit_store_flag): Do not emit setcc followed by a
6296 conditional move when trapping comparison was split to a
6297 non-trapping one (and vice versa).
6299 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
6301 * config/s390/s390.c (s390_function_profiler): Generate nops
6302 instead of profiler call sequences.
6303 * config/s390/s390.opt: Add the new option.
6305 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
6307 * config/s390/s390.c (s390_function_profiler): Generate
6308 __mcount_loc section.
6309 * config/s390/s390.opt: Add the new option.
6311 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
6313 * common.opt: Add the new warning.
6314 * config/s390/s390.c (s390_function_profiler): Emit "brasl
6315 %r0,__fentry__" when -mfentry is specified.
6316 (s390_option_override_internal): Disallow -mfentry for 31-bit
6318 * config/s390/s390.opt: Add the new option.
6320 2018-07-16 Richard Biener <rguenther@suse.de>
6323 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
6324 for function-local FUNCTION_DECL and RESULT_DECL immediately.
6326 2018-07-16 Martin Liska <mliska@suse.cz>
6329 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
6332 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
6334 * config/arc/arcHS.md: Update ARCHS scheduling rules.
6336 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
6338 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
6340 * config/arc/arc-cpus.def (hs4x): New cpu.
6342 * config/arc/arc-tables.opt: Regenerate.
6343 * config/arc/arc.c (arc_sched_issue_rate): New function.
6344 (TARGET_SCHED_ISSUE_RATE): Define.
6345 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
6346 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
6348 (attr tune): Add ARCHS4x tune values.
6349 (attr tune_dspmpy): Define.
6350 (*tst): Correct instruction type.
6351 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
6352 * config/arc/arcHS4x.md: New file.
6353 * config/arc/fpu.md: Update instruction type attributes.
6354 * config/arc/t-multilib: Regenerate.
6356 2018-07-16 Tom de Vries <tdevries@suse.de>
6359 * var-tracking.c (vt_initialize): Fix pre_dec handling.
6361 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6363 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
6366 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
6368 PR tree-optimization/86514
6369 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
6370 conversion to a boolean type from a type with greater precision.
6372 2018-07-16 Tom de Vries <tdevries@suse.de>
6374 * var-tracking.c (vt_initialize): Print adjusted insn slim if
6375 dump_flags request TDF_SLIM.
6377 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
6379 * fold-const.c (int_const_binop_1): Abstract...
6380 (wide_int_binop): ...wide int code here.
6381 (poly_int_binop): ...poly int code here.
6382 Abstract the rest of int_const_binop_1 into int_const_binop.
6383 * fold-const.h (wide_int_binop): New.
6384 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
6385 Remove useless PLUS/MINUS_EXPR case.
6386 (zero_nonzero_bits_from_vr): Move wide int code...
6387 (zero_nonzero_bits_from_bounds): ...here.
6388 (extract_range_from_binary_expr_1): Move mask optimization code...
6389 (range_easy_mask_min_max): ...here.
6390 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
6391 (range_easy_mask_min_max): New.
6393 2018-07-15 Jeff Law <law@redhat.com>
6396 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
6399 2018-07-14 Jim Wilson <jimw@sifive.com>
6401 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
6403 2018-07-14 Paul Koning <ni1d@arrl.net>
6405 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
6407 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
6409 * lto-streamer-out.c (copy_function_or_variable): Dump info about
6412 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
6413 Steve Munroe <munroesj52@gmail.com>
6415 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
6416 (_mm_andnot_si128): Likewise.
6417 (_mm_or_si128): Likewise.
6418 (_mm_xor_si128): Likewise.
6420 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
6423 * builtins.c (expand_builtin_memcmp): Inline the calls first
6424 when result_eq is false.
6425 (expand_builtin_strcmp): Inline the calls first.
6426 (expand_builtin_strncmp): Likewise.
6427 (inline_string_cmp): New routine. Expand a string compare
6428 call by using a sequence of char comparison.
6429 (inline_expand_builtin_string_cmp): New routine. Inline expansion
6430 a call to str(n)cmp/memcmp.
6431 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
6433 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
6435 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
6437 * config/arm/driver-arm.c: Include arm-native.h.
6438 (host_detect_local_cpu): Use auto-generated data tables.
6439 (vendors, arm_cpu_table): Delete. Move part information to ...
6440 * config/arm/arm-cpus.in: ... here.
6441 * config/arm/parsecpu.awk (gen_native): New function.
6442 (vendor, part): New CPU fields.
6443 (END): Add support for building the native CPU detection tables.
6444 * config/arm/t-arm (arm-native.h): Add build rule.
6445 (driver-arm.o): Add dependency on arm-native.h.
6447 2018-07-13 Richard Biener <rguenther@suse.de>
6450 * match.pd (addr1 - addr2): Allow either of the operand to
6453 2018-07-13 Tom de Vries <tdevries@suse.de>
6455 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
6458 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
6460 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
6461 arrays instead of numbered variables.
6463 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
6465 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
6466 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
6468 2018-07-13 Richard Biener <rguenther@suse.de>
6471 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
6472 instead of get_context_die.
6474 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
6475 Richard Biener <rguenther@suse.de>
6478 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
6479 that the loop latch destination where phi is defined.
6481 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
6483 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
6484 (riscv_expand_epilogue): Add assertion to check interrupt mode.
6485 (riscv_set_current_function): Extract getting interrupt type to new
6487 (riscv_get_interrupt_type): New function.
6488 (riscv_merge_decl_attributes): New function, checking interrupt type is
6490 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
6492 2018-07-12 Paul Koning <ni1d@arrl.net>
6494 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
6497 2018-07-12 Paul Koning <ni1d@arrl.net>
6499 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
6500 zero reference, add doloop_end instead.
6501 * doc/md.texi (decrement_and_branch_until_zero): Remove.
6502 (Looping patterns): Remove decrement_and_branch_until_zero. Add
6503 detail for doloop_end.
6505 2018-07-12 Martin Sebor <msebor@redhat.com>
6508 * attribs.c (decl_attributes): Reject conflicting attributes before
6509 calling attribute handlers.
6511 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
6513 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
6515 (gcc::dump_manager::get_dump_file_name): likewise.
6516 (dump_begin): Likewise.
6517 * dumpfile.h (dump_begin): Update prototype.
6518 (gcc::dump_manager::get_dump_file_name,
6519 gcc::dump_manager::get_dump_file_name): Update prototype.
6521 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6523 * internal-fn.h (vectorizable_internal_fn_p): New function.
6524 * tree-vect-slp.c (compatible_calls_p): Likewise.
6525 (vect_build_slp_tree_1): Remove nops argument. Handle calls
6526 to internal functions.
6527 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
6529 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6531 * fold-const.h (inverse_conditions_p): Declare.
6532 * fold-const.c (inverse_conditions_p): New function.
6533 * match.pd: Use inverse_conditions_p. Add folds of view_converts
6534 that test the inverse condition of a conditional internal function.
6535 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
6536 * internal-fn.c (internal_fn_mask_index): Handle conditional
6538 (vectorized_internal_fn_supported_p): New function.
6539 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
6540 (any_pred_load_store): Replace with...
6541 (need_to_predicate): ...this new variable.
6542 (redundant_ssa_names): New variable.
6543 (ifcvt_can_use_mask_load_store): Move initial checks to...
6544 (ifcvt_can_predicate): ...this new function. Handle tree codes
6545 for which a conditional internal function exists.
6546 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
6547 instead of ifcvt_can_use_mask_load_store. Update after variable
6549 (predicate_load_or_store): New function, split out from
6550 predicate_mem_writes.
6551 (check_redundant_cond_expr): New function.
6552 (value_available_p): Likewise.
6553 (predicate_rhs_code): Likewise.
6554 (predicate_mem_writes): Rename to...
6555 (predicate_statements): ...this. Use predicate_load_or_store
6556 and predicate_rhs_code.
6557 (combine_blocks, tree_if_conversion): Update after above name changes.
6558 (ifcvt_local_dce): Handle redundant_ssa_names.
6559 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
6560 general conditional functions.
6561 * tree-vect-stmts.c (vectorizable_call): Likewise.
6563 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6564 Alan Hayward <alan.hayward@arm.com>
6565 David Sherwood <david.sherwood@arm.com>
6567 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
6568 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
6569 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
6570 plus and minus and convert them into IFN_COND_FMA-based sequences.
6571 (convert_mult_to_fma): Handle conditional plus and minus.
6573 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6575 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
6576 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
6577 (cond_fnms_optab): New optabs.
6578 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
6580 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
6581 * internal-fn.h (get_conditional_internal_fn): Declare.
6582 (get_unconditional_internal_fn): Likewise.
6583 * internal-fn.c (cond_ternary_direct): New macro.
6584 (expand_cond_ternary_optab_fn): Likewise.
6585 (direct_cond_ternary_optab_supported_p): Likewise.
6586 (FOR_EACH_COND_FN_PAIR): Likewise.
6587 (get_conditional_internal_fn): New function.
6588 (get_unconditional_internal_fn): Likewise.
6589 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
6590 (gimple_match_op::gimple_match_op): Add a new overload for 5
6592 (gimple_match_op::set_op): Likewise.
6593 (gimple_resimplify5): Declare.
6594 * genmatch.c (decision_tree::gen): Generate simplifications for
6596 * gimple-match-head.c (gimple_simplify): Define an overload for
6597 5 operands. Handle calls with 5 arguments in the top-level overload.
6598 (convert_conditional_op): Handle conversions from unconditional
6599 internal functions to conditional ones.
6600 (gimple_resimplify5): New function.
6601 (build_call_internal): Pass a fifth operand.
6602 (maybe_push_res_to_seq): Likewise.
6603 (try_conditional_simplification): Try converting conditional
6604 internal functions to unconditional internal functions.
6605 Handle 3-operand unconditional forms.
6606 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
6607 Define ternary equivalents of the current rules for binary conditional
6609 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
6611 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
6612 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
6613 (optab): Handle them.
6614 (SVE_COND_FP_TERNARY): New int iterator.
6615 (sve_fmla_op, sve_fmad_op): New int attributes.
6616 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
6617 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
6618 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
6620 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6622 * target.def (preferred_else_value): New target hook.
6623 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
6624 * doc/tm.texi: Regenerate.
6625 * targhooks.h (default_preferred_else_value): Declare.
6626 * targhooks.c (default_preferred_else_value): New function.
6627 * internal-fn.h (conditional_internal_fn_code): Declare.
6628 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
6629 (get_conditional_internal_fn): Use it.
6630 (conditional_internal_fn_code): New function.
6631 * gimple-match.h (gimple_match_cond): New struct.
6632 (gimple_match_op): Add a cond member function.
6633 (gimple_match_op::gimple_match_op): Update all forms to take a
6635 * genmatch.c (expr::gen_transform): Use the same condition as res_op
6636 for the suboperation, but don't specify a particular else_value.
6637 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
6638 (visit_nary_op, visit_reference_op_load): Pass
6639 gimple_match_cond::UNCOND to the gimple_match_op constructor.
6640 * gimple-match-head.c: Include tree-eh.h
6641 (convert_conditional_op): New function.
6642 (maybe_resimplify_conditional_op): Likewise.
6643 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
6644 (gimple_resimplify2): Likewise.
6645 (gimple_resimplify3): Likewise.
6646 (gimple_resimplify4): Likewise.
6647 (maybe_push_res_to_seq): Return null for conditional operations.
6648 (try_conditional_simplification): New function.
6649 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
6651 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
6653 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
6655 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
6657 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
6659 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
6661 (hash_tree): Do not hash DECL_FCONTEXT
6662 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
6663 Do not stream DECL_FCONTEXT.
6664 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
6665 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
6667 2018-07-12 Richard Biener <rguenther@suse.de>
6670 * dwarf2out.c (gen_block_die): Only output blocks when they have
6671 at least one !DECL_IGNORED_P variable.
6673 2018-07-12 Richard Biener <rguenther@suse.de>
6676 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
6677 Remove -mieee-fp handling.
6679 2018-07-12 Richard Biener <rguenther@suse.de>
6681 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
6682 left-over from last patch.
6684 2018-07-12 Jakub Jelinek <jakub@redhat.com>
6686 PR tree-optimization/86492
6687 * gimple-ssa-store-merging.c
6688 (imm_store_chain_info::coalesce_immediate_stores): Call
6689 check_no_overlap even for the merge_overlapping case. Formatting fix.
6691 2018-07-12 Richard Biener <rguenther@suse.de>
6694 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
6695 move possibly trapping operations into the conditional.
6697 2018-07-12 Richard Biener <rguenther@suse.de>
6699 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
6700 (vn_lookup_simplify_result): Remove recursion limit applied
6702 (vn_nary_build_or_lookup_1): Adjust.
6703 (try_to_simplify): Likewise.
6704 * gimple-match-head.c (gimple_resimplify1): Instead apply one
6706 (gimple_resimplify2): Likewise.
6707 (gimple_resimplify3): Likewise.
6708 (gimple_resimplify4): Likewise.
6710 2018-07-11 Jakub Jelinek <jakub@redhat.com>
6712 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
6713 Use __mmask64 type instead of __mmask8 for __M argument.
6714 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
6715 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
6717 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
6718 __mmask16 for __M argument.
6719 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
6720 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
6721 to __mmask16 instead of __mmask8.
6722 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
6723 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
6724 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
6725 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
6726 instead of __mmask16 for __U argument.
6727 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
6728 __mmask16 instead of __mmask8 for __U argument.
6729 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
6731 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
6733 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
6735 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
6737 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
6739 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
6741 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
6743 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
6745 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
6747 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
6749 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
6750 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
6751 return type as well as __M argument type and all casts from __mmask8
6753 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
6754 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
6755 return type as well as __M argument type and all casts from __mmask8
6757 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
6758 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
6759 return type as well as __M argument type and all casts from __mmask8
6761 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
6762 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
6763 return type as well as __M argument type and all casts from __mmask8
6765 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
6766 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
6769 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
6771 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
6772 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
6775 2018-07-11 Paul Koning <ni1d@arrl.net>
6777 * doc/md.texi (define_subst): Document how multiple occurrences of
6778 the same argument in the replacement pattern are handled.
6780 2018-07-11 Paul Koning <ni1d@arrl.net>
6782 * doc/extend.texi (Common Variable Attributes): Move "mode" into
6784 (Common Type Attributes): Add "mode" attribute.
6786 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
6788 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
6789 stream DECL_ORIGINAL_TYPE.
6790 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
6791 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
6792 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
6793 Do not walk original type.
6794 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
6796 (write_ts_decl_non_common_tree_pointers): Do not stream
6798 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
6799 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
6801 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
6803 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
6804 thread twice from the same starting edge.
6806 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
6808 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
6809 * gimple.c (gimple_call_nonnull_result_p): ...here...
6810 (gimple_call_nonnull_arg): ...and here.
6811 * gimple.h (gimple_call_nonnull_result_p): New.
6812 (gimple_call_nonnull_arg): New.
6814 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
6816 * config/arm/arm-cpus.in: Move information from fpu field of each
6817 cpu definition to the isa field.
6818 * config/arm/parsecpu.awk (fpu): Delete match rule.
6819 (gen_comm_data): Don't add bits from the CPU's FPU entry.
6821 2018-07-11 Richard Biener <rguenther@suse.de>
6824 * dwarf2out.c (init_sections_and_labels): Use
6825 output_asm_line_debug_info consistently.
6826 (dwarf2out_early_finish): Likewise.
6827 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
6830 2018-07-11 Richard Biener <rguenther@suse.de>
6832 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
6833 Rework father_bb setting in a way to avoid propagating constants
6834 multiple times on a loop body.
6836 2018-07-10 Mark Wielaard <mark@klomp.org>
6839 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
6841 2018-07-10 Richard Biener <rguenther@suse.de>
6843 * hash-map.h (hash_map::iterator::operator*): Return
6844 references to key and value.
6846 2018-07-10 Jakub Jelinek <jakub@redhat.com>
6849 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
6850 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
6851 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
6852 (gimplify_omp_for): For composite loops, move outer
6853 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
6854 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
6855 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
6856 TREE_LIST for both the original class iterator and the "last" helper
6857 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
6858 loop, remember has_decl_expr from outer composite loops for the
6859 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
6861 2018-07-09 Martin Sebor <msebor@redhat.com>
6865 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
6866 accessing implicitly initialized array elements.
6867 * expr.c (string_constant): Handle string initializers of
6868 character arrays within aggregates.
6869 * gimple-fold.c (fold_array_ctor_reference): Add argument.
6870 Store element offset. As a special case, handle zero size.
6871 (fold_nonarray_ctor_reference): Same.
6872 (fold_ctor_reference): Add argument. Store subobject offset.
6873 * gimple-fold.h (fold_ctor_reference): Add argument.
6875 2018-07-09 Paul Koning <ni1d@arrl.net>
6877 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
6878 (pdp11_insn_cost): New function.
6879 (pdp11_md_asm_adjust): New function.
6880 (TARGET_INVALID_WITHIN_DOLOOP): Define.
6881 (pdp11_rtx_costs): Update to match machine better.
6882 (output_addr_const_pdp11): Correct format mismatch warnings.
6883 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
6884 * config/pdp11/pdp11.md: General change to add base_cost and/or
6885 length attributes for use by new pdp11_insn_cost function.
6886 (MIN_BRANCH): Correct definition.
6888 (doloop_end): Use standard pattern name for looping pattern.
6889 (doloop_end_nocc): New.
6890 (movsf): Add another constraint alternative.
6891 (zero_extendqihi2): Add constraint alternatives for not in place
6893 (zero_extendhisi2): Remove.
6894 (shift patterns): Add CC handling variants.
6898 (define_peephole2): New peephole to recognize mov that sets CC for
6901 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6903 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
6904 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
6907 2018-07-09 Jakub Jelinek <jakub@redhat.com>
6910 * real.c (real_nextafter): Return true if result is denormal.
6912 2018-07-09 Martin Liska <mliska@suse.cz>
6914 * common.opt: Add back wrongly removed attribute.
6916 2018-07-09 Richard Biener <rguenther@suse.de>
6919 * dwarf2out.c (gen_block_die): For an early generated DIE
6920 always output high/low PC attributes.
6922 2018-07-09 Tom de Vries <tdevries@suse.de>
6924 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
6925 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
6926 onto VAR_DECL with abstract origin.
6928 2018-07-07 Jim Wilson <jimw@sifive.com>
6930 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
6932 2018-07-07 Tom de Vries <tdevries@suse.de>
6934 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
6936 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
6938 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
6940 (combine_bound): Use wide-int overflow calculation instead of
6942 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
6944 * fold-const.c (int_const_binop_2): Same.
6945 (extract_muldiv_1): Same.
6946 (fold_div_compare): Same.
6947 (fold_abs_const): Same.
6949 * poly-int.h (add): Same.
6953 * predict.c (predict_iv_comparison): Same.
6954 * profile-count.c (slow_safe_scale_64bit): Same.
6955 * simplify-rtx.c (simplify_const_binary_operation): Same.
6956 * tree-chrec.c (tree_fold_binomial): Same.
6957 * tree-data-ref.c (split_constant_offset_1): Same.
6958 * tree-if-conv.c (idx_within_array_bound): Same.
6959 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
6960 * tree-ssa-phiopt.c (minmax_replacement): Same.
6961 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
6962 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
6963 * vr-values.c (vr_values::adjust_range_with_scev): Same.
6964 * wide-int.cc (wi::add_large): Same.
6965 (wi::mul_internal): Same.
6966 (wi::sub_large): Same.
6967 (wi::divmod_internal): Same.
6968 * wide-int.h: Change overflow type to overflow_type for neg, add,
6969 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
6970 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
6971 mul_internal, divmod_internal.
6972 (overflow_type): New enum.
6973 (accumulate_overflow): New.
6975 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
6977 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
6978 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
6980 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
6982 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
6983 argument is checked for zero before entering loop, avoid checking again.
6985 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
6987 * gimplify.h (generic_expr_could_trap_p): Set as global function.
6988 * gimplify.h (generic_expr_could_trap_p): Likwise.
6989 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
6991 2018-07-06 Jakub Jelinek <jakub@redhat.com>
6993 PR tree-optimization/86401
6994 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
6995 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
6996 (fold_bit_and_mask): ... here. New helper function for match.pd.
6997 * fold-const.h (fold_bit_and_mask): Declare.
6998 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
7000 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
7003 * target.def (translate_mode_attribute): New hook.
7004 * targhooks.h (default_translate_mode_attribute): Declare.
7005 * targhooks.c (default_translate_mode_attribute): New function.
7006 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
7007 * doc/tm.texi: Regenerate.
7008 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
7009 (rs6000_translate_mode_attribute): New function.
7011 2018-07-06 Paul Koning <ni1d@arrl.net>
7013 * doc/md.texi (define_split): Document DONE and FAIL.
7014 (define_peephole2): Ditto.
7016 2018-07-05 Jeff Law <law@redhat.com>
7018 PR tree-optimization/86010
7019 * tree-ssa-dse.c (compute_trims): More aggressively trim at
7020 both the head and tail of mem* and str* calls.
7022 2018-07-05 Jim Wilson <jimw@sifive.com>
7024 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
7026 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
7028 * config/aarch64/aarch64-simd.md: correct flags text for
7029 MIN_EXPR replacement.
7031 2018-07-05 James Clarke <jrtc27@jrtc27.com>
7033 * configure: Regenerated.
7035 2018-07-05 Carl Love <cel@us.ibm.com>
7037 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
7038 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
7039 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
7040 VSX_BUILTIN_DOUBLEL_V4SF.
7042 2018-07-05 Martin Sebor <msebor@redhat.com>
7045 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
7046 than its domain to compute its the upper bound of a char array.
7048 2018-07-05 Nathan Sidwell <nathan@acm.org>
7050 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
7051 * doc/cpp.texi: Update comment.
7052 * doc/tm.texi: Rebuilt.
7053 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
7054 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
7055 * doc/extend.texi (Backwards Compatibility): Clarify it is system
7056 headers affected by extern "C".
7057 * system.h: Poison NO_IMPLICIT_EXTERN_C.
7058 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
7059 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
7060 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
7061 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
7062 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
7063 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
7064 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
7065 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
7066 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
7067 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
7068 config/sparc/sp64-elf.h, config/spu/spu.h,
7069 config/stormy16/stormy16.h, config/v850/v850.h,
7070 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
7071 define NO_IMPLICIT_EXTERN_C.
7072 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
7074 2018-07-05 Tamar Christina <tamar.christina@arm.com>
7077 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
7078 instead of GET_MODE_SIZE when comparing Units.
7080 2018-07-05 Tamar Christina <tamar.christina@arm.com>
7083 * rtlanal.c (set_noop_p): Constrain on mode change,
7084 include hard-reg-set.h
7086 2018-07-05 Tamar Christina <tamar.christina@arm.com>
7088 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
7090 2018-07-05 Jakub Jelinek <jakub@redhat.com>
7093 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
7096 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
7098 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
7100 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
7103 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
7105 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
7107 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
7110 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
7112 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
7114 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
7115 neg_*_op* variables.
7117 2018-07-04 Martin Liska <mliska@suse.cz>
7119 * tree-switch-conversion.c: Define
7120 max_ratio_for_speed and max_ratio_for_size constants.
7122 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
7123 Martin Liska <mliska@suse.cz>
7128 * common.opt: Rename align options with 'str_' prefix.
7129 * common/config/i386/i386-common.c (set_malign_value): New
7131 (ix86_handle_option): Use it to set -falign-* options/
7132 * config/aarch64/aarch64-protos.h (struct tune_params): Change
7133 type from int to string.
7134 * config/aarch64/aarch64.c: Update default values from int
7136 * config/alpha/alpha.c (alpha_override_options_after_change):
7138 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
7139 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7140 max skip conditionally.
7141 * config/i386/freebsd.h (SUBALIGN_LOG): New.
7142 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7143 max skip conditionally.
7144 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7145 max skip conditionally.
7146 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
7147 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7148 max skip conditionally.
7149 * config/i386/i386.c (struct ptt): Change type from int to
7151 (ix86_default_align): Set default values.
7152 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
7153 max skip conditionally.
7154 * config/i386/iamcu.h (SUBALIGN_LOG): New.
7155 (ASM_OUTPUT_MAX_SKIP_ALIGN):
7156 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
7157 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7158 max skip conditionally.
7159 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
7160 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
7161 * config/i386/x86-64.h (SUBALIGN_LOG): New.
7162 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7163 max skip conditionally.
7164 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
7165 * config/ia64/ia64.c (ia64_option_override): Set default values
7166 for alignment options.
7167 * config/m68k/m68k.c: Handle new str_align_* options.
7168 * config/mips/mips.c (mips_set_compression_mode): Change
7170 (mips_option_override): Set default values for options.
7171 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
7173 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7175 * config/rx/rx.c (rx_option_override): Likewise.
7176 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
7177 (LABEL_ALIGN): Use align_labels_log.
7178 (LOOP_ALIGN): Use align_loops_align.
7179 * config/s390/s390.c (s390_asm_output_function_label): Use new
7181 * config/sh/sh.c (sh_override_options_after_change):
7182 Change type of constants.
7183 * config/spu/spu.c (spu_sched_init): Likewise.
7184 * config/sparc/sparc.c (sparc_option_override): Set default
7186 * config/visium/visium.c (visium_option_override): Likewise.
7187 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
7188 emit p2align format with last argument if it's not needed.
7189 * doc/invoke.texi: Document extended format of -falign-*.
7190 * final.c: Use align_labels alignment.
7191 * flags.h (struct target_flag_state): Change type to use
7193 (struct align_flags_tuple): New.
7194 (struct align_flags): Likewise.
7195 (align_loops_log): Redefine macro to use new types.
7196 (align_loops_max_skip): Redefine macro to use new types.
7197 (align_jumps_log): Redefine macro to use new types.
7198 (align_jumps_max_skip): Redefine macro to use new types.
7199 (align_labels_log): Redefine macro to use new types.
7200 (align_labels_max_skip): Redefine macro to use new types.
7201 (align_functions_log): Redefine macro to use new types.
7202 (align_loops): Redefine macro to use new types.
7203 (align_jumps): Redefine macro to use new types.
7204 (align_labels): Redefine macro to use new types.
7205 (align_functions): Redefine macro to use new types.
7206 (align_functions_max_skip): Redefine macro to use new types.
7207 (align_loops_value): New macro.
7208 (align_jumps_value): New macro.
7209 (align_labels_value): New macro.
7210 (align_functions_value): New macro.
7211 * function.c (invoke_set_current_function_hook): Propagate
7212 alignment values from flags to global variables default in
7214 * ipa-icf.c (sem_function::equals_wpa): Use
7215 cl_optimization_option_eq instead of memcmp.
7216 * lto-streamer.h (cl_optimization_stream_out): Support streaming
7218 (cl_optimization_stream_in): Likewise.
7219 * optc-save-gen.awk: Support strings in cl_optimization.
7220 * opth-gen.awk: Likewise.
7221 * opts.c (finish_options): Remove error checking of invalid
7223 (MAX_CODE_ALIGN): Remove.
7224 (MAX_CODE_ALIGN_VALUE): Likewise.
7225 (parse_and_check_align_values): New function.
7226 (check_alignment_argument): Likewise.
7227 (common_handle_option): Use check_alignment_argument.
7228 * opts.h (parse_and_check_align_values): Declare.
7229 * toplev.c (init_alignments): Remove.
7230 (read_log_maxskip): New.
7231 (parse_N_M): Likewise.
7232 (parse_alignment_opts): Likewise.
7233 (backend_init_target): Remove usage of init_alignments.
7234 * toplev.h (parse_alignment_opts): Declare.
7235 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
7237 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
7238 * tree.c (cl_option_hasher::equal): New.
7239 * varasm.c: Use new global macros.
7241 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
7243 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
7244 Use a simpler align directive also if MAXSKIP = ALIGN-1.
7245 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7246 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7247 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7248 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
7249 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
7250 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
7251 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
7252 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7253 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7254 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7255 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7257 2018-07-04 Martin Liska <mliska@suse.cz>
7258 Jonathan Wakely <jwakely@redhat.com>
7260 * coverage.c: Use correct type.
7261 * doc/invoke.texi: Language correction.
7263 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
7266 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
7267 ENDBRANCH for non-tail call which may return via indirect branch.
7268 * doc/extend.texi: Document indirect_return attribute.
7270 2018-07-03 Martin Sebor <msebor@redhat.com>
7272 PR tree-optimization/86274
7273 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
7275 (format_floating): Correct handling of infinities and NaNs.
7277 2018-07-03 Martin Sebor <msebor@redhat.com>
7279 * print-tree.c (print_real_cst): New function.
7280 (print_node_brief): Call it.
7281 (print_node): Ditto.
7283 2018-07-03 Jeff Law <law@redhat.com>
7285 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
7286 into a single pattern.
7288 * config/h8300/h8300.md (ors code_iterator): New.
7289 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
7290 a single pattern and single splitter.
7291 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
7292 (iorqi3_1, xorqi3_1): Likewise.
7293 (iorqi3, xorqi3 expanders): Similarly.
7295 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
7296 (movmd_internal) into a single pattern using the P mode iterator.
7297 (movmd splitters): Similarly.
7298 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
7299 (movsd splitters): Similarly.
7301 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
7302 ADDB, ADDW and ADDL into a single ADD attribute which selects the
7303 right table based on the size of the operand.
7304 * config/h8300/h8300.md (length_table): Corresponding changes. All
7305 references to "addb", "addw" and "addl" changed to "add".
7306 (btst patterns): Merge two variants into a single pattern.
7307 (tstqi, tsthi): Likewise.
7308 (addhi3_incdec, addsi3_incdec): Likewise.
7309 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
7310 (mulhi3, mulsi3): Likewise.
7311 (udivhi3, udivsi3): Likewise.
7312 (divhi3, divsi3): Likewise.
7313 (andorqi3, andorhi3, andorsi3): Likewise.
7315 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
7318 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
7319 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
7321 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7323 PR tree-optimization/85694
7324 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
7325 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
7327 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
7328 (<u>avg<mode>3_ceil): New patterns.
7330 2018-07-03 David Malcolm <dmalcolm@redhat.com>
7332 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
7333 scan-tree-dump directive.
7334 * gcc.dg/vect/slp-perm-2.c: Likewise.
7335 * gcc.dg/vect/slp-perm-3.c: Likewise.
7336 * gcc.dg/vect/slp-perm-5.c: Likewise.
7337 * gcc.dg/vect/slp-perm-6.c: Likewise.
7338 * gcc.dg/vect/slp-perm-7.c: Likewise.
7339 * gcc.dg/vect/slp-perm-8.c: Likewise.
7341 2018-07-03 Marek Polacek <polacek@redhat.com>
7344 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
7346 2018-07-03 Richard Biener <rguenther@suse.de>
7349 * tree-ssa-structalias.c (find_func_clobbers): Properly
7350 handle indirect calls.
7352 2018-07-03 Jeff Law <law@redhat.com>
7354 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
7355 (shifts): New code iterator.
7356 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
7357 expander. Fix HImode handling on H8/SX.
7358 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
7359 (subqi3, subhi3, subsi3 expanders): Likewise.
7360 (andqi3, andhi3, andsi3 expanders): Likewise.
7361 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
7362 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
7363 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
7364 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
7365 (zero_extendqihi2, zero_extendqisi2): Likewise.
7366 (extendqihi2, extendqisi2): Likewise.
7367 (rotlqi3, rotlhi3, rotlsi3): Likewise.
7368 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
7369 (rotlqi3_1, rotlhi3_1): Likewise.
7370 (logicalhi3_sn, logicalsi3_sn): Likewise.
7371 (logicalhi3, logicalsi3): Likewise.
7373 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7375 * tree-vect-patterns.c (vect_recog_rotate_pattern)
7376 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
7377 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
7378 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
7379 type to append_pattern_def_seq instead of creating a stmt_vec_info
7381 (build_mask_conversion): Likewise. Remove vinfo argument.
7382 (vect_add_conversion_to_patterm): Likewise, renaming to...
7383 (vect_add_conversion_to_pattern): ...this.
7384 (vect_recog_mask_conversion_pattern): Update call to
7385 build_mask_conversion. Pass the vector type to
7386 append_pattern_def_seq here too.
7387 (vect_recog_gather_scatter_pattern): Update call to
7388 vect_add_conversion_to_pattern.
7390 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7392 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
7393 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7394 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
7395 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
7396 STMT_VINFO_PATTERN_DEF_SEQ to null here.
7397 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
7398 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
7399 append_pattern_def_seq instead of new_pattern_def_seq.
7400 (vect_recog_divmod_pattern): Do both of the above.
7401 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
7404 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7406 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
7407 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
7408 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
7409 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
7410 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
7411 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
7412 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
7413 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
7414 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
7415 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
7416 parameter with a single stmt_vec_info.
7417 (vect_recog_func_ptr): Likewise.
7418 (vect_recog_gather_scatter_pattern): Likewise, folding in...
7419 (vect_try_gather_scatter_pattern): ...this.
7420 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
7421 the stmt_vec_info of the statement to be matched. Don't clear
7422 STMT_VINFO_RELATED_STMT.
7423 (vect_pattern_recog): Update call accordingly.
7425 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7427 PR tree-optimization/85694
7428 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
7429 (uavgM3_ceil): Document new optabs.
7430 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
7431 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
7433 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
7434 (savg_ceil_optab): New optabs.
7435 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
7436 (vect_vect_recog_func_ptrs): Add it.
7437 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
7438 constant directly from the associated lhs.
7440 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7442 * tree-vect-patterns.c (vect_split_statement): New function.
7443 (vect_convert_input): Use it to try to split an existing cast.
7445 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7447 * poly-int.h (print_hex): New function.
7448 * dumpfile.h (dump_dec, dump_hex): Declare.
7449 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
7450 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
7451 min_input_precision, operation_precision and operation_sign.
7452 * tree-vect-patterns.c (vect_get_range_info): New function.
7453 (vect_same_loop_or_bb_p, vect_single_imm_use)
7454 (vect_operation_fits_smaller_type): Delete.
7455 (vect_look_through_possible_promotion): Add an optional
7456 single_use_p parameter.
7457 (vect_recog_over_widening_pattern): Rewrite to use new
7458 stmt_vec_info infomration. Handle one operation at a time.
7459 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
7460 (vect_truncatable_operation_p, vect_set_operation_type)
7461 (vect_set_min_input_precision): New functions.
7462 (vect_determine_min_output_precision_1): Likewise.
7463 (vect_determine_min_output_precision): Likewise.
7464 (vect_determine_precisions_from_range): Likewise.
7465 (vect_determine_precisions_from_users): Likewise.
7466 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
7467 (vect_vect_recog_func_ptrs): Put over_widening first.
7469 (vect_pattern_recog): Call vect_determine_precisions.
7471 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7473 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
7474 statements that have been replaced by further pattern statements.
7475 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
7477 2018-07-03 Richard Biener <rguenther@suse.de>
7479 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
7480 always set *dt. Dump vectype in vectype overload.
7481 * dumpfile.h (dump_gimple_expr): New function.
7482 (dump_gimple_expr_loc): Likewise.
7483 * dumpfile.c (dump_gimple_expr): New function.
7484 (dump_gimple_expr_loc): Likewise.
7486 2018-07-02 Jeff Law <law@redhat.com>
7488 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
7489 the H8/300, H8/300H and H8/S variants into a single pattern.
7490 (movhi_h8300, movqi_h8300hs): Similarly.
7491 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
7492 (QHI mode iterator): New.
7494 * config/h8300/h8300.md: Remove trailing whitespace.
7496 2018-07-02 Jim Wilson <jimw@sifive.com>
7498 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
7499 instead of emit_insn for interrupt returns.
7500 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
7501 (riscv_sret, riscv_uret): Likewise.
7503 2018-07-02 David Malcolm <dmalcolm@redhat.com>
7505 * pretty-print.c (selftest::test_pp_format): Move save and restore
7506 of quotes to class auto_fix_quotes, and add an instance.
7507 * selftest.c: Include "intl.h".
7508 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
7509 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
7510 * selftest.h (selftest::auto_fix_quotes): New class.
7512 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
7514 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
7515 (aarch64_sve_prepare_conditional_op): Remove.
7516 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
7517 Allow aarch64_simd_reg_or_zero as select operand; remove
7518 the aarch64_sve_prepare_conditional_op call.
7519 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
7520 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
7521 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
7522 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
7523 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
7524 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
7525 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
7526 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
7527 and a splitters to match all of the *_any patterns.
7528 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
7530 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
7531 (SVE_COND_FP_BINARY_REV): Remove.
7532 (sve_int_op_rev, sve_fp_op_rev): New.
7533 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
7534 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
7535 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
7536 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
7537 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
7538 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
7539 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
7540 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
7541 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
7543 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
7544 Remove match_dup 1 from the inner unspec.
7545 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
7547 * config/aarch64/aarch64.md (movprfx): New attr.
7548 (length): Default movprfx to 8.
7549 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
7550 (*madd<SVE_I>, *msub<SVE_I): Likewise.
7551 (*<su>mul<SVE_I>3_highpart): Likewise.
7552 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
7553 (*v<ASHIFT><SVE_I>3): Likewise.
7554 (*<su><MAXMIN><SVE_I>3): Likewise.
7555 (*<su><MAXMIN><SVE_F>3): Likewise.
7556 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
7557 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
7558 (*div<SVE_F>4): Likewise.
7560 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
7562 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
7565 2018-07-02 Richard Biener <rguenther@suse.de>
7567 PR tree-optimization/86363
7568 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
7569 memset argument refers to a non-variable address.
7571 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
7573 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
7574 {PLUS,MINUS}_EXPR code to...
7575 (adjust_symbolic_bound): ...here,
7576 (combine_bound): ...here,
7577 (set_value_range_with_overflow): ...and here.
7579 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
7581 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
7583 (extract_range_from_abs_expr): ...here.
7585 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
7587 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
7588 -fno-omit-frame-pointer when not optimizing.
7590 2018-07-02 Martin Liska <mliska@suse.cz>
7593 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
7594 (propagate_nothrow): Likewise.
7596 2018-07-02 Martin Liska <mliska@suse.cz>
7599 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
7601 2018-07-02 David Malcolm <dmalcolm@redhat.com>
7603 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
7604 function in r262149, changing "loc" param from source_location to
7605 const dump_location_t &.
7606 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
7607 declaration, as above.
7609 2018-07-01 Paul Koning <ni1d@arrl.net>
7611 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
7612 -munit-asm, -mgnu-asm, -mdec-asm.
7613 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
7614 (pdp11_output_labelref): New.
7615 (pdp11_output_def): New.
7616 (pdp11_output_addr_vec_elt): New.
7617 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
7618 %# and %@ format codes.
7619 (pdp11_option_override): New.
7620 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
7621 (pdp11_output_ident): New.
7622 (pdp11_asm_named_section): New.
7623 (pdp11_asm_init_sections): New.
7624 (pdp11_file_start): New.
7625 (pdp11_file_end): New.
7626 (output_ascii): Use .ascii/.asciz for -mdec-asm.
7627 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
7628 %o, like %c but octal.
7629 (pdp11_option_override): New.
7630 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
7632 (DATA_SECTION_ASM_OP): Ditto.
7633 (READONLY_DATA_SECTION_ASM_OP): New.
7634 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
7635 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
7636 (ASM_OUTPUT_LABELREF): Ditto.
7637 (ASM_OUTPUT_DEF): Ditto.
7638 (ASM_OUTPUT_EXTERNAL): New.
7639 (ASM_OUTPUT_SOURCE_FILENAME): New.
7640 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
7641 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
7642 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
7643 %# and %@ format codes.
7644 * config/pdp11/pdp11.opt (mgnu-asm): New.
7645 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
7646 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
7647 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
7649 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
7651 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
7652 dereferencing path[] beyond its length.
7654 (debug_all_paths): New.
7655 (rewire_first_differing_edge): New.
7656 (adjust_paths_after_duplication): New.
7657 (duplicate_thread_path): Call adjust_paths_after_duplication.
7659 (thread_through_all_blocks): Add new argument to
7660 duplicate_thread_path.
7662 2018-06-30 Jim Wilson <jimw@sifive.com>
7664 * config/riscv/predicates.md (p2m1_shift_operand): New.
7665 (high_mask_shift_operand): New.
7666 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
7667 pattern using p2m1_shift_operand.
7668 (lshsi3_zero_extend_3+2): New combiner pattern using
7669 high_mask_shift_operand.
7671 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7673 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
7675 (vect_recog_rotate_pattern): ...here.
7676 (vect_convert_input): Try to insert casts of invariants in the
7678 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
7679 preheader to be empty.
7681 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7683 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
7684 vector type. If given, install it in the new statement's
7686 (vect_element_precision): New function.
7687 (vect_unpromoted_value): New struct.
7688 (vect_unpromoted_value::vect_unpromoted_value): New function.
7689 (vect_unpromoted_value::set_op): Likewise.
7690 (vect_look_through_possible_promotion): Likewise.
7691 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
7692 (vect_widened_op_tree, vect_convert_input): Likewise.
7693 (vect_convert_inputs, vect_convert_output): Likewise.
7694 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
7695 to handle the optional cast of the multiplication result and
7696 vect_widened_op_tree to detect the widened multiplication itself.
7697 Do not require the input and output of promotion casts to have
7698 the same sign, but base the signedness of the operation on the
7699 input rather than the result. If the pattern includes two
7700 promotions, check that those promotions have the same sign.
7701 Do not restrict the MULT_EXPR handling to a double-width result;
7702 handle quadruple-width results and wider. Use vect_convert_inputs
7703 to convert the inputs to the common type.
7704 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
7705 to handle the optional cast of the ABS result. Also allow a sign
7706 change or a sign extension between the ABS and MINUS.
7707 Use vect_widened_op_tree to detect the widened subtraction and use
7708 vect_convert_inputs to convert the inputs to the common type.
7709 (vect_handle_widen_op_by_const): Delete.
7710 (vect_recog_widen_op_pattern): New function.
7711 (vect_recog_widen_mult_pattern): Use it.
7712 (vect_recog_widen_shift_pattern): Likewise.
7713 (vect_recog_widen_sum_pattern): Use
7714 vect_look_through_possible_promotion to handle the promoted
7717 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7719 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
7720 the containing gimple_seq *.
7721 * gimple-iterator.h (gsi_for_stmt): Declare it.
7722 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
7723 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
7724 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
7725 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
7726 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
7728 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
7730 (vect_mark_pattern_stmts): ...here. Handle cases in which the
7731 statement being replaced is part of an existing pattern
7732 definition sequence, inserting the new pattern statements before
7734 (vect_pattern_recog_1): Don't return a bool. If the statement
7735 is already part of a pattern, instead apply pattern matching
7736 to the pattern definition statements. Don't clear the
7737 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
7738 (vect_pattern_recog): Don't break after the first match;
7739 continue processing the pattern definition statements instead.
7740 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
7742 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7744 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7745 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7746 (vect_recog_widen_sum_pattern): Use it.
7748 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7750 * tree-vect-loop.c (vectorizable_reduction): Assert that the
7751 phi is not a pattern statement and has not been replaced by
7752 a pattern statement.
7753 * tree-vect-patterns.c (type_conversion_p): Don't check
7754 STMT_VINFO_IN_PATTERN_P.
7755 (vect_recog_vector_vector_shift_pattern): Likewise.
7756 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
7757 the pattern statement rather than the original statement; check
7758 directly for a WIDEN_MULT_EXPR here.
7759 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
7760 vect_is_simple_use to return the pattern statement rather
7761 than the original statement; use is_pattern_stmt_p to check
7762 for such a pattern statement.
7763 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
7764 to return the pattern statement rather than the original statement;
7765 don't do the same transformation here.
7766 (vect_is_simple_use): If the defining statement has been replaced
7767 by a pattern statement, return the pattern statement instead.
7768 Remove the corresponding (local) transformation from the vectype
7771 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7773 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
7774 end and default to null.
7775 * tree-vect-loop.c (vect_create_epilog_for_reduction)
7776 (vectorizable_reduction): Update calls accordingly, dropping the
7777 gimple ** argument if the passed-back statement isn't needed.
7778 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
7779 (vect_recog_rotate_pattern): Likewise.
7780 (vect_recog_mask_conversion_pattern): Likewise.
7781 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
7782 (vect_mask_constant_operand_p): Likewise.
7783 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
7784 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
7785 (get_group_load_store_type, get_load_store_type): Likewise.
7786 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
7787 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
7788 (vectorizable_conversion, vectorizable_assignment): Likewise.
7789 (vectorizable_shift, vectorizable_operation): Likewise.
7790 (vectorizable_store, vect_is_simple_cond): Likewise.
7791 (vectorizable_condition, vectorizable_comparison): Likewise.
7792 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
7793 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
7794 and move it to the end. Cope with null def_stmt_outs.
7796 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
7798 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
7800 2018-06-29 Jeff Law <law@redhat.com>
7802 * config/v850/v850.c (v850_legitimate_address_p): Handle large
7803 displacements for TARGET_V850E2V3 and newer.
7804 (TARGET_LRA_P): Remove. Defaults to LRA now.
7805 * config/v850/v850.md (sign23byte_load): Remove.
7806 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
7807 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
7809 2018-06-29 Martin Liska <mliska@suse.cz>
7812 * coverage.c (coverage_init): Mangle full path name.
7813 * doc/invoke.texi: Document the change.
7814 * gcov-io.c (mangle_path): New.
7815 * gcov-io.h (mangle_path): Likewise.
7816 * gcov.c (mangle_name): Use mangle_path for path mangling.
7818 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7820 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
7821 if starting source register is not even.
7823 2018-06-29 Martin Liska <mliska@suse.cz>
7825 PR tree-optimization/86263
7826 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
7827 Make edge redirection.
7829 2018-06-29 David Malcolm <dmalcolm@redhat.com>
7831 * dumpfile.c (dump_loc): Add indentation based on scope depth.
7832 (dump_scope_depth): New variable.
7833 (get_dump_scope_depth): New function.
7834 (dump_begin_scope): New function.
7835 (dump_end_scope): New function.
7836 * dumpfile.h (get_dump_scope_depth): New declaration.
7837 (dump_begin_scope): New declaration.
7838 (dump_end_scope): New declaration.
7839 (class auto_dump_scope): New class.
7840 (AUTO_DUMP_SCOPE): New macro.
7841 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
7844 2018-06-29 Richard Biener <rguenther@suse.de>
7846 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
7847 compute_all_dependences succeeds.
7848 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
7849 exceed --param loop-max-datarefs-for-datadeps.
7851 2018-06-29 Jakub Jelinek <jakub@redhat.com>
7853 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
7855 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
7858 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
7859 alternative 0 in preferred_for_speed attribute.
7861 2018-06-28 Paul Koning <ni1d@arrl.net>
7863 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
7864 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
7865 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
7866 * config/pdp11/pdp11.md: Correct "length" attribute calculation
7867 for shift insn patterns.
7869 2018-06-28 David Malcolm <dmalcolm@redhat.com>
7871 * cgraph.c (cgraph_node::get_body): Replace assignments to
7872 "dump_file" with calls to set_dump_file.
7873 * dumpfile.c (alt_dump_file): Make static, and group with...
7874 (alt_flags): ...this definition.
7875 (dumps_are_enabled): New variable.
7876 (refresh_dumps_are_enabled): New function.
7877 (set_dump_file): New function.
7878 (set_alt_dump_file): New function.
7879 (gcc::dump_manager::dump_start): Replace assignments to
7880 "dump_file" and "alt_dump_file" with calls to set_dump_file and
7882 (gcc::dump_manager::dump_finish): Likewise.
7883 * dumpfile.h (alt_dump_file): Delete decl.
7884 (dumps_are_enabled): New variable decl.
7885 (set_dump_file): New function decl.
7886 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
7888 * tree-nested.c (lower_nested_functions): Replace assignments to
7889 "dump_file" with calls to set_dump_file.
7891 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
7893 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
7894 goto_locus of each outgoing edge of each basic block.
7896 2018-06-28 Richard Biener <rguenther@suse.de>
7898 * dwarf2out.c (decl_scope_table): Remove.
7899 (push_decl_scope): Likewise.
7900 (pop_decl_scope): Likewise.
7901 (gen_type_die_for_member): Do not call push/pop_decl_scope.
7902 (gen_struct_or_union_type_die): Likewise.
7903 (gen_tagged_type_die): Likewise.
7904 (dwarf2out_init): Do not initialize decl_scope_table.
7905 (dwarf2out_c_finalize): Do not free it.
7907 2018-06-28 Richard Biener <rguenther@suse.de>
7909 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
7910 deciding whether to not re-use a DIE.
7912 2018-06-28 Richard Biener <rguenther@suse.de>
7914 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
7915 DW_AT_abstract_origin attribute.
7917 2018-06-28 Martin Liska <mliska@suse.cz>
7919 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
7920 Use newly introduced constants.
7921 * tree-switch-conversion.h (struct jump_table_cluster):
7922 Define max_ratio_for_size and max_ratio_for_speed.
7924 2018-06-28 Martin Liska <mliska@suse.cz>
7926 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
7927 Add new checking assert to catch invalid state.
7928 (jump_table_cluster::can_be_handled): Handle single case
7930 (jump_table_cluster::is_beneficial): Bail out for such case.
7931 (bit_test_cluster::find_bit_tests):
7932 Add new checking assert to catch invalid state.
7933 (bit_test_cluster::can_be_handled): Handle single case
7935 (bit_test_cluster::is_beneficial): Bail out for such case.
7936 (switch_decision_tree::analyze_switch_statement):
7939 2018-06-28 Martin Liska <mliska@suse.cz>
7941 * common.opt: Introduce -completion option.
7942 * gcc.c (driver_handle_option): Handle it.
7943 (driver::main): Print completions if completion
7945 * opt-suggestions.c (option_proposer::get_completions):
7947 (option_proposer::suggest_completion): Likewise.
7948 (option_proposer::find_param_completions): Likewise.
7949 (verify_autocompletions): Likewise.
7950 (test_completion_valid_options): Likewise.
7951 (test_completion_valid_params): Likewise.
7952 (in_completion_p): Likewise.
7953 (empty_completion_p): Likewise.
7954 (test_completion_partial_match): Likewise.
7955 (test_completion_garbage): Likewise.
7956 (opt_proposer_c_tests): Likewise.
7957 * opt-suggestions.h: Declare new functions.
7958 * opts.c (common_handle_option): Handle OPT__completion_.
7959 * selftest-run-tests.c (selftest::run_tests): Add
7960 opt_proposer_c_tests.
7961 * selftest.c (assert_str_startswith): New.
7962 * selftest.h (assert_str_startswith): Likewise.
7963 (opt_proposer_c_tests): New.
7964 (ASSERT_STR_STARTSWITH): Likewise.
7966 2018-06-28 Martin Liska <mliska@suse.cz>
7968 * Makefile.in: Add opt-suggestions.o.
7969 * gcc-main.c: Include opt-suggestions.h.
7970 * gcc.c (driver::driver): Likewise.
7971 (driver::~driver): Remove m_option_suggestions.
7972 (driver::build_option_suggestions): Moved to option_proposer.
7973 (driver::suggest_option): Likewise.
7974 (driver::handle_unrecognized_options): Use option_proposer.
7975 * gcc.h (class driver): Add new memver m_option_proposer.
7976 * opt-suggestions.c: New file.
7977 * opt-suggestions.h: New file.
7979 2018-06-28 Martin Liska <mliska@suse.cz>
7981 * vec.h (class auto_string_vec): New (moved from auto_argvec).
7982 (auto_string_vec::~auto_string_vec): Likewise.
7984 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
7986 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
7987 prevent_decl_creation_for_types fields up and add reset_location field.
7988 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
7989 statement if id->reset_location is true.
7990 (copy_edges_for_bb): Do not set goto_locus on the new edges if
7991 id->reset_location is true.
7992 (copy_phis_for_bb): Force input_location on the arguments if
7993 id->reset_location is true.
7994 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
7995 is set on the function to be inlined.
7997 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
7999 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
8001 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
8003 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
8004 registers for Pmode.
8005 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
8006 hard registers for the clobbered pseudo.
8008 2018-06-27 Paul Koning <ni1d@arrl.net>
8010 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
8011 mutually exclusive options.
8012 * config/pdp11/constraints.md (h): New constraint.
8013 (O): Update definition to match shift code generation.
8014 (D): New constraint.
8015 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
8017 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
8019 (output_jump): Change arguments.
8020 (pdp11_fixed_cc_regs): New function.
8021 (pdp11_cc_mode): Ditto.
8022 (pdp11_expand_shift): Ditto.
8023 (pdp11_assemble_shift): Ditto.
8024 (pdp11_small_shift): Ditto.
8025 (pdp11_branch_cost): Remove.
8026 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
8028 (pdp11_register_move_cost): Update for CC registers.
8029 (pdp11_rtx_costs): Add case for LSHIFTRT.
8030 (pdp11_output_jump): Add CCNZ mode conditional branches.
8031 (notice_update_cc_on_set): Remove.
8032 (pdp11_cc_mode): New function.
8033 (simple_memory_operand): Correct pre/post decrement case.
8034 (no_side_effect_operand): New function.
8035 (pdp11_regno_reg_class): Add CC_REGS class.
8036 (pdp11_fixed_cc_regs): New function.
8037 (pdp11_small_shift): New function.
8038 (pdp11_expand_shift): New function to expand shift insns.
8039 (pdp11_assemble_shift): New function to output shifts.
8040 (pdp11_branch_cost): Remove.
8041 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
8042 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
8043 (WCHAR_TYPE): Ditto.
8044 (PTRDIFF_TYPE): Ditto.
8045 (ADJUST_INSN_LENGTH): New macro.
8046 (FIXED_REGISTERS): Add CC registers.
8047 (CALL_USED_REGISTERS): Ditto.
8049 (REG_CLASS_NAMES): Ditto.
8050 (REG_CLASS_CONTENTS): Ditto.
8051 (SELECT_CC_MODE): Use new function.
8052 (TARGET_FLAGS_REGNUM): New macro.
8053 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
8054 (cc0_reg_rtx): Remove.
8055 (CC_STATUS_MDEP): Remove.
8056 (CC_STATUS_MDEFP_INIT): Remove.
8057 (CC_IN_FPU): Remove.
8058 (NOTICE_UPDATE_CC): Remove.
8059 (REGISTER_NAMES): Add CC registers.
8060 (BRANCH_COST): Change to constant 1.
8061 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
8063 * config/pdp11/pdp11.opt (mbcopy): Remove.
8064 (mbcopy-builtin): Remove.
8065 (mbranch-cheap): Remove.
8066 (mbranch-expensive): Remove.
8067 * config/pdp11/predicates.md (expand_shift_operand): Update to
8068 match shift code generation.
8069 (ccnz_operator): New predicate.
8070 * doc/invoke.texi (PDP-11 Options): Remove deleted options
8071 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
8072 Remove non-existent option -mabshi, -mno-abshi. Document mutually
8074 * doc/md.texi (PDP-11): Document new D and h constraints. Update
8075 description of O constraint.
8077 2018-06-27 Jeff Law <law@redhat.com>
8078 Austin Law <austinklaw@gmail.com>
8080 * config/v850/v850.md (addsi3_set_flags): New pattern.
8081 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
8082 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
8083 (zero_extendhisi2_v850_set_flags): Likewise.
8084 (zero_extendqisi2_v850_set_flags): Likewise.
8085 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
8086 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
8087 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
8089 * config/v850/v850-protos.h (notice_update_cc): Remove.
8090 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
8091 (v850_print_operand): Handle 'D' and "d".
8092 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
8093 Add handling of arithmetic/logical operations compared against zero.
8094 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
8095 Do not look at v850_compare_op, instead get mode from last argument.
8096 (v850_gen_compare): Remove
8097 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
8098 after reload for prologue insns.
8099 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
8101 (construct_save_jarl): Likewise.
8102 (TARGET_FLAGS_REGNUM): Define.
8103 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
8104 (NOTICE_UPDATE_CC): Remove.
8105 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
8106 than cc0. Conditionalize on reload_completed.
8107 (cmpsi_insn, setfcc_insn): Likewise.
8108 (tst1 splitter): Turn into define_and_split which sets the flags
8110 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
8111 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
8112 (cstoresf4, cstoredf4): Clobber the flags.
8113 (cmpsi, cmpsf, cmpdf): Remove expanders.
8114 (setf_insn): Remove pattern.
8115 (addsi3): Turn into define_and_split which clobbers the flags after
8116 reload and a suitable pattern (addsi3_clobber_flags) for use after
8118 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
8119 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
8120 (ashrsi3, ashrsi3_v850e2): Likewise.
8121 (bins): Clobber the flags.
8122 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
8123 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
8124 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
8125 (call_value_internal_short, call_value_internal_long): Likewise.
8126 (callt_save_interrupt, callt_return_interrupt): Likewise.
8127 (save_interrupt, return_interrupt): Likewise.
8128 (callt_save_all_interrupt, save_all_interrupt): Likewise.
8129 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
8130 (restore_all_interrupt, _restore_all_interrupt): Likewise.
8131 (All FP comparisons): Only allow after reload has completed.
8133 (divh, divhu): Tweak output template.
8134 (branch_z_normal, branch_z_invert): Remove
8135 (branch_nz_normal, branch_nz_invert): Likewise.
8136 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
8138 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
8139 * config/v850/v850.c (notice_update_cc): Remove.
8140 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
8141 (CC_NO_CARRY): Likewise.
8142 (NOTICE_UPDATE_CC): Define to nothing.
8143 * config/v850/v850.md: Remove block comment on cc0 handling
8144 Remove "cc" attribute from all patterns. Remove cc_status handling
8145 from all patterns. Minor formatting fixes.
8147 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8149 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
8150 (cortex-a76.cortex-a55): Likewise.
8151 * config/aarch64/aarch64-tune.md: Regenerate.
8152 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
8153 cortex-a76.cortex-a55.
8155 2018-06-27 Jeff Law <law@redhat.com>
8157 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
8158 (MULTILIB_DIRNAMES): Similarly.
8160 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
8162 * gimple.h (gimple_return_retbnd): Delete.
8163 (gimple_return_set_retbnd): Likewise.
8164 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
8165 gimple_return_set_retbnd.
8166 * gimple-pretty-print.c (dump_gimple_return): Remove call to
8167 gimple_return_retbnd and adjust.
8168 * tree-inline.h (struct copy_body_data): Remove retbnd field.
8169 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
8170 Explicitly return NULL in a couple more cases. Move assertion
8171 on debug statements and remove unreachable code.
8172 (reset_debug_binding): Do not test id->retbnd.
8173 (expand_call_inline): Do not set it.
8175 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
8177 * configure.ac: Add --disable-gcov option.
8178 * configure: Regenerate.
8179 * Makefile.in: Honour @enable_gcov@.
8180 * doc/install.texi: Document --disable-gcov.
8182 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8184 * config/arm/arm-cpus.in (cortex-a76): New entry.
8185 (cortex-a76.cortex-a55): Likewise.
8186 * config/arm/arm-tables.opt: Regenerate.
8187 * config/arm/arm-tune.md: Likewise.
8188 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
8189 * doc/invoke.texi (ARM Options): Document cortex-a76 and
8190 cortex-a76.cortex-a55.
8192 2018-06-27 Tamar Christina <tamar.christina@arm.com>
8195 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
8197 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
8199 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
8201 (EPILOGUE_USES): Likewise.
8203 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
8205 * tree-inline.c (remap_location): New function extracted from...
8206 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
8207 (copy_phis_for_bb): ...here. Call remap_location.
8208 (copy_cfg_body): Adjust call to copy_edges_for_bb.
8210 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
8212 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
8213 unaligned vsx for 16B memset.
8215 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
8218 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
8219 ieee128_float_type_node to long_double_type_node unless
8220 TARGET_LONG_DOUBLE_128 is set.
8222 2018-06-26 David Malcolm <dmalcolm@redhat.com>
8224 * cfgloop.c (get_loop_location): Convert return type from
8225 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
8226 by implicit construction from rtx_insn *, and using
8227 dump_user_location_t::from_function_decl for the fallback case.
8228 * cfgloop.h (get_loop_location): Convert return type from
8229 location_t to dump_user_location_t.
8230 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
8231 dump_printf_loc to pass in a dump_location_t rather than a
8232 location_t, via the gimple stmt.
8233 * coverage.c (get_coverage_counts): Update calls to
8234 dump_printf_loc to pass in dump_location_t rather than a
8236 * doc/optinfo.texi (Dump types): Convert example of
8237 dump_printf_loc from taking "locus" to taking "insn". Update
8238 description of the "_loc" calls to cover dump_location_t.
8239 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
8241 (dump_user_location_t::dump_user_location_t): New constructors,
8242 from gimple *stmt and rtx_insn *.
8243 (dump_user_location_t::from_function_decl): New function.
8244 (dump_loc): Make static.
8245 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
8246 const dump_location_t &.
8247 (dump_generic_expr_loc): Delete.
8248 (dump_printf_loc): Convert param "loc" from location_t to
8249 const dump_location_t &.
8250 (selftest::test_impl_location): New function.
8251 (selftest::dumpfile_c_tests): New function.
8252 * dumpfile.h: Include "profile-count.h".
8253 (class dump_user_location_t): New class.
8254 (struct dump_impl_location_t): New struct.
8255 (class dump_location_t): New class.
8256 (dump_printf_loc): Convert 2nd param from source_location to
8257 const dump_location_t &.
8258 (dump_generic_expr_loc): Delete.
8259 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
8260 const dump_location_t &.
8261 * gimple-fold.c (fold_gimple_assign): Update call to
8262 dump_printf_loc to pass in a dump_location_t rather than a
8263 location_t, via the gimple stmt.
8264 (gimple_fold_call): Likewise.
8265 * gimple-loop-interchange.cc
8266 (loop_cand::analyze_iloop_reduction_var): Update for change to
8267 check_reduction_path.
8268 (tree_loop_interchange::interchange): Update for change to
8270 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
8271 change in return-type of find_loop_location.
8272 (graphite_regenerate_ast_isl): Likewise.
8273 * graphite-optimize-isl.c (optimize_isl): Likewise.
8274 * graphite.c (graphite_transform_loops): Likewise.
8275 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
8276 pass in a dump_location_t rather than a location_t, via the
8278 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
8279 * ipa.c (walk_polymorphic_call_targets): Likewise.
8280 * loop-unroll.c (report_unroll): Convert "locus" param from
8281 location_t to dump_location_t.
8282 (decide_unrolling): Update for change to get_loop_location's
8284 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
8285 location_t to dump_user_location_t.
8286 (grid_find_single_omp_among_assignments_1): Updates calls to
8287 dump_printf_loc to pass in a dump_location_t rather than a
8288 location_t, via the gimple stmt.
8289 (grid_parallel_clauses_gridifiable): Convert "tloc" from
8290 location_t to dump_location_t. Updates calls to dump_printf_loc
8291 to pass in a dump_location_t rather than a location_t, via the
8293 (grid_inner_loop_gridifiable_p): Likewise.
8294 (grid_dist_follows_simple_pattern): Likewise.
8295 (grid_gfor_follows_tiling_pattern): Likewise.
8296 (grid_target_follows_gridifiable_pattern): Likewise.
8297 (grid_attempt_target_gridification): Convert initialization
8298 of local "grid" from memset to zero-initialization; FIXME: does
8299 this require C++11? Update call to dump_printf_loc to pass in a
8300 optinfo_location rather than a location_t, via the gimple stmt.
8301 * profile.c (read_profile_edge_counts): Updates call to
8302 dump_printf_loc to pass in a dump_location_t rather than a
8304 (compute_branch_probabilities): Likewise.
8305 * selftest-run-tests.c (selftest::run_tests): Call
8307 * selftest.h (dumpfile_c_tests): New decl.
8308 * tree-loop-distribution.c (pass_loop_distribution::execute):
8309 Update for change in return type of find_loop_location.
8310 * tree-parloops.c (parallelize_loops): Likewise.
8311 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
8312 "locus" from location_t to dump_user_location_t.
8313 (canonicalize_loop_induction_variables): Likewise.
8314 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
8315 for change in return type of find_loop_location.
8316 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
8317 to dump_printf_loc to pass in a dump_location_t rather than a
8318 location_t, via the stmt.
8319 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
8321 * tree-vect-loop-manip.c (find_loop_location): Convert return
8322 type from source_location to dump_user_location_t.
8323 (vect_do_peeling): Update for above change.
8324 (vect_loop_versioning): Update for change in type of
8326 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
8327 from location_t to dump_user_location_t.
8328 (vect_estimate_min_profitable_iters): Update for change in type
8330 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
8331 location_t to dump_location_t.
8332 (vect_slp_bb): Update for change in type of vect_location.
8333 * tree-vectorizer.c (vect_location): Convert from source_location
8334 to dump_user_location_t.
8335 (try_vectorize_loop_1): Update for change in vect_location's type.
8336 (vectorize_loops): Likewise.
8337 (increase_alignment): Likewise.
8338 * tree-vectorizer.h (vect_location): Convert from source_location
8339 to dump_user_location_t.
8340 (find_loop_location): Convert return type from source_location to
8341 dump_user_location_t.
8342 (check_reduction_path): Convert 1st param from location_t to
8343 dump_user_location_t.
8344 * value-prof.c (check_counter): Update call to dump_printf_loc to
8345 pass in a dump_user_location_t rather than a location_t; update
8346 call to error_at for change in type of "locus".
8347 (check_ic_target): Update call to dump_printf_loc to
8348 pass in a dump_user_location_t rather than a location_t, via the
8351 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
8353 * config/s390/s390.h (enum processor_flags): Do not use
8354 default tune parameter when -march was specified.
8356 2018-06-26 Jakub Jelinek <jakub@redhat.com>
8359 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
8360 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
8363 2018-06-26 Richard Biener <rguenther@suse.de>
8365 PR tree-optimization/86287
8367 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
8368 (vect_analyze_loop): Initialize n_stmts.
8370 2018-06-26 Richard Biener <rguenther@suse.de>
8373 * fold-const.c (fold_convertible_p): Pointer extension
8376 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
8379 * dwarf2out.c (loc_list_has_views): Adjust comments.
8380 (dw_loc_list): Split single cross-partition range with
8383 2018-06-25 Jeff Law <law@redhat.com>
8385 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
8386 on -mbig-switch by default.
8388 * config/v850/predicates.md (const_float_1_operand): Fix match_code
8390 (const_float_0_operand): Remove unused predicate.
8391 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
8392 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
8393 (recipsf2): New expander. Original pattern now called
8395 (recipdf2, recipdf2_insn): Similarly.
8396 (rsqrtsf2, rsqrtsf2_insn): Similarly
8397 (rsqrtdf2, rsqrtdf2_insn): Similarly
8399 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
8401 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
8402 Simplify logic for FreeBSD (twice).
8404 2018-06-25 Martin Sebor <msebor@redhat.com>
8406 PR tree-optimization/86204
8407 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
8408 a strnlen result if it's less than the length of the string.
8410 2018-06-25 Martin Sebor <msebor@redhat.com>
8412 PR tree-optimization/85700
8413 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
8414 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
8415 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
8417 2018-06-25 Martin Sebor <msebor@redhat.com>
8419 * doc/extend.texi (Zero-length arrays): Update and clarify.
8421 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
8423 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
8424 added IEEE/IBM long double multilib support on PowerPC little
8425 endian Linux systems.
8426 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
8427 (MULTILIB_DEFAULTS): Likewise.
8428 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8430 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
8431 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
8432 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
8434 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
8437 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
8438 (REORDER_45): Likewise.
8440 2018-06-25 Jeff Law <law@redhat.com>
8442 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
8443 dividend to 32 bits. Adjust length.
8444 (udivmodhi4): Cleanup output template. Fix length.
8446 2018-06-25 Carl Love <cel@us.ibm.com>
8448 * config/rs6000/vsx.md: Change word selector to prefered location.
8450 2018-06-25 Richard Biener <rguenther@suse.de>
8452 PR tree-optimization/86304
8453 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
8454 epilogue-if-converted loops as well.
8456 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
8458 * lto-section-out.c (lto_begin_section): Do not print section
8459 name for noaddr and unnumbered dumps.
8461 2018-06-25 Richard Biener <rguenther@suse.de>
8463 * tree-vectorizer.h (struct vec_info_shared): New structure
8464 with parts split out from struct vec_info and loop_nest from
8465 struct _loop_vec_info.
8466 (struct vec_info): Adjust accordingly.
8467 (struct _loop_vec_info): Likewise.
8468 (LOOP_VINFO_LOOP_NEST): Adjust.
8469 (LOOP_VINFO_DATAREFS): Likewise.
8470 (LOOP_VINFO_DDRS): Likewise.
8471 (struct _bb_vec_info): Likewise.
8472 (BB_VINFO_DATAREFS): Likewise.
8473 (BB_VINFO_DDRS): Likewise.
8474 (struct _stmt_vec_info): Add dr_aux member.
8475 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
8476 (DR_MISALIGNMENT_UNINITIALIZED): New.
8477 (set_dr_misalignment): Adjust.
8478 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
8479 (vect_analyze_loop): Adjust prototype.
8480 (vect_analyze_loop_form): Likewise.
8481 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
8482 Compute dependences lazily.
8483 (vect_record_base_alignments): Use shared datarefs/ddrs.
8484 (vect_verify_datarefs_alignment): Likewise.
8485 (vect_analyze_data_refs_alignment): Likewise.
8486 (vect_analyze_data_ref_accesses): Likewise.
8487 (vect_analyze_data_refs): Likewise.
8488 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
8489 constructor parameter for shared part.
8490 (vect_analyze_loop_form): Pass in shared part and adjust.
8491 (vect_analyze_loop_2): Pass in storage for the number of
8492 stmts. Move loop nest finding to the caller. Compute
8494 (vect_analyze_loop): Pass in shared part.
8495 (vect_transform_loop): Verify shared datarefs are unchanged.
8496 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
8497 constructor parameter for shared part.
8498 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
8499 (vect_slp_bb): Verify shared datarefs are unchanged before
8501 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
8503 (new_stmt_vec_info): Initialize DR_AUX misalignment to
8504 DR_MISALIGNMENT_UNINITIALIZED.
8505 * tree-vectorizer.c (vec_info::vec_info): Add constructor
8506 parameter for shared part.
8507 (vec_info::~vec_info): Adjust.
8508 (vec_info_shared::vec_info_shared): New.
8509 (vec_info_shared::~vec_info_shared): Likewise.
8510 (vec_info_shared::save_datarefs): Likewise.
8511 (vec_info_shared::check_datarefs): Likewise.
8512 (try_vectorize_loop_1): Construct shared part live for analyses
8513 of a single loop for multiple vector sizes.
8514 * tree-parloops.c (gather_scalar_reductions): Adjust.
8516 2018-06-25 Richard Biener <rguenther@suse.de>
8518 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
8519 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
8520 (vect_analyze_data_refs): Remove similar code from here and
8521 simplify accordingly.
8523 2018-06-25 Richard Biener <rguenther@suse.de>
8525 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
8526 for reverse storage order accesses rather than asserting
8527 they cannot happen here.
8529 2018-06-25 Tom de Vries <tdevries@suse.de>
8532 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
8533 Use data16 instead of .byte for insn prefix.
8535 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
8538 * parser.c (make_char_string_pack): Pass this literal chars
8539 through cpp_interpret_string.
8540 (cp_parser_userdef_numeric_literal): Check the result of
8541 make_char_string_pack.
8543 2018-06-24 Maya Rashish <coypu@sdf.org>
8545 * ginclude/stddef.h: Simplify conditions around avoiding
8546 re-definition of __size_t.
8548 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
8550 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
8551 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
8553 2018-06-22 Maya Rashish <coypu@sdf.org>
8555 * doc/invoke.texi (mno-fancy-math-387): Update for changes
8556 made to OpenBSD and NetBSD through the years.
8558 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8560 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
8561 behavior of vec_pack (vector double, vector double) to match
8562 behavior of vec_float2 (vector double, vector double).
8564 2018-06-22 Olivier Hainque <hainque@adacore.com>
8566 * gimplify.c (gimplify_function_tree): Prevent creation
8567 of a trampoline for the address of the current function
8568 passed to entry/exit instrumentation hooks.
8570 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
8573 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
8576 2018-06-22 Martin Liska <mliska@suse.cz>
8578 PR tree-optimization/86263
8579 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
8580 Bail out if is_enabled is false.
8581 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
8583 (jump_table_cluster::is_enabled): New function.
8585 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
8587 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
8588 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
8589 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
8590 (lto_input_ts_binfo_tree_pointers): Likewise.
8591 * tree-streamer-out.c (streamer_write_tree_bitfields,
8592 write_ts_binfo_tree_pointers): Likewise.
8593 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
8595 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
8597 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
8599 2018-06-22 Martin Liska <mliska@suse.cz>
8601 * symbol-summary.h (get): Make it pure and inline move
8602 functionality from ::get function.
8603 (get): Remove and inline into ::get and ::get_create.
8604 (get_create): Move code from ::get function.
8606 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8609 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
8610 -x assembler-with-cpp.
8612 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8614 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
8615 _FILE_OFFSET_BITS=64 for C++.
8617 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
8619 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
8620 conversion insn that shows up when pr85657-3.c is compiled using
8621 IEEE 128-bit long double.
8622 (neg<mode>2_internal): Use the correct mode to check whether the
8623 mode is IBM extended.
8624 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
8625 multiply and divide external functions from being created more
8628 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8630 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
8632 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
8633 the edge can be forwarded.
8634 (cfg_layout_merge_blocks): Likewise.
8636 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8638 * except.c (finish_eh_generation): Commit edge insertions only after
8639 the EH edges have been redirected from post-landing to landing pads.
8641 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8643 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
8644 create_tmp_var_for to create the FRAME decl.
8645 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
8647 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8649 * tree-inline.c (copy_edges_for_bb): Minor tweak.
8650 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
8651 debug statement when resetting its value.
8652 (expand_call_inline): Copy the locus of the call onto the assignment
8653 of the return value, if any. Use local variable in more cases.
8655 2018-06-21 Martin Liska <mliska@suse.cz>
8657 * ipa-pure-const.c (propagate_nothrow): Use
8658 funct_state_summaries->get.
8659 (dump_malloc_lattice): Likewise.
8660 (propagate_malloc): Likewise.
8662 2018-06-21 Richard Biener <rguenther@suse.de>
8664 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
8665 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
8666 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
8668 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
8669 BLOCK_ABSTRACT_ORIGIN unconditionally.
8671 2018-06-21 David Malcolm <dmalcolm@redhat.com>
8673 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
8675 * ipa-reference.c (ipa_reference_c_finalize): Delete
8676 ipa_ref_opt_sum_summaries and set it to NULL.
8678 2018-06-21 Tom de Vries <tdevries@suse.de>
8680 PR tree-optimization/85859
8681 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
8682 test with comment from bb_no_side_effects_p.
8684 2018-06-21 Richard Biener <rguenther@suse.de>
8686 PR tree-optimization/86232
8687 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
8688 max for constant niter.
8690 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
8692 * config/aarch64/aarch64-simd.md
8693 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
8695 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
8697 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
8698 Make opernads of the unspec commutative.
8700 2018-06-21 Richard Biener <rguenther@suse.de>
8702 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
8703 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
8704 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
8705 (vect_analyze_data_ref_dependence): Re-order checks to deal with
8707 (vect_record_base_alignments): Do not record base alignment
8708 for gathers or scatters.
8709 (vect_compute_data_ref_alignment): Drop return value that is always
8710 true. Bail out early for gathers or scatters.
8711 (vect_enhance_data_refs_alignment): Bail out early for gathers
8713 (vect_find_same_alignment_drs): Likewise.
8714 (vect_analyze_data_refs_alignment): Remove dead code.
8715 (vect_slp_analyze_and_verify_node_alignment): Likewise.
8716 (vect_analyze_data_refs): For possible gathers or scatters do
8717 not create an alternate DR, just check their possible validity
8718 and mark them. Adjust DECL_NONALIASED handling to not rely
8720 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
8721 update inits of gathers or scatters.
8722 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
8723 Also copy gather/scatter flag to pattern vinfo.
8725 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
8727 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
8728 behavior of vec_packsu (vector unsigned long long, vector unsigned
8729 long long) to match behavior of vec_packs with same signature.
8731 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
8732 Thomas Schwinge <thomas@codesourcery.com>
8733 Cesar Philippidis <cesar@codesourcery.com>
8735 * gimplify.c (gimplify_scan_omp_clauses): Add support for
8736 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
8737 (gimplify_adjust_omp_clauses): Likewise.
8738 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
8739 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
8740 (gimplify_omp_target_update): Update handling of acc update and
8742 * omp-low.c (install_var_field): Remove unused parameter
8743 base_pointers_restrict.
8744 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
8745 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
8747 (omp_target_base_pointers_restrict_p): Delete.
8748 (scan_omp_target): Update call to scan_sharing_clauses.
8749 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
8751 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
8752 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
8753 (convert_local_omp_clauses): Likewise.
8754 * tree-pretty-print.c (dump_omp_clause): Likewise.
8755 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
8757 (omp_clause_code_name): Likewise.
8759 2018-06-20 Jakub Jelinek <jakub@redhat.com>
8762 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
8765 PR tree-optimization/86231
8766 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
8767 anti-range don't overwrite *vr0min before using it to compute *vr0max.
8769 2018-06-20 Tom de Vries <tdevries@suse.de>
8771 PR tree-optimization/86097
8772 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
8773 iv type if signedness of iv type is not the same as that of *nit.
8775 2018-06-20 Jakub Jelinek <jakub@redhat.com>
8777 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
8778 EDGE_EH edges, verify they are all EDGE_EH.
8780 2018-06-20 Maya Rashish <coypu@sdf.org>
8782 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
8784 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8786 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
8787 * config/aarch64/aarch64.c (xgene1_tunings): Add
8788 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
8789 (aarch64_mode_valid_for_sched_fusion_p):
8790 Allow 16-byte modes.
8791 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
8792 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
8794 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
8796 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
8797 * config/aarch64/iterators.md (VQ2): New mode iterator.
8799 2018-06-20 Martin Liska <mliska@suse.cz>
8801 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
8802 Change default ratio from 10 to 8.
8804 2018-06-20 Martin Liska <mliska@suse.cz>
8806 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
8808 (bit_test_cluster::find_bit_tests): Likewise.
8809 (switch_decision_tree::analyze_switch_statement): Find clusters.
8810 * tree-switch-conversion.h (struct jump_table_cluster): Document
8813 2018-06-20 Martin Liska <mliska@suse.cz>
8815 * tree-switch-conversion.c (switch_conversion::collect):
8816 Record m_uniq property.
8817 (switch_conversion::expand): Bail out for special conditions.
8818 (group_cluster::~group_cluster): New.
8819 (group_cluster::group_cluster): Likewise.
8820 (group_cluster::dump): Likewise.
8821 (jump_table_cluster::emit): New.
8822 (switch_decision_tree::fix_phi_operands_for_edges): New.
8823 (struct case_node): Remove struct.
8824 (jump_table_cluster::can_be_handled): New.
8825 (case_values_threshold): Moved to header.
8826 (reset_out_edges_aux): Likewise.
8827 (jump_table_cluster::is_beneficial): New.
8828 (bit_test_cluster::can_be_handled): Likewise.
8829 (add_case_node): Remove.
8830 (bit_test_cluster::is_beneficial): New.
8831 (case_bit_test::cmp): New.
8832 (bit_test_cluster::emit): New.
8833 (expand_switch_as_decision_tree_p): Remove.
8834 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
8835 (fix_phi_operands_for_edge): Likewise.
8836 (switch_decision_tree::analyze_switch_statement): New.
8837 (compute_cases_per_edge): Move ...
8838 (switch_decision_tree::compute_cases_per_edge): ... here.
8839 (try_switch_expansion): Likewise.
8840 (switch_decision_tree::try_switch_expansion): Likewise.
8841 (record_phi_operand_mapping): Likewise.
8842 (switch_decision_tree::record_phi_operand_mapping): Likewise.
8843 (emit_case_decision_tree): Likewise.
8844 (switch_decision_tree::emit): Likewise.
8845 (balance_case_nodes): Likewise.
8846 (switch_decision_tree::balance_case_nodes): Likewise.
8847 (dump_case_nodes): Likewise.
8848 (switch_decision_tree::dump_case_nodes): Likewise.
8849 (emit_jump): Likewise.
8850 (switch_decision_tree::emit_jump): Likewise.
8851 (emit_cmp_and_jump_insns): Likewise.
8852 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
8853 (emit_case_nodes): Likewise.
8854 (switch_decision_tree::emit_case_nodes): Likewise.
8855 (conditional_probability): Remove.
8856 * tree-switch-conversion.h (enum cluster_type): New.
8858 (struct cluster): Likewise.
8859 (cluster::cluster): Likewise.
8860 (struct simple_cluster): Likewise.
8861 (simple_cluster::simple_cluster): Likewise.
8862 (struct group_cluster): Likewise.
8863 (struct jump_table_cluster): Likewise.
8864 (struct bit_test_cluster): Likewise.
8865 (struct min_cluster_item): Likewise.
8866 (struct case_tree_node): Likewise.
8867 (case_tree_node::case_tree_node): Likewise.
8868 (jump_table_cluster::case_values_threshold): Likewise.
8869 (struct case_bit_test): Likewise.
8870 (struct switch_decision_tree): Likewise.
8871 (struct switch_conversion): Likewise.
8872 (switch_decision_tree::reset_out_edges_aux): Likewise.
8874 2018-06-20 Martin Liska <mliska@suse.cz>
8876 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
8877 (hoist_edge_and_branch_if_true): Likewise.
8878 (expand_switch_using_bit_tests_p): Likewise.
8879 (struct case_bit_test): Likewise.
8880 (case_bit_test_cmp): Likewise.
8881 (emit_case_bit_tests): Likewise.
8882 (switch_conversion::switch_conversion): New class.
8883 (struct switch_conv_info): Remove old struct.
8884 (collect_switch_conv_info): More to ...
8885 (switch_conversion::collect): ... this.
8886 (check_range): Likewise.
8887 (switch_conversion::check_range): Likewise.
8888 (check_all_empty_except_final): Likewise.
8889 (switch_conversion::check_all_empty_except_final): Likewise.
8890 (check_final_bb): Likewise.
8891 (switch_conversion::check_final_bb): Likewise.
8892 (create_temp_arrays): Likewise.
8893 (switch_conversion::create_temp_arrays): Likewise.
8894 (free_temp_arrays): Likewise.
8895 (gather_default_values): Likewise.
8896 (switch_conversion::gather_default_values): Likewise.
8897 (build_constructors): Likewise.
8898 (switch_conversion::build_constructors): Likewise.
8899 (constructor_contains_same_values_p): Likewise.
8900 (switch_conversion::contains_same_values_p): Likewise.
8901 (array_value_type): Likewise.
8902 (switch_conversion::array_value_type): Likewise.
8903 (build_one_array): Likewise.
8904 (switch_conversion::build_one_array): Likewise.
8905 (build_arrays): Likewise.
8906 (switch_conversion::build_arrays): Likewise.
8907 (gen_def_assigns): Likewise.
8908 (switch_conversion::gen_def_assigns): Likewise.
8909 (prune_bbs): Likewise.
8910 (switch_conversion::prune_bbs): Likewise.
8911 (fix_phi_nodes): Likewise.
8912 (switch_conversion::fix_phi_nodes): Likewise.
8913 (gen_inbound_check): Likewise.
8914 (switch_conversion::gen_inbound_check): Likewise.
8915 (process_switch): Use the newly created class.
8916 (switch_conversion::expand): New.
8917 (switch_conversion::~switch_conversion): New.
8918 * tree-switch-conversion.h: New file.
8920 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8922 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
8923 tree-vect-patterns.c.
8924 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
8925 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
8926 (vect_recog_sad_pattern): Likewise.
8927 (vect_recog_widen_sum_pattern): Likewise.
8928 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
8929 (vect_recog_widen_shift_pattern): Remove the type_in argument.
8930 (vect_recog_rotate_pattern): Likewise.
8931 (vect_recog_mult_pattern): Likewise.
8932 (vect_recog_vector_vector_shift_pattern): Likewise.
8933 (vect_recog_divmod_pattern): Likewise.
8934 (vect_recog_mixed_size_cond_pattern): Likewise.
8935 (vect_recog_bool_pattern): Likewise.
8936 (vect_recog_mask_conversion_pattern): Likewise.
8937 (vect_try_gather_scatter_pattern): Likewise.
8938 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
8939 (vect_recog_over_widening_pattern): Likewise.
8940 (vect_recog_gather_scatter_pattern): Likewise.
8941 (vect_recog_func_ptr): Move from tree-vectorizer.h
8942 (vect_vect_recog_func_ptrs): Move further down the file.
8943 (vect_recog_func): Likewise. Remove the third argument.
8944 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
8945 (vect_pattern_recog_1): Expect the pattern function to do any
8946 necessary target tests. Also expect it to provide a vector type.
8947 Remove the type_in handling.
8949 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8951 * tree-vect-patterns.c (vect_pattern_detected): New function.
8952 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
8953 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
8954 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
8955 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
8956 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
8957 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
8958 (vect_recog_mask_conversion_pattern)
8959 (vect_try_gather_scatter_pattern): Likewise.
8961 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8963 * tree-vect-patterns.c (vect_get_internal_def): New function.
8964 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
8965 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
8966 (search_type_for_mask_1): Use it.
8968 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8970 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
8971 redundant WIDEN_SUM_EXPR handling.
8972 (vect_recog_sad_pattern): Likewise.
8974 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8976 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
8977 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
8978 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
8979 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
8980 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
8982 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8984 * tree-vect-stmts.c (vectorizable_call): Make sure that we
8985 use the stmt_vec_info of the original bb statement for the
8986 new zero assignment, even if the call is part of a pattern.
8988 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8990 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
8991 that the sequence is attached to the original statement rather
8992 than the pattern statement.
8993 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
8994 PATTERN_DEF_SEQ from the original statement rather than
8995 the main pattern statement.
8996 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
8997 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
8998 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
9000 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
9002 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
9003 definition statements before the early exit for statements that aren't
9005 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
9007 (vect_transform_loop): ...here. Process pattern definition
9008 statements without first checking whether the main pattern
9009 statement is live or relevant.
9011 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
9013 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
9014 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
9016 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
9018 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
9019 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
9020 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
9021 (expand_block_compare): Change select_block_compare_mode call.
9022 (expand_strncmp_align_check): Use new functions, fix comment.
9023 (emit_final_str_compare_gpr): New function.
9024 (expand_strn_compare): Refactor and clean up code.
9025 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
9027 2018-06-19 Tony Reix <tony.reix@atos.com>
9028 Damien Bergamini <damien.bergamini@atos.com>
9029 David Edelsohn <dje.gcc@gmail.com>
9031 * collect2.c (static_obj): New variable.
9032 (static_libs): New variable.
9033 (is_in_list): Uncomment declaration.
9034 (main): Track AIX libraries linked statically.
9035 (is_in_list): Uncomment definition.
9036 (scan_prog_file): Don't add AIX shared libraries initializer
9037 to constructor list if linking statically.
9039 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
9041 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
9043 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
9045 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
9047 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
9050 2018-06-19 Martin Liska <mliska@suse.cz>
9052 * config/i386/i386.c (ix86_can_inline_p): Do not use
9053 ipa_fn_summaries::get_create.
9054 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
9056 (devirtualization_time_bonus): Likewise.
9057 (ipcp_propagate_stage): Likewise.
9058 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
9059 (edge_set_predicate): Likewise.
9060 (evaluate_conditions_for_known_args): Likewise.
9061 (evaluate_properties_for_edge): Likewise.
9062 (ipa_call_summary::reset): Tranform to ...
9063 (ipa_call_summary::~ipa_call_summary): ... this.
9064 (ipa_fn_summary::reset): Transform to ...
9065 (ipa_fn_summary::~ipa_fn_summary): ... this.
9066 (ipa_fn_summary_t::remove): Rename to ...
9067 (ipa_fn_summary_t::remove_callees): ... this.
9068 (ipa_fn_summary_t::duplicate): Use placement new
9069 instead of memory copy.
9070 (ipa_call_summary_t::duplicate): Likewise.
9071 (ipa_call_summary_t::remove): Remove.
9072 (dump_ipa_call_summary): Change get_create to get.
9073 (ipa_dump_fn_summary): Dump only when summary exists.
9074 (analyze_function_body): Use symbol_summary::get instead
9076 (compute_fn_summary): Likewise.
9077 (estimate_edge_devirt_benefit): Likewise.
9078 (estimate_edge_size_and_time): Likewise.
9079 (inline_update_callee_summaries): Likewise.
9080 (remap_edge_change_prob): Likewise.
9081 (remap_edge_summaries): Likewise.
9082 (ipa_merge_fn_summary_after_inlining): Likewise.
9083 (write_ipa_call_summary): Likewise.
9084 (ipa_fn_summary_write): Likewise.
9085 (ipa_free_fn_summary): Likewise.
9086 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
9087 (struct ipa_call_summary): Likewise.
9088 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
9090 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
9091 (estimate_size_after_inlining): Likewise.
9092 (estimate_growth): Likewise.
9093 (growth_likely_positive): Likewise.
9094 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
9095 (inline_call): Likewise.
9096 * ipa-inline.c (caller_growth_limits): Likewise.
9097 (can_inline_edge_p): Likewise.
9098 (can_inline_edge_by_limits_p): Likewise.
9099 (compute_uninlined_call_time): Likewise.
9100 (compute_inlined_call_time): Likewise.
9101 (want_inline_small_function_p): Likewise.
9102 (edge_badness): Likewise.
9103 (update_caller_keys): Likewise.
9104 (update_callee_keys): Likewise.
9105 (inline_small_functions): Likewise.
9106 (inline_to_all_callers_1): Likewise.
9107 (dump_overall_stats): Likewise.
9108 (early_inline_small_functions): Likewise.
9109 (early_inliner): Likewise.
9110 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
9111 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
9112 * ipa-pure-const.c (malloc_candidate_p): Likewise.
9113 * ipa-split.c (execute_split_functions): Likewise.
9114 * symbol-summary.h: Likewise.
9115 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
9117 2018-06-19 Richard Biener <rguenther@suse.de>
9119 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
9120 (vectorize_loops): ... here. Fix dbgcnt handling.
9121 (try_vectorize_loop): Wrap try_vectorize_loop_1.
9123 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
9126 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
9127 ieee128 argument takes up only one (vector) register, not two (floating
9130 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
9132 * gimplify.c (gimplify_init_constructor): Really never clear for an
9133 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
9135 2018-06-19 Richard Biener <rguenther@suse.de>
9137 PR tree-optimization/86179
9138 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
9139 after failed recognition.
9141 2018-06-18 Martin Sebor <msebor@redhat.com>
9144 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
9145 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
9146 Handle integer subtraction.
9147 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
9148 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
9150 2018-06-18 David Malcolm <dmalcolm@redhat.com>
9152 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
9153 param from rtx to rtx_insn *.
9154 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
9156 (frv_ifcvt_modify_insn): Likwise.
9157 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
9158 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
9159 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
9160 as_a <rtx_insn *> cast to local "unprotected_region" once
9161 it's been established that it's not NULL or pc_rtx.
9162 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
9163 param "sethi" from rtx to rtx_insn *.
9164 (nds32_group_float_insns): Likewise for param "insn".
9165 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
9167 (vax_output_int_subtract): Likewise.
9168 * config/vax/vax.c (vax_output_int_add): Likewise for param
9170 (vax_output_int_subtract): Likewise.
9171 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
9172 (emit_pattern_after): Likewise for param "after".
9173 (emit_insn_after): Likewise.
9174 (emit_jump_insn_after): Likewise.
9175 (emit_call_insn_after): Likewise.
9176 (emit_debug_insn_after): Likewise.
9177 (emit_pattern_before): Likewise for param "before".
9178 (emit_insn_before): Likewise.
9179 (emit_jump_insn_before): Likewise.
9180 * final.c (get_insn_template): Likewise for param "insn", removing
9182 * output.h (get_insn_template): Likewise for 2nd param.
9183 * rtl.h (emit_insn_before): Likewise.
9184 (emit_jump_insn_before): Likewise.
9185 (emit_debug_insn_before_noloc): Likewise.
9186 (emit_insn_after): Likewise.
9187 (emit_jump_insn_after): Likewise.
9188 (emit_call_insn_after): Likewise.
9189 (emit_debug_insn_after): Likewise.
9190 (set_insn_deleted): Likewise for param.
9192 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
9195 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
9196 floating point modes, so that IFmode is numerically greater than
9197 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
9198 to declare the ordering. This prevents IFmode from being
9199 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
9200 machine. Include rs6000-modes.h to share the fractional values
9201 between genmodes* and the rest of the compiler.
9205 * config/rs6000/rs6000-modes.h: New file.
9206 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
9207 meaning of rs6000_long_double_size so that 126..128 selects an
9208 appropriate 128-bit floating point type.
9209 (rs6000_option_override_internal): Likewise.
9210 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
9211 (TARGET_LONG_DOUBLE_128): Change the meaning of
9212 rs6000_long_double_size so that 126..128 selects an appropriate
9213 128-bit floating point type.
9214 (LONG_DOUBLE_TYPE_SIZE): Update comment.
9215 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
9216 source and destination to match the standard usage.
9217 (truncifkf2): Likewise.
9218 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
9219 ISA 2.07 to use an explicit clobber, instead of passing in a
9221 (copysign<mode>3_soft): Likewise.
9223 2018-06-18 David Malcolm <dmalcolm@redhat.com>
9225 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
9226 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
9227 (vect_slp_analyze_instance_dependence): Likewise.
9228 (vect_enhance_data_refs_alignment): Likewise.
9229 (vect_analyze_data_refs_alignment): Likewise.
9230 (vect_slp_analyze_and_verify_instance_alignment
9231 (vect_analyze_data_ref_accesses): Likewise.
9232 (vect_prune_runtime_alias_test_list): Likewise.
9233 (vect_analyze_data_refs): Likewise.
9234 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
9235 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9236 (vect_analyze_scalar_cycles_1): Likewise.
9237 (vect_get_loop_niters): Likewise.
9238 (vect_analyze_loop_form_1): Likewise.
9239 (vect_update_vf_for_slp): Likewise.
9240 (vect_analyze_loop_operations): Likewise.
9241 (vect_analyze_loop): Likewise.
9242 (vectorizable_induction): Likewise.
9243 (vect_transform_loop): Likewise.
9244 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
9245 * tree-vect-slp.c (vect_analyze_slp): Likewise.
9246 (vect_make_slp_decision): Likewise.
9247 (vect_detect_hybrid_slp): Likewise.
9248 (vect_slp_analyze_operations): Likewise.
9249 (vect_slp_bb): Likewise.
9250 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
9251 (vectorizable_bswap): Likewise.
9252 (vectorizable_call): Likewise.
9253 (vectorizable_simd_clone_call): Likewise.
9254 (vectorizable_conversion): Likewise.
9255 (vectorizable_assignment): Likewise.
9256 (vectorizable_shift): Likewise.
9257 (vectorizable_operation): Likewise.
9258 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
9260 2018-06-18 Martin Sebor <msebor@redhat.com>
9262 PR tree-optimization/81384
9263 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
9264 * builtins.c (expand_builtin_strnlen): New function.
9265 (expand_builtin): Call it.
9266 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
9267 * builtins.def (BUILT_IN_STRNLEN): New.
9268 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
9269 Warn for bounds in excess of maximum object size.
9270 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
9271 single-value ranges. Handle strnlen.
9272 (handle_builtin_strlen): Handle strnlen.
9273 (strlen_check_and_optimize_stmt): Same.
9274 * doc/extend.texi (Other Builtins): Document strnlen.
9276 2018-06-18 Maya Rashish <coypu@sdf.org>
9278 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
9279 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
9280 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
9282 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
9284 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
9286 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9288 * tree.c (escaped_string::escape): Replace cast to char * by
9289 const_cast<char *> (unescaped).
9291 2018-06-18 Nick Clifton <nickc@redhat.com>
9294 * tree.c (escaped_string): New class. Converts an unescaped
9295 string into its escaped equivalent.
9296 (warn_deprecated_use): Use the new class to convert the
9297 deprecation message, if present.
9298 (test_escaped_strings): New self test.
9299 (test_c_tests): Add test_escaped_strings.
9300 * doc/extend.texi (deprecated): Add a note that the
9301 deprecation message is affected by the -fmessage-length
9302 option, and that control characters will be escaped.
9303 (#pragma GCC error): Document this pragma.
9304 (#pragma GCC warning): Likewise.
9305 * doc/invoke.texi (-fmessage-length): Document this option's
9306 effect on the #warning and #error preprocessor directives and
9307 the deprecated attribute.
9309 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
9311 * tree.c (decl_value_expr_lookup): Revert latest change.
9312 (decl_value_expr_insert): Likewise.
9314 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
9316 * gimplify.c (nonlocal_vlas): Delete.
9317 (nonlocal_vla_vars): Likewise.
9318 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
9320 (gimplify_body): Do not create and destroy nonlocal_vlas.
9321 * tree-nested.c: Include diagnostic.h.
9322 (use_pointer_in_frame): Tweak.
9323 (lookup_field_for_decl): Add assertion and declare the transformation.
9324 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
9325 internal error when the reference is in a wrong context. Do not
9326 create a debug decl by default.
9327 (note_nonlocal_block_vlas): Delete.
9328 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
9329 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
9330 create a debug decl by default.
9331 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
9332 call is in a wrong context.
9333 (fixup_vla_decls): New function.
9334 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
9335 debug variables were created.
9336 * tree.c (decl_value_expr_lookup): Add checking assertion.
9337 (decl_value_expr_insert): Likewise.
9339 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
9342 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
9343 * tree-scalar-evolution.c (interpret_expr): Likewise.
9344 (expression_expensive_p): Likewise.
9345 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
9346 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
9347 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
9348 (ssa_defined_by_minus_one_stmt_p): New.
9350 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
9353 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
9354 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
9355 * dojump.c (do_jump): Likewise.
9356 * expr.c (expand_expr_real_2): Check operand type's sign.
9357 * fold-const.c (const_unop): Handle ABSU_EXPR.
9358 (fold_abs_const): Likewise.
9359 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
9360 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
9361 (strip_sign_op_1): Likesise.
9362 * match.pd: Add new pattern to generate ABSU_EXPR.
9363 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
9364 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
9365 * tree-eh.c (operation_could_trap_helper_p): Likewise.
9366 * tree-inline.c (estimate_operator_cost): Likewise.
9367 * tree-pretty-print.c (dump_generic_node): Likewise.
9368 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
9369 * tree.def (ABSU_EXPR): New.
9371 2018-06-16 Jakub Jelinek <jakub@redhat.com>
9374 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
9375 documented as preserved for backward compatibility only.
9376 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
9378 PR rtl-optimization/86108
9379 * bb-reorder.c (create_forwarder_block): Renamed to ...
9380 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
9381 jump from new landing pad to the second part.
9382 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
9385 2018-06-15 Jakub Jelinek <jakub@redhat.com>
9388 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
9389 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
9390 Only call store_expr for halves if the mode is the same.
9393 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
9394 Fix up comment formatting.
9396 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
9398 * typed-splay-tree.h (typed_splay_tree::remove): New function.
9399 (typed_splay_tree::closure,
9400 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
9401 (typed_splay_tree::typed_splay_tree,
9402 typed_splay_tree::operator =): Declared private.
9403 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
9404 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
9405 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
9406 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
9407 typed_splay_tree::splay_tree_splay,
9408 typed_splay_tree::splay_tree_foreach_helper,
9409 typed_splay_tree::splay_tree_insert,
9410 typed_splay_tree::splay_tree_remove,
9411 typed_splay_tree::splay_tree_lookup,
9412 typed_splay_tree::splay_tree_predecessor,
9413 typed_splay_tree::splay_tree_successor,
9414 typed_splay_tree::splay_tree_min,
9415 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
9416 (typed_splay_tree::root, typed_splay_tree::comp,
9417 typed_splay_tree::delete_key,
9418 typed_splay_tree::delete_value): New data members.
9419 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
9420 typed_splay_tree::remove.
9422 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
9424 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
9425 -mginv and -mno-ginv to the assembler.
9426 * config/mips/mips.opt (-mcrc): New option.
9428 * doc/invoke.text (-mcrc): Document.
9431 2018-06-15 Nick Clifton <nickc@redhat.com>
9434 * tree.c (escaped_string): New class. Converts an unescaped
9435 string into its escaped equivalent.
9436 (warn_deprecated_use): Use the new class to convert the
9437 deprecation message, if present.
9438 (test_escaped_strings): New self test.
9439 (test_c_tests): Add test_escaped_strings.
9440 * doc/extend.texi (deprecated): Add a note that the
9441 deprecation message is affected by the -fmessage-length
9442 option, and that control characters will be escaped.
9443 (#pragma GCC error): Document this pragma.
9444 (#pragma GCC warning): Likewise.
9445 * doc/invoke.texi (-fmessage-length): Document this option's
9446 effect on the #warning and #error preprocessor directives and
9447 the deprecated attribute.
9449 2018-06-15 Richard Biener <rguenther@suse.de>
9451 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
9452 here, also noting vector size used.
9453 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
9454 size used in MSG_OPTIMIZED_LOCATIONS dump.
9455 (pass_slp_vectorize::execute): Adjust.
9457 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
9460 * config/arc/arc.c (arc_return_address_register): Fix
9463 2018-06-15 Richard Biener <rguenther@suse.de>
9466 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
9467 leave useless conversion stripping to force_gimple_operand_gsi.
9468 (gimplify_build2): Likewise.
9469 (gimplify_build1): Likewise.
9471 2018-06-15 Richard Biener <rguenther@suse.de>
9474 * tree-cfg.c (move_stmt_op): unshare invariant addresses
9475 before adjusting their block.
9477 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
9479 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
9480 multilibs for *-*-rtems*.
9481 * config/riscv/t-rtems: New file.
9483 2018-06-14 Jakub Jelinek <jakub@redhat.com>
9486 * match.pd ((A +- CST1) +- CST2): Punt if last resort
9487 unsigned_type_for returns NULL.
9490 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
9491 subregs of multi-word pseudos unless the float mode has word size.
9493 2018-06-14 Richard Biener <rguenther@suse.de>
9496 * tree-vect-generic.c (build_word_mode_vector_type): Remove
9497 duplicate and harmful type_hash_canon.
9498 * tree.c (type_hash_canon): Assert we didn't find ourselves.
9500 2018-06-14 Richard Biener <rguenther@suse.de>
9503 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
9506 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
9508 * config/rtems.h (STDINT_LONG32): Define.
9510 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
9511 Prachi Godbole <prachi.godbole@imgtec.com>
9513 * config/mips/mips-cpus.def: Define P6600.
9514 * config/mips/mips-tables.opt: Regenerate.
9515 * config/mips/mips.c (mips_ucbranch_type): New enum.
9516 (mips_rtx_cost_data): Add support for P6600.
9517 (mips_issue_rate): Likewise.
9518 (mips_multipass_dfa_lookahead): Likewise.
9519 (mips_avoid_hazard): Likewise.
9520 (mips_reorg_process_insns): Likewise.
9521 (mips_classify_branch_p6600): New function.
9522 * config/mips/mips.h (TUNE_P6600): New define.
9523 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
9524 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
9525 * config/mips/mips.md: Include p6600.md.
9526 (processor): Add p6600.
9527 * config/mips/p6600.md: New file.
9528 * doc/invoke.texi: Add p6600 to supported architectures.
9530 2018-06-13 Martin Sebor <msebor@redhat.com>
9532 PR tree-optimization/86114
9533 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
9535 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
9537 2018-06-13 Richard Biener <rguenther@suse.de>
9539 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
9540 Properly set vector type of the intermediate stmt.
9541 * tree-vect-stmts.c (vectorizable_operation): The destination
9542 var always has vectype_out type.
9544 2018-06-13 Jeff Law <law@redhat.com>
9546 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
9547 integer 0 for argument to print_rtl_with_bb.
9548 (rl78_reorg): Likewise.
9550 2018-06-13 David Malcolm <dmalcolm@redhat.com>
9552 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
9553 from rtx to rtx_insn *.
9554 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
9556 (add_sched_insns_for_speculation): Likewise for local "target",
9557 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
9558 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
9559 from rtx_insn ** to rtx_code_label **.
9560 (reorg_emit_nops): Likewise.
9561 (c6x_reorg): Likewise for local "call_labels".
9562 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
9564 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
9565 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
9566 the loops over LABEL_REFS.
9567 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
9569 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
9570 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
9571 (split_branches): Strengthen local "olabel" from rtx to
9572 rtx_insn *, adding a safe_as_a cast.
9573 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
9575 (add_insn_after): Likewise for first two params.
9576 (add_insn_before): Likewise.
9577 (remove_insn): Likewise for param.
9578 (emit_pattern_before_noloc): Likewise for second and third params.
9579 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
9580 (emit_call_insn_before_noloc): Likewise.
9581 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
9583 (emit_barrier_before): Likewise.
9584 (emit_label_before): Strengthen "label" param from "rtx" to
9585 "rtx_code_label *". Strengthen "before" param from "rtx" to
9587 (emit_insn_after_1): Strengthen "after" param from "rtx" to
9589 (emit_pattern_after_noloc): Likewise.
9590 (emit_insn_after_noloc): Likewise.
9591 (emit_jump_insn_after_noloc): Likewise.
9592 (emit_call_insn_after_noloc): Likewise.
9593 (emit_debug_insn_after_noloc): Likewise.
9594 (emit_barrier_after): Likewise.
9595 (emit_label_after): Likewise for both params.
9596 (emit_pattern_after_setloc): Likewise for "after" param. Convert
9597 "loc" param from "int" to "location_t".
9598 (emit_insn_after_setloc): Likewise.
9599 (emit_jump_insn_after_setloc): Likewise.
9600 (emit_call_insn_after_setloc): Likewise.
9601 (emit_debug_insn_after_setloc): Likewise.
9602 (emit_pattern_before_setloc): Likewise for "before" param. Convert
9603 "loc" param from "int" to "location_t".
9604 (emit_pattern_before): Convert NULL_RTX to NULL.
9605 (emit_insn_before_setloc): Convert "loc" param from "int" to
9607 (emit_jump_insn_before_setloc): Likewise.
9608 (emit_call_insn_before_setloc): Likewise.
9609 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
9610 rtx_insn *. Convert "loc" param from "int" to "location_t".
9611 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
9612 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
9613 Convert 3rd param from "int" to "location_t".
9614 (emit_barrier_before, emit_barrier_after, next_real_insn):
9615 Strengthen param from rtx to rtx_insn *.
9616 (emit_label_before): Strengthen 1st param from "rtx" to
9617 "rtx_code_label *". Strengthen 2nd param from "rtx" to
9619 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
9620 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
9621 Strengthen 2nd param from "rtx" to "rtx_insn *".
9622 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
9623 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
9624 Likewise. Convert 3rd param from "int" to "location_t".
9625 (emit_label_after): Strengthen 1st param from "rtx" to
9627 (next_real_insn, remove_insn): Strengthen param from "rtx" to
9629 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
9630 from "rtx" to "rtx_insn *".
9632 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
9634 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
9635 bodies streamed in with -Q.
9636 * dumpfile.c (dump_files): Add lto-stream-out dump file.
9637 * dumpfile.h (tree_dump_index): Add lto_stream_out.
9638 * gimple-streamer-out.c: Include gimple-pretty-print.h
9639 (output_bb): Dump stmts streamed.
9640 * lto-section-out.c: Include print-tree.h
9641 (lto_begin_section): Dump sections created.
9642 (lto_output_decl_index): Dump decl encoded.
9643 * lto-streamer-out.c: Include print-tree.h
9644 (create_output_block): Dump output block created.
9645 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
9646 (output_function): Dump function output.
9647 (output_constructor): Dump constructor streamed.
9648 (write_global_stream): Output indexes encoded.
9649 (produce_asm_for_decls): Dump streams encoded.
9650 * lto-streamer.c (streamer_dump_file): New global var.
9651 * lto-streamer.h (streamer_dump_file): Declare.
9652 * passes.c (ipa_write_summaries): Initialize streamer dump.
9653 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
9656 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
9659 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
9660 offsets for register save directives. Emit a second batch of save
9661 directives, if need be, when the function accesses prior frames.
9663 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9665 * config/arc/fpu.md (fmasf4): Force operand to register.
9666 (fnmasf4): Likewise.
9668 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9670 * config/arc/arc-protos.h (arc_pad_return): Remove.
9671 * config/arc/arc.c (machine_function): Remove force_short_suffix
9673 (arc_print_operand): Adjust printing of '&'.
9674 (arc_verify_short): Remove conditional printing of short suffix.
9675 (arc_final_prescan_insn): Remove reference to size_reason.
9676 (pad_return): New function.
9677 (arc_reorg): Call pad_return.
9678 (arc_pad_return): Remove.
9679 (arc_init_machine_status): Remove reference to force_short_suffix.
9680 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
9681 (attr length): When attribute iscompact is true force to 2
9682 regardless; in the case of maybe check if we want to force the
9683 instruction to have 4 bytes length.
9684 (nopv): Change it to generate 4 byte long nop as well.
9685 (blockage): New pattern.
9686 (simple_return): Remove call to arc_pad_return.
9687 (p_return_i): Likewise.
9689 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9691 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
9693 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9695 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
9698 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9700 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
9701 for ARC700 and ARCv2.
9703 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
9706 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
9707 operands[2] instead of operands[1].
9710 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
9712 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
9713 case, check whether the outer register overlaps an unallocatable
9714 register, not just whether it fits the required class.
9716 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
9718 * poly-int.h (can_div_trunc_p): Add new overload in which all values
9720 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
9721 (memrefs_conflict_p): Likewise.
9722 (init_alias_analysis): Likewise.
9723 * cfgexpand.c (expand_debug_expr): Likewise.
9724 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
9725 * cse.c (fold_rtx): Likewise.
9726 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
9727 * expr.c (emit_block_move_hints): Likewise.
9728 (clear_storage_hints, push_block, emit_push_insn): Likewise.
9729 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
9730 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
9731 (emit_group_store): Likewise.
9732 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
9733 to read the PRE/POST_MODIFY increment.
9734 * calls.c (store_one_arg): Use strip_offset.
9735 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
9737 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
9739 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
9740 (simplify_binary_operation_1): Extend CONST_INT handling to
9742 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
9743 than a HOST_WIDE_INT.
9744 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
9746 (adjust_mems, add_stores): Update accodingly.
9747 (vt_canonicalize_addr): Track polynomial offsets.
9748 (emit_note_insn_var_location): Likewise.
9749 (vt_add_function_parameter): Likewise.
9750 (vt_initialize): Likewise.
9752 2018-06-12 Jeff Law <law@redhat.com>
9754 * config.gcc (alpha*-*-freebsd*): Remove.
9755 * config/alpha/freebsd.h: Remove.
9757 2018-06-12 David Malcolm <dmalcolm@redhat.com>
9760 * spellcheck-tree.c (levenshtein_distance): Rename to...
9761 (get_edit_distance): ...this, and update for underlying renaming.
9762 * spellcheck-tree.h (levenshtein_distance): Rename to...
9763 (get_edit_distance): ...this.
9764 * spellcheck.c (levenshtein_distance): Rename to...
9765 (get_edit_distance): ...this. Convert from Levenshtein distance
9766 to Damerau-Levenshtein distance by supporting transpositions of
9767 adjacent characters. Rename "v1" to "v_next" and "v0" to
9769 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
9770 (selftest::test_edit_distance_unit_test_oneway): ...this, and
9771 update for underlying renaming.
9772 (selftest::levenshtein_distance_unit_test): Rename to...
9773 (selftest::test_get_edit_distance_unit): ...this, and update for
9774 underlying renaming.
9775 (selftest::test_find_closest_string): Add example from PR 69968
9776 where transposition helps
9777 (selftest::test_metric_conditions): Update for renaming.
9778 (selftest::test_metric_conditions): Likewise.
9779 (selftest::spellcheck_c_tests): Likewise.
9780 * spellcheck.h (levenshtein_distance): Rename both overloads to...
9781 (get_edit_distance): ...this.
9782 (best_match::consider): Update for renaming.
9784 2018-06-12 Martin Sebor <msebor@redhat.com>
9786 PR tree-optimization/85259
9787 * builtins.c (compute_objsize): Handle constant offsets.
9788 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
9789 true iff a warning has been issued.
9790 * gimple.h (gimple_nonartificial_location): New function.
9791 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
9792 gimple_nonartificial_location and handle -Wno-system-headers.
9793 (handle_builtin_stxncpy): Same.
9795 2018-06-12 Martin Sebor <msebor@redhat.com>
9798 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
9800 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
9802 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
9803 BUILTIN_VEC_XST entries for pointer to double and long long.
9805 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
9808 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
9810 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
9813 2018-06-12 Martin Liska <mliska@suse.cz>
9815 * doc/options.texi: Document IntegerRange.
9817 2018-06-12 Martin Liska <mliska@suse.cz>
9819 * config/i386/i386.opt: Make MPX-related options as Deprecated.
9820 * opt-functions.awk: Handle Deprecated flag.
9821 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
9823 (read_cmdline_option): Report warning for a deprecated option.
9824 * opts.h (struct cl_option): Add new field cl_deprecated.
9825 (CL_ERR_DEPRECATED): New.
9827 2018-06-12 Martin Liska <mliska@suse.cz>
9829 * doc/options.texi: Document Deprecated option flag.
9831 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9833 * config/arc/arc-arch.h (arc_extras): New enum.
9834 (arc_cpu_t):Add field extra.
9835 (arc_cpu_types): Consider the extras.
9836 * config/arc/arc-cpus.def: Add extras info.
9837 * config/arc/arc-opts.h (processor_type): Consider extra field.
9838 * config/arc/arc.c (arc_override_options): Handle extra field.
9840 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9842 * config/arc/arc-arch.h: Update ARC_OPTX macro.
9843 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
9845 * config/arc/arc.c (arc_init): Update pic warning.
9846 (irq_range): Update irq range parsing warnings.
9847 (arc_override_options): Update various warning messages.
9848 (arc_handle_aux_attribute): Likewise.
9850 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
9852 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
9854 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9856 * doc/sourcebuild.texi: Document usage of line number 0 in verify
9857 compiler messages directives.
9859 2018-06-12 Matthew Fortune <mfortune@gmail.com>
9861 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
9862 * config/mips/mips-tables.opt: Regenerate.
9863 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
9865 * doc/invoke.texi: Document -march=i6500.
9867 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
9869 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
9870 (i6400_gpmul): Add cpu_unit.
9871 (i6400_gpdiv): Likewise.
9872 (i6400_msa_add_d): Update reservations.
9873 (i6400_msa_int_add) Likewise.
9874 (i6400_msa_short_logic3) Likewise.
9875 (i6400_msa_short_logic2) Likewise.
9876 (i6400_msa_short_logic) Likewise.
9877 (i6400_msa_move) Likewise.
9878 (i6400_msa_cmp) Likewise.
9879 (i6400_msa_short_float2) Likewise.
9880 (i6400_msa_div_d) Likewise.
9881 (i6400_msa_long_logic1) Likewise.
9882 (i6400_msa_long_logic2) Likewise.
9883 (i6400_msa_mult) Likewise.
9884 (i6400_msa_long_float2) Likewise.
9885 (i6400_msa_long_float4) Likewise.
9886 (i6400_msa_long_float5) Likewise.
9887 (i6400_msa_long_float8) Likewise.
9888 (i6400_fpu_fadd): Include frint type.
9889 (i6400_fpu_store): New define_insn_reservation.
9890 (i6400_fpu_load): Likewise.
9891 (i6400_fpu_move): Likewise.
9892 (i6400_fpu_fcmp): Likewise.
9893 (i6400_fpu_fmadd): Likewise.
9894 (i6400_int_mult): Include imul3nc type and update reservation.
9895 (i6400_int_div): Include idiv3 type and update reservation.
9896 (i6400_int_load): Update to check type not move_type.
9897 (i6400_int_store): Likewise.
9898 (i6400_int_prefetch): Set zero latency.
9900 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
9902 * gcc.c: Document new %@{...} sequence.
9903 (LINK_COMMAND_SPEC): Use it for the -L switches.
9904 (cpp_unique_options): Use it for the -I switches.
9905 (at_file_argbuf): New global variable.
9906 (in_at_file): Likewise.
9907 (alloc_args): Create at_file_argbuf.
9908 (clear_args): Truncate at_file_argbuf.
9909 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
9910 (open_at_file): New function.
9911 (close_at_file): Likewise.
9912 (create_at_file): Delete.
9913 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
9916 (validate_switches_from_spec): Deal with %@{...} sequence.
9917 (validate_switches): Likewise.
9918 (driver::finalize): Call clear_args.
9920 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
9922 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
9924 2018-06-11 Martin Sebor <msebor@redhat.com>
9926 * doc/invoke.texi (-Wall): List -Wc++17-compat.
9927 (Wno-class-memaccess): Add @opindex.
9928 (Wno-templates, Wno-multiple-inheritance): Same.
9929 (Wno-virtual-inheritance, Wno-namespaces): Same.
9930 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
9931 (Wno-format-overflow, Wno-format-truncation): Same.
9932 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
9933 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
9934 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
9935 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
9936 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
9937 (Wno-misspelled-isr): Same.
9939 2018-06-11 Martin Sebor <msebor@redhat.com>
9941 * PR tree-optimization/86083
9942 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
9944 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
9946 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
9948 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
9951 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
9952 on the correct operand.
9953 (*movdi_internal64): Ditto.
9955 2018-06-11 Martin Liska <mliska@suse.cz>
9957 PR tree-optimization/86089
9958 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
9960 2018-06-11 Julia Koval <julia.koval@intel.com>
9962 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
9963 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
9964 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
9966 2018-06-11 Olivier Hainque <hainque@adacore.com>
9968 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
9969 for Ada with strict dwarf2.
9971 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
9974 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
9977 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
9979 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
9981 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
9983 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
9986 2018-06-08 Martin Liska <mliska@suse.cz>
9988 * tree-cfg.h (debug_function): Fix argument type to match
9991 2018-06-08 Martin Liska <mliska@suse.cz>
9993 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
9994 Remove usage of MPX-related (and removed) fields.
9995 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
9997 2018-06-08 David Malcolm <dmalcolm@redhat.com>
9999 * cfg.c (debug): Use TDF_NONE rather than 0.
10000 * cfghooks.c (debug): Likewise.
10001 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
10002 (struct dump_option_value_info): Convert to...
10003 (struct kv_pair): ...this template type.
10004 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
10006 (optinfo_verbosity_options): Likewise.
10007 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
10009 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
10010 than int for "optgroup_flags" param.
10011 (dump_generic_expr_loc): Use dump_flags_t rather than int for
10013 (dump_dec): Likewise.
10014 (dump_finish): Use TDF_NONE rather than 0.
10015 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
10016 rather than int for "optgroup_flags" param. Use TDF_NONE rather
10017 than 0. Update for change to option_ptr.
10018 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
10019 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
10020 0. Update for changes to optinfo_verbosity_options and
10022 (opt_info_switch_p): Convert optgroup_flags from int to
10024 (dump_basic_block): Use dump_flags_t rather than int
10025 for "dump_kind" param.
10026 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
10027 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
10028 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
10029 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
10030 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
10031 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
10032 TDF_NONE): Convert from macros to...
10033 (enum dump_flag): ...this new enum.
10034 (dump_flags_t): Update to use enum.
10035 (operator|, operator&, operator~, operator|=, operator&=):
10036 Implement for dump_flags_t.
10037 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
10038 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
10039 Convert from macros to...
10040 (enum optgroup_flag): ...this new enum.
10041 (optgroup_flags_t): New typedef.
10042 (operator|, operator|=): Implement for optgroup_flags_t.
10043 (struct dump_file_info): Convert field "alt_flags" to
10044 dump_flags_t. Convert field "optgroup_flags" to
10046 (dump_basic_block): Use dump_flags_t rather than int for param.
10047 (dump_generic_expr_loc): Likewise.
10048 (dump_dec): Likewise.
10049 (dump_register): Convert param "optgroup_flags" to
10051 (opt_info_enable_passes): Likewise.
10052 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
10054 * gimple-pretty-print.c (debug): Likewise.
10055 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
10056 (merged_store_group::apply_stores): Likewise.
10057 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
10058 * gimple.c (verify_gimple_pp): Likewise.
10059 * graphite-poly.c (print_pbb_body): Likewise.
10060 * passes.c (pass_manager::register_one_dump_file): Convert
10061 local "optgroup_flags" to optgroup_flags_t.
10062 * print-tree.c (print_node): Use TDF_NONE rather than 0.
10064 (debug_body): Likewise.
10065 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
10066 to optgroup_flags_t.
10067 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
10069 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
10070 (convert_mult_to_fma): Likewise.
10071 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
10072 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
10073 * tree-vect-data-refs.c (dump_lower_bound): Convert param
10074 "dump_kind" to dump_flags_t.
10076 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
10078 * config/rs6000/rs6000.c (min, max): Delete.
10080 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
10082 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
10083 -mabi=spe and -mabi=no-spe.
10085 2018-06-08 Martin Liska <mliska@suse.cz>
10087 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
10088 where we expect an existing summary.
10090 2018-06-08 Martin Liska <mliska@suse.cz>
10092 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
10093 * ipa-inline.h (estimate_edge_growth): Likewise.
10095 2018-06-08 Martin Liska <mliska@suse.cz>
10097 * cgraph.c (function_version_hasher::hash): Use
10098 cgraph_node::get_uid ().
10099 (function_version_hasher::equal):
10100 * cgraph.h (cgraph_node::get_uid): New method.
10101 * ipa-inline.c (update_caller_keys): Use
10102 cgraph_node::get_uid ().
10103 (update_callee_keys): Likewise.
10104 * ipa-utils.c (searchc): Likewise.
10105 (ipa_reduced_postorder): Likewise.
10106 * lto-cgraph.c (input_node): Likewise.
10107 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
10108 * symbol-summary.h (symtab_insertion): Likewise.
10109 (symtab_removal): Likewise.
10110 (symtab_duplication): Likewise.
10111 * tree-pretty-print.c (dump_function_header): Likewise.
10112 * tree-sra.c (convert_callers_for_node): Likewise.
10114 2018-06-08 Martin Liska <mliska@suse.cz>
10116 * cgraph.c (symbol_table::create_edge): Always assign a new
10118 (symbol_table::free_edge): Do not recycle numbers.
10119 * cgraph.h (cgraph_edge::get): New method.
10120 * symbol-summary.h (symtab_removal): Use it.
10121 (symtab_duplication): Likewise.
10122 (call_summary::hashable_uid): Remove.
10124 2018-06-08 Martin Liska <mliska@suse.cz>
10126 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
10127 (initialize_growth_caches): Remove.
10128 (free_growth_caches): Likewise.
10129 (do_estimate_edge_time): Use edge_growth_cache.
10130 (do_estimate_edge_size): Likewise.
10131 (do_estimate_edge_hints): Likewise.
10132 * ipa-inline.c (reset_edge_caches): Likewise.
10133 (recursive_inlining): Likewise.
10134 (inline_small_functions): Likewise.
10135 * ipa-inline.h (initialize_growth_caches): Remove.
10136 (estimate_edge_size): Likewise.
10137 (estimate_edge_time): Likewise.
10138 (estimate_edge_hints): Likewise.
10139 (reset_edge_growth_cache): Likewise.
10140 * symbol-summary.h (call_summary::remove): New method.
10142 2018-06-08 Martin Liska <mliska@suse.cz>
10144 * ipa-cp.c (class edge_clone_summary): New summary.
10145 (grow_edge_clone_vectors): Remove.
10146 (ipcp_edge_duplication_hook): Remove.
10147 (class edge_clone_summary_t): New call_summary class.
10148 (ipcp_edge_removal_hook): Remove.
10149 (edge_clone_summary_t::duplicate): New function.
10150 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
10151 (create_specialized_node): Likewise.
10152 (ipcp_driver): Initialize edge_clone_summaries and do not
10155 2018-06-08 Martin Liska <mliska@suse.cz>
10157 * symbol-summary.h (get): New function.
10158 (call_summary::m_initialize_when_cloning): New class member.
10160 2018-06-08 Martin Liska <mliska@suse.cz>
10162 * cgraph.c (cgraph_node::remove): Do not recycle uid.
10163 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
10164 (symbol_table::allocate_cgraph_symbol): Do not set uid.
10165 * passes.c (uid_hash_t): Record removed_nodes by their uids.
10166 (remove_cgraph_node_from_order): Use the removed_nodes set.
10167 (do_per_function_toporder): Likwise.
10168 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
10169 instead of summary_uid.
10170 (symtab_removal): Likewise.
10171 (symtab_duplication): Likewise.
10173 2018-06-08 Martin Liska <mliska@suse.cz>
10175 * ipa-cp.c (ipcp_store_bits_results): Use
10176 ipcp_transformation_sum.
10177 (ipcp_store_vr_results): Likewise.
10178 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
10180 (ipcp_transformation_initialize): ... this.
10181 (ipa_set_node_agg_value_chain):
10182 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
10183 (write_ipcp_transformation_info): Likewise.
10184 (read_ipcp_transformation_info): Likewise.
10185 (ipcp_update_bits): Likewise.
10186 (ipcp_update_vr): Likewise.
10187 (ipcp_transform_function): Likewise.
10188 * ipa-prop.h: Rename ipcp_transformation_summary to
10189 ipcp_transformation.
10190 (class ipcp_transformation_t): New function summary.
10191 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
10192 (ipa_get_agg_replacements_for_node): Likewise.
10194 2018-06-08 Martin Liska <mliska@suse.cz>
10196 * ipa-pure-const.c (struct funct_state_d): Do it class instead
10198 (class funct_state_summary_t): New function_summary class.
10199 (has_function_state): Remove.
10200 (get_function_state): Likewise.
10201 (set_function_state): Likewise.
10202 (add_new_function): Likewise.
10203 (funct_state_summary_t::insert): New function.
10204 (duplicate_node_data): Remove.
10205 (remove_node_data): Remove.
10206 (funct_state_summary_t::duplicate): New function.
10207 (register_hooks): Create new funct_state_summaries.
10208 (pure_const_generate_summary): Use it.
10209 (pure_const_write_summary): Likewise.
10210 (pure_const_read_summary): Likewise.
10211 (propagate_pure_const): Likewise.
10212 (propagate_nothrow): Likewise.
10213 (dump_malloc_lattice): Likewise.
10214 (propagate_malloc): Likewise.
10215 (execute): Do not register hooks, just remove summary
10217 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
10220 2018-06-08 Martin Liska <mliska@suse.cz>
10222 * ipa-reference.c (remove_node_data): Remove.
10223 (duplicate_node_data): Likewise.
10224 (class ipa_ref_var_info_summary_t): New class.
10225 (class ipa_ref_opt_summary_t): Likewise.
10226 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
10227 (get_reference_optimization_summary): Use
10228 ipa_ref_opt_sum_summaries.
10229 (set_reference_vars_info): Remove.
10230 (set_reference_optimization_summary): Likewise.
10231 (ipa_init): Create summaries.
10232 (init_function_info): Use function summary.
10233 (ipa_ref_opt_summary_t::duplicate): New function.
10234 (ipa_ref_opt_summary_t::remove): New function.
10235 (get_read_write_all_from_node): Fix GNU coding style.
10236 (propagate): Use function summary.
10237 (write_node_summary_p): Fix GNU coding style.
10238 (stream_out_bitmap): Likewise.
10239 (ipa_reference_read_optimization_summary): Use function summary.
10240 (ipa_reference_c_finalize): Do not release hooks.
10242 2018-06-08 Martin Liska <mliska@suse.cz>
10244 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
10245 (analyze_function_body): Extract multiple calls of get_create.
10246 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
10247 * ipa-inline.c (recursive_inlining): Use ::get method.
10248 * ipa-inline.h (estimate_edge_growth): Likewise.
10250 2018-06-08 Martin Liska <mliska@suse.cz>
10252 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
10254 (hsa_function_summary::hsa_function_summary): Use the new enum
10256 (hsa_gpu_implementation_p): Use hsa_summaries::get.
10257 * hsa-gen.c (hsa_get_host_function): Likewise.
10258 (get_brig_function_name): Likewise.
10259 * ipa-hsa.c (process_hsa_functions): Likewise.
10260 (ipa_hsa_write_summary): Likewise.
10261 * symbol-summary.h (symtab_duplication): Use ::get function/
10262 (get): New function.
10264 2018-06-08 Martin Liska <mliska@suse.cz>
10266 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
10268 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
10269 (hsa_register_kernel): Likewise.
10270 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
10271 * hsa-gen.c (hsa_get_host_function): Likewise.
10272 (get_brig_function_name): Likewise.
10273 (generate_hsa): Likewise.
10274 (pass_gen_hsail::execute): Likewise.
10275 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
10276 (devirtualization_time_bonus): Likewise.
10277 (ipcp_propagate_stage): Likewise.
10278 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
10279 (edge_set_predicate): Likewise.
10280 (evaluate_conditions_for_known_args): Likewise.
10281 (evaluate_properties_for_edge): Likewise.
10282 (ipa_fn_summary::reset): Likewise.
10283 (ipa_fn_summary_t::duplicate): Likewise.
10284 (dump_ipa_call_summary): Likewise.
10285 (ipa_dump_fn_summary): Likewise.
10286 (analyze_function_body): Likewise.
10287 (compute_fn_summary): Likewise.
10288 (estimate_edge_devirt_benefit): Likewise.
10289 (estimate_edge_size_and_time): Likewise.
10290 (estimate_calls_size_and_time): Likewise.
10291 (estimate_node_size_and_time): Likewise.
10292 (inline_update_callee_summaries): Likewise.
10293 (remap_edge_change_prob): Likewise.
10294 (remap_edge_summaries): Likewise.
10295 (ipa_merge_fn_summary_after_inlining): Likewise.
10296 (ipa_update_overall_fn_summary): Likewise.
10297 (read_ipa_call_summary): Likewise.
10298 (inline_read_section): Likewise.
10299 (write_ipa_call_summary): Likewise.
10300 (ipa_fn_summary_write): Likewise.
10301 (ipa_free_fn_summary): Likewise.
10302 * ipa-hsa.c (process_hsa_functions): Likewise.
10303 (ipa_hsa_write_summary): Likewise.
10304 (ipa_hsa_read_section): Likewise.
10305 * ipa-icf.c (sem_function::merge): Likewise.
10306 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
10307 (do_estimate_edge_time): Likewise.
10308 (estimate_size_after_inlining): Likewise.
10309 (estimate_growth): Likewise.
10310 (growth_likely_positive): Likewise.
10311 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
10312 (inline_call): Likewise.
10313 * ipa-inline.c (caller_growth_limits): Likewise.
10314 (can_inline_edge_p): Likewise.
10315 (can_inline_edge_by_limits_p): Likewise.
10316 (compute_uninlined_call_time): Likewise.
10317 (compute_inlined_call_time): Likewise.
10318 (want_inline_small_function_p): Likewise.
10319 (edge_badness): Likewise.
10320 (update_caller_keys): Likewise.
10321 (update_callee_keys): Likewise.
10322 (recursive_inlining): Likewise.
10323 (inline_small_functions): Likewise.
10324 (inline_to_all_callers_1): Likewise.
10325 (dump_overall_stats): Likewise.
10326 (early_inline_small_functions): Likewise.
10327 (early_inliner): Likewise.
10328 * ipa-inline.h (estimate_edge_growth): Likewise.
10329 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
10330 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
10331 * ipa-prop.h (IPA_NODE_REF): Likewise.
10332 (IPA_EDGE_REF): Likewise.
10333 * ipa-pure-const.c (malloc_candidate_p): Likewise.
10334 (propagate_malloc): Likewise.
10335 * ipa-split.c (execute_split_functions): Likewise.
10336 * symbol-summary.h: Rename get to get_create.
10338 (get_create): Likewise.
10339 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
10341 2018-06-08 Martin Liska <mliska@suse.cz>
10343 * symbol-summary.h (release): Move definition out of class
10345 (symtab_removal): Likewise.
10346 (symtab_duplication): Likewise.
10348 2018-06-08 Martin Liska <mliska@suse.cz>
10350 * symbol-summary.h (function_summary): Move constructor
10351 implementation out of class declaration.
10352 (release): Likewise.
10353 (symtab_insertion): Likewise.
10354 (symtab_removal): Likewise.
10355 (symtab_duplication): Likewise.
10358 2018-06-08 Martin Liska <mliska@suse.cz>
10360 * Makefile.in: Remove support for MPX (macros, related functions,
10361 fields in cgraph_node, ...).
10362 * builtin-types.def (BT_BND): Likewise.
10363 (BT_FN_BND_CONST_PTR): Likewise.
10364 (BT_FN_CONST_PTR_BND): Likewise.
10365 (BT_FN_VOID_PTR_BND): Likewise.
10366 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
10367 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
10368 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
10369 (expand_builtin_mempcpy_with_bounds): Likewise.
10370 (expand_builtin_memset_with_bounds): Likewise.
10371 (expand_builtin_memset_args): Likewise.
10372 (std_expand_builtin_va_start): Likewise.
10373 (expand_builtin): Likewise.
10374 (expand_builtin_with_bounds): Likewise.
10375 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
10376 (DEF_LIB_BUILTIN_CHKP): Likewise.
10377 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
10378 (DEF_CHKP_BUILTIN): Likewise.
10379 (BUILT_IN_MEMCPY): Likewise.
10380 (BUILT_IN_MEMMOVE): Likewise.
10381 (BUILT_IN_MEMPCPY): Likewise.
10382 (BUILT_IN_MEMSET): Likewise.
10383 (BUILT_IN_STPCPY): Likewise.
10384 (BUILT_IN_STRCAT): Likewise.
10385 (BUILT_IN_STRCHR): Likewise.
10386 (BUILT_IN_STRCPY): Likewise.
10387 (BUILT_IN_STRLEN): Likewise.
10388 (BUILT_IN_MEMCPY_CHK): Likewise.
10389 (BUILT_IN_MEMMOVE_CHK): Likewise.
10390 (BUILT_IN_MEMPCPY_CHK): Likewise.
10391 (BUILT_IN_MEMSET_CHK): Likewise.
10392 (BUILT_IN_STPCPY_CHK): Likewise.
10393 (BUILT_IN_STRCAT_CHK): Likewise.
10394 (BUILT_IN_STRCPY_CHK): Likewise.
10395 * calls.c (store_bounds): Likewise.
10396 (emit_call_1): Likewise.
10397 (special_function_p): Likewise.
10398 (maybe_warn_nonstring_arg): Likewise.
10399 (initialize_argument_information): Likewise.
10400 (finalize_must_preallocate): Likewise.
10401 (compute_argument_addresses): Likewise.
10402 (expand_call): Likewise.
10403 * cfgexpand.c (expand_call_stmt): Likewise.
10404 (expand_return): Likewise.
10405 (expand_gimple_stmt_1): Likewise.
10406 (pass_expand::execute): Likewise.
10407 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
10408 (cgraph_node::remove): Likewise.
10409 (cgraph_node::dump): Likewise.
10410 (cgraph_node::verify_node): Likewise.
10411 * cgraph.h (chkp_function_instrumented_p): Likewise.
10412 (symtab_node::get_alias_target): Likewise.
10413 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
10414 (cgraph_local_p): Likewise.
10415 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
10416 (cgraph_edge::rebuild_references): Likewise.
10417 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
10418 (walk_polymorphic_call_targets): Likewise.
10419 (cgraph_node::expand_thunk): Likewise.
10420 (symbol_table::output_weakrefs): Likewise.
10421 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
10422 (ix86_handle_option): Likewise.
10423 * config/i386/constraints.md: Likewise.
10424 * config/i386/i386-builtin-types.def (BND): Likewise.
10428 * config/i386/i386-builtin.def (BDESC_END): Likewise.
10429 (BDESC_FIRST): Likewise.
10431 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
10432 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
10433 * config/i386/i386.c (enum reg_class): Likewise.
10434 (ix86_target_string): Likewise.
10435 (ix86_option_override_internal): Likewise.
10436 (ix86_conditional_register_usage): Likewise.
10437 (ix86_valid_target_attribute_inner_p): Likewise.
10438 (ix86_set_indirect_branch_type): Likewise.
10439 (ix86_set_current_function): Likewise.
10440 (ix86_function_arg_regno_p): Likewise.
10441 (init_cumulative_args): Likewise.
10442 (ix86_function_arg_advance): Likewise.
10443 (ix86_function_arg): Likewise.
10444 (ix86_pass_by_reference): Likewise.
10445 (ix86_function_value_regno_p): Likewise.
10446 (ix86_function_value_1): Likewise.
10447 (ix86_function_value_bounds): Likewise.
10448 (ix86_return_in_memory): Likewise.
10449 (ix86_setup_incoming_vararg_bounds): Likewise.
10450 (ix86_va_start): Likewise.
10451 (indirect_thunk_need_prefix): Likewise.
10452 (print_reg): Likewise.
10453 (ix86_print_operand): Likewise.
10454 (ix86_expand_call): Likewise.
10455 (ix86_output_function_return): Likewise.
10456 (reg_encoded_number): Likewise.
10457 (BDESC_VERIFYS): Likewise.
10458 (ix86_init_mpx_builtins): Likewise.
10459 (ix86_init_builtins): Likewise.
10460 (ix86_emit_cmove): Likewise.
10461 (ix86_emit_move_max): Likewise.
10462 (ix86_expand_builtin): Likewise.
10463 (ix86_builtin_mpx_function): Likewise.
10464 (ix86_get_arg_address_for_bt): Likewise.
10465 (ix86_load_bounds): Likewise.
10466 (ix86_store_bounds): Likewise.
10467 (ix86_load_returned_bounds): Likewise.
10468 (ix86_store_returned_bounds): Likewise.
10469 (ix86_class_likely_spilled_p): Likewise.
10470 (ix86_hard_regno_mode_ok): Likewise.
10471 (x86_order_regs_for_local_alloc): Likewise.
10472 (ix86_mitigate_rop): Likewise.
10473 (ix86_bnd_prefixed_insn_p): Likewise.
10474 (ix86_mpx_bound_mode): Likewise.
10475 (ix86_make_bounds_constant): Likewise.
10476 (ix86_initialize_bounds): Likewise.
10477 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
10478 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
10479 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
10480 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
10481 (TARGET_CHKP_BOUND_MODE): Likewise.
10482 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
10483 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
10484 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
10485 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
10486 * config/i386/i386.h (TARGET_MPX): Likewise.
10487 (TARGET_MPX_P): Likewise.
10488 (VALID_BND_REG_MODE): Likewise.
10489 (FIRST_BND_REG): Likewise.
10490 (LAST_BND_REG): Likewise.
10491 (enum reg_class): Likewise.
10492 (BND_REG_P): Likewise.
10493 (BND_REGNO_P): Likewise.
10494 (BNDmode): Likewise.
10495 (ADJUST_INSN_LENGTH): Likewise.
10496 * config/i386/i386.md: Likewise.
10497 * config/i386/i386.opt: Likewise.
10498 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
10499 (defined): Likewise.
10500 (LINK_MPX): Likewise.
10501 (MPX_SPEC): Likewise.
10502 (LIBMPX_SPEC): Likewise.
10503 (LIBMPXWRAPPERS_SPEC): Likewise.
10504 (CHKP_SPEC): Likewise.
10505 * config/i386/predicates.md: Likewise.
10506 * dbxout.c (dbxout_type): Likewise.
10507 * doc/extend.texi: Likewise.
10508 * doc/invoke.texi: Likewise.
10509 * doc/md.texi: Likewise.
10510 * doc/tm.texi: Likewise.
10511 * doc/tm.texi.in: Likewise.
10512 * dwarf2out.c (is_base_type): Likewise.
10513 (gen_formal_types_die): Likewise.
10514 (gen_subprogram_die): Likewise.
10515 (gen_type_die_with_usage): Likewise.
10516 (gen_decl_die): Likewise.
10517 (dwarf2out_late_global_decl): Likewise.
10518 * expr.c (expand_assignment): Likewise.
10519 (emit_storent_insn): Likewise.
10520 (store_expr_with_bounds): Likewise.
10521 (store_expr): Likewise.
10522 (expand_expr_real_1): Likewise.
10523 * expr.h (store_expr_with_bounds): Likewise.
10524 * function.c (use_register_for_decl): Likewise.
10525 (struct bounds_parm_data): Likewise.
10526 (assign_parms_augmented_arg_list): Likewise.
10527 (assign_parm_find_entry_rtl): Likewise.
10528 (assign_parm_is_stack_parm): Likewise.
10529 (assign_parm_load_bounds): Likewise.
10530 (assign_bounds): Likewise.
10531 (assign_parms): Likewise.
10532 (expand_function_start): Likewise.
10533 * gcc.c (CHKP_SPEC): Likewise.
10534 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
10535 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
10536 (wrestrict_dom_walker::check_call): Likewise.
10537 * gimple.c (gimple_build_call_from_tree): Likewise.
10538 * gimple.h (enum gf_mask): Likewise.
10539 (gimple_call_with_bounds_p): Likewise.
10540 (gimple_call_set_with_bounds): Likewise.
10541 * gimplify.c (gimplify_init_constructor): Likewise.
10542 * ipa-cp.c (initialize_node_lattices): Likewise.
10543 (propagate_constants_across_call): Likewise.
10544 (find_more_scalar_values_for_callers_subset): Likewise.
10545 * ipa-hsa.c (process_hsa_functions): Likewise.
10546 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
10547 * ipa-icf.c (sem_function::merge): Likewise.
10548 * ipa-inline.c (early_inliner): Likewise.
10549 * ipa-pure-const.c (warn_function_noreturn): Likewise.
10550 (warn_function_cold): Likewise.
10551 (propagate_pure_const): Likewise.
10552 * ipa-ref.h (enum GTY): Likewise.
10553 * ipa-split.c (find_retbnd): Likewise.
10554 (consider_split): Likewise.
10555 (split_function): Likewise.
10556 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
10557 * ipa.c (walk_polymorphic_call_targets): Likewise.
10558 (symbol_table::remove_unreachable_nodes): Likewise.
10559 (process_references): Likewise.
10560 (cgraph_build_static_cdtor_1): Likewise.
10561 * lto-cgraph.c (lto_output_node): Likewise.
10562 (output_refs): Likewise.
10563 (compute_ltrans_boundary): Likewise.
10564 (input_overwrite_node): Likewise.
10565 (input_node): Likewise.
10566 (input_cgraph_1): Likewise.
10567 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
10568 * passes.c (pass_manager::execute_early_local_passes): Likewise.
10569 (class pass_chkp_instrumentation_passes): Likewise.
10570 (make_pass_chkp_instrumentation_passes): Likewise.
10571 * passes.def: Likewise.
10572 * rtl.h (struct GTY): Likewise.
10573 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
10574 * stor-layout.c (layout_type): Likewise.
10575 * symtab.c: Likewise.
10576 * target.def: Likewise.
10577 * targhooks.c (default_chkp_bound_type): Likewise.
10578 (default_chkp_bound_mode): Likewise.
10579 (default_builtin_chkp_function): Likewise.
10580 (default_chkp_function_value_bounds): Likewise.
10581 (default_chkp_make_bounds_constant): Likewise.
10582 (default_chkp_initialize_bounds): Likewise.
10583 * targhooks.h (default_chkp_bound_type): Likewise.
10584 (default_chkp_bound_mode): Likewise.
10585 (default_builtin_chkp_function): Likewise.
10586 (default_chkp_function_value_bounds): Likewise.
10587 (default_chkp_make_bounds_constant): Likewise.
10588 (default_chkp_initialize_bounds): Likewise.
10589 * toplev.c (compile_file): Likewise.
10590 (process_options): Likewise.
10591 * tree-core.h (DEF_BUILTIN): Likewise.
10592 (DEF_BUILTIN_CHKP): Likewise.
10593 * tree-inline.c (declare_return_variable): Likewise.
10594 (remap_gimple_stmt): Likewise.
10595 (copy_bb): Likewise.
10596 (initialize_inlined_parameters): Likewise.
10597 (expand_call_inline): Likewise.
10598 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
10599 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
10600 (make_pass_ipa_chkp_produce_thunks): Likewise.
10601 (make_pass_chkp): Likewise.
10602 (make_pass_chkp_opt): Likewise.
10603 (make_pass_chkp_instrumentation_passes): Likewise.
10604 * tree-pretty-print.c (dump_generic_node): Likewise.
10605 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
10606 * tree-ssa-dce.c (propagate_necessity): Likewise.
10607 (eliminate_unnecessary_stmts): Likewise.
10608 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
10609 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
10610 * tree-ssa-sccvn.h: Likewise.
10611 * tree-ssa-strlen.c (get_string_length): Likewise.
10612 (valid_builtin_call): Likewise.
10613 (adjust_last_stmt): Likewise.
10614 (handle_builtin_strchr): Likewise.
10615 (handle_builtin_strcpy): Likewise.
10616 (handle_builtin_stxncpy): Likewise.
10617 (handle_builtin_memcpy): Likewise.
10618 (handle_builtin_strcat): Likewise.
10619 (strlen_check_and_optimize_stmt): Likewise.
10620 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
10621 * tree-streamer-in.c: Likewise.
10622 * tree-streamer.c (record_common_node): Likewise.
10623 * tree.c (tree_code_size): Likewise.
10624 (wide_int_to_tree_1): Likewise.
10625 (type_contains_placeholder_1): Likewise.
10626 (build_common_tree_nodes): Likewise.
10627 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
10628 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
10629 (POINTER_BOUNDS_P): Likewise.
10630 (BOUNDED_TYPE_P): Likewise.
10631 (BOUNDED_P): Likewise.
10632 (CALL_WITH_BOUNDS_P): Likewise.
10633 (pointer_bounds_type_node): Likewise.
10634 * value-prof.c (gimple_ic): Likewise.
10635 * var-tracking.c (vt_add_function_parameters): Likewise.
10636 * varasm.c (make_decl_rtl): Likewise.
10637 (assemble_start_function): Likewise.
10638 (output_constant): Likewise.
10639 (maybe_assemble_visibility): Likewise.
10640 * varpool.c (ctor_for_folding): Likewise.
10641 * chkp-builtins.def: Remove.
10642 * ipa-chkp.c: Remove.
10643 * ipa-chkp.h: Remove.
10644 * rtl-chkp.c: Remove.
10645 * rtl-chkp.h: Remove.
10646 * tree-chkp-opt.c: Remove.
10647 * tree-chkp.c: Remove.
10648 * tree-chkp.h: Remove.
10650 2018-06-07 Carl Love <cel@us.ibm.com>
10652 * config/rs6000/vsx.md (vextract_fp_from_shorth,
10653 vextract_fp_from_shortl): Add BE support.
10655 2018-06-07 Paul Koning <ni1d@arrl.net>
10657 * compare-elim.c (try_merge_compare): Don't merge compare if
10658 address contains a side effect.
10659 (try_eliminate_compare): Likewise.
10661 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
10663 * config.gcc: Support "tremont".
10664 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
10665 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10667 * config/i386/i386.c (m_TREMONT): Define.
10668 (processor_target_table): Add "tremont".
10669 (PTA_TREMONT): Define.
10670 (ix86_lea_outperforms): Add TARGET_TREMONT.
10671 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
10672 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
10673 and M_INTEL_GOLDMONT_PLUS.
10674 (fold_builtin_cpu): Add "tremont".
10675 (ix86_add_stmt_cost): Add TARGET_TREMONT.
10676 (ix86_option_override_internal): Add "tremont".
10677 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
10678 (processor_type): Add PROCESSOR_TREMONT.
10679 * config/i386/x86-tune.def: Add m_TREMONT.
10680 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
10682 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10684 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
10685 symbol defined for msp430i* devices to be lower case.
10687 2018-06-07 Richard Biener <rguenther@suse.de>
10689 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
10690 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
10691 Properly wrap signed arithmetic if overflow wraps.
10693 2018-06-07 Jakub Jelinek <jakub@redhat.com>
10695 PR tree-optimization/69615
10696 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
10697 of a cast from a same precision integral SSA_NAME in a bb dominated
10698 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
10699 cast to utype if rhs2 has already a compatible type.
10701 2018-06-07 Richard Biener <rguenther@suse.de>
10703 PR tree-optimization/85935
10704 * graphite-scop-detection.c (find_params_in_bb): Analyze
10705 condition operands with respect to the correct loop. Assert
10706 the analysis doesn't fail.
10708 2018-06-04 Carl Love <cel@us.ibm.com>
10710 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
10711 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
10712 as it is slightly cheaper.
10713 (first_match_or_eos_index_<mode>):
10714 Calculate index using natural element order.
10715 (first_match_index_<mode>):
10716 Calculate index using natural element order.
10717 (first_match_or_eos_index_<mode>):
10718 Calculate index using natural order.
10719 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
10720 for BE and LE modes.
10721 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
10722 P9V_BUILTIN_VCLZLSBB_V16QI.
10723 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
10726 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
10728 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
10729 indentation and line wrap for many prototypes. Add missing
10730 @smallexample directives around block of prototypes for vec_xl and
10733 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
10735 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
10736 track if we pass or return IEEE 128-bit floating point.
10737 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
10738 C++ mangling that is compatible with GCC 8.1.
10739 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
10740 (init_cumulative_args): Note if we pass or return IEEE 128-bit
10741 floating point types.
10742 (rs6000_function_arg_advance_1): Likewise.
10743 (rs6000_mangle_type): Optionally generate mangled names that match
10744 what GCC 8.1 generated for IEEE 128-bit floating point types.
10745 (rs6000_globalize_decl_name): If we have an external function that
10746 passes or returns IEEE 128-bit types, generate a weak reference
10747 from the mangled name used in GCC 8.1 to the current mangled
10749 (rs6000_init_builtins): Make __ibm128 use the long double type if
10750 long double is IBM extended double. Make __float128 use the long
10751 double type if long double is IEEE 128-bit.
10754 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
10755 macro for __ibm128 built-in functions.
10756 (PACK_IF): Add __ibm128 pack/unpack functions.
10757 (UNPACK_IF): Likewise.
10758 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
10759 enable long double built-in functions if long double is IEEE
10760 128-bit floating point.
10761 (rs6000_invalid_builtin): Update long double built-in function
10763 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
10764 functions, adjust the built-in function to use the long double
10765 built-in function if __ibm128 and long double are the same type.
10766 * doc/extend.texi (PowerPC builtins): Update documention for
10767 __builtin_{,un}pack_longdouble. Add documentation for
10768 __builtin_{,un}pack_ibm128.
10770 2018-06-06 Jim Wilson <jimw@sifive.com>
10772 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
10773 (struct machine_function): New field interrupt_mode.
10774 (riscv_handle_type_attribute): New function. Add forward declaration.
10775 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
10776 (riscv_expand_epilogue): Check interrupt_mode field.
10777 (riscv_set_current_function): Check interrupt attribute args and
10778 set interrupt_mode field.
10779 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
10780 (riscv_sret, riscv_uret): New.
10781 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
10782 new arguments to interrupt attribute.
10784 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
10787 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
10788 Don't handle -mcpu=power8 if -mpower9-vector is also used.
10790 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
10792 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
10793 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
10794 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
10795 several redundant entries.
10797 2018-06-06 David Malcolm <dmalcolm@redhat.com>
10799 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
10800 type from "rtx" to "rtx_insn *".
10801 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
10802 for local "call_insn", removing cast.
10803 (ix86_expand_call): Likewise, introducing a "call_insn" local.
10805 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
10807 PR tree-optimization/86066
10808 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
10809 for BIT_INSERT_EXPR stores.
10811 2018-06-06 Richard Biener <rguenther@suse.de>
10813 PR tree-optimization/86062
10814 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
10815 component refs ontop
10816 of to be offsetted base.
10818 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10820 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
10821 to be static and remove check on interrupt attribute name.
10823 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
10825 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
10826 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
10828 2018-06-05 Steve Ellcey <sellcey@cavium.com>
10831 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
10833 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
10834 Remove second argument, change how error is called.
10835 (aarch64_layout_arg): Remove second argument from
10836 aarch64_err_no_fpadvsimd call.
10837 (aarch64_init_cumulative_args): Ditto.
10838 (aarch64_gimplify_va_arg_expr): Ditto.
10839 * config/aarch64/aarch64.md (mov<mode>): Ditto.
10841 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
10843 * config/i386/i386.md (simple_return_indirect_internal): New expander.
10844 (*simple_return_indirect_internal<mode>): Rename from
10845 simple_return_indirect_internal. Use W mode iterator.
10846 (rstorssp): New expander.
10847 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
10848 (clrssbsy): New expander.
10849 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
10851 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
10853 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
10855 (cmse_check_pointed_object): Likewise.
10857 2018-06-05 Martin Liska <mliska@suse.cz>
10859 PR gcov-profile/47618
10860 * doc/invoke.texi: Document how -fprofile-dir format
10863 2018-06-05 Richard Biener <rguenther@suse.de>
10865 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
10866 removal pretend DOM info isn't available so we do not update
10867 it and only remove edges, not dominated blocks. Actually free
10868 DOM info in case we removed something. Remove unreachable blocks.
10869 (mfb_keep_latches): Work with either DOM info or marked backedges.
10870 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
10871 first. Mark backedges if DOM info isn't available.
10872 (Re-)compute DOM info after cleanup_control_flow_pre.
10874 2018-06-05 Richard Biener <rguenther@suse.de>
10876 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
10877 (locus_discrim_hasher::hash): Adjust.
10878 (locus_discrim_hasher::equal): Likewise.
10879 (next_discriminator_for_locus): Work on line directly.
10880 (same_line_p): Pass in expanded locus1 as well.
10881 (assign_discriminators): Avoid redundant location expansions.
10883 2018-06-05 Richard Biener <rguenther@suse.de>
10885 PR tree-optimization/86046
10886 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
10887 if required after clearing TREE_ADDRESSABLE.
10889 2018-06-05 Richard Biener <rguenther@suse.de>
10891 PR tree-optimization/86047
10892 * tree-ssa-loop.c (for_each_index): Glob handling of all
10893 decls and constants and really handle all of them.
10895 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10898 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
10899 qualifier_void_pointer and qualifier_const_void_pointer.
10900 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
10901 (arm_init_builtins): Handle the above.
10902 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
10903 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
10904 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
10907 2018-06-05 Martin Liska <mliska@suse.cz>
10909 * auto-profile.c (read_autofdo_file): Do not use
10910 gcov_ctr_summary struct.
10911 (afdo_callsite_hot_enough_for_early_inline): Likewise.
10912 * coverage.c (struct counts_entry): Likewise.
10913 (read_counts_file): Read just single summary entry.
10914 (get_coverage_counts): Use gcov_summary struct.
10915 * coverage.h (get_coverage_counts): Likewise.
10916 * gcov-dump.c (dump_working_sets): Likewise.
10917 (tag_summary): Dump just single summary.
10918 * gcov-io.c (gcov_write_summary): Write just histogram
10920 (gcov_read_summary): Read just single summary.
10921 (compute_working_sets): Use gcov_summary struct.
10922 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
10923 of GCOV_COUNTERS_SUMMABLE.
10924 (GCOV_COUNTERS_SUMMABLE): Remove.
10925 (GCOV_FIRST_VALUE_COUNTER): Replace with
10926 GCOV_COUNTER_V_INTERVAL.
10927 (struct gcov_ctr_summary): Remove.
10928 (struct gcov_summary): Directly use fields of former
10930 (compute_working_sets): Use gcov_summary struct.
10931 * gcov.c (read_count_file): Do not use ctrs fields.
10932 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
10934 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
10936 * profile.c: Likewise.
10937 * profile.h: Likewise.
10939 2018-06-05 Martin Liska <mliska@suse.cz>
10941 PR gcov-profile/84846
10942 * gcov.c (output_lines): Print working directory only
10943 in intermediate format.
10945 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
10947 * config/s390/s390-builtin-types.def: Add void function type.
10948 * config/s390/s390-builtins.def: Use the function type for the
10951 2018-06-04 Jim Wilson <jimw@sifive.com>
10953 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
10955 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
10956 and maybe_eh_return. Change regno to unsigned int. Use new args to
10957 handle EH_RETURN_DATA_REGNO registers properly.
10958 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
10959 (riscv_expand_epilogue): Update comment. Change argument name and
10960 type. Update code to use new name and type. Pass new args to
10961 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
10963 * config/riscv/riscv.md (NORMAL_RETURN): New.
10964 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
10965 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
10966 (eh_return): Call gen_eh_return_internal and emit barrier.
10967 (eh_return_internal): Call riscv_expand_epilogue.
10969 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
10971 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
10972 bit_insertion field and declare can_be_merged_into method.
10973 (merged_store_group::can_be_merged_into): New method.
10974 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
10975 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
10976 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
10978 2018-06-04 Richard Biener <rguenther@suse.de>
10980 PR tree-optimization/85955
10981 * builtins.c (fold_builtin_sincos): Convert pointers to
10982 destination to appropriate type before dereferencing.
10984 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
10986 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
10988 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
10990 * expr.c (expand_expr_real_1): Force the operand into memory if
10991 its TYPE_MODE is BLKmode and if there is no integer mode for
10992 the number of bits being extracted.
10994 2018-06-04 Jakub Jelinek <jakub@redhat.com>
10998 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
10999 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
11001 2018-06-04 Richard Biener <rguenther@suse.de>
11003 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
11004 (cleanup_tree_cfg_noloop): ... single caller. Do
11005 start_recording_case_labels later.
11007 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
11009 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
11010 to _IMMINTRIN_H_INCLUDED.
11011 * config/i386/pconfigintrin.h: Ditto.
11012 * config/i386/waitpkgintrin.h: Ditto.
11013 * config/i386/immintrin.h: Add includes for sgxintrin.h,
11014 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
11015 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
11016 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
11017 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
11018 waitpkgintrin.h and cldemoteintrin.h.
11020 2018-06-04 Richard Biener <rguenther@suse.de>
11022 PR tree-optimization/86038
11023 * tracer.c (find_best_successor): Check probability for
11024 being initialized, bail out if not.
11026 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
11029 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
11030 of bits to ignore when comparing architectures.
11032 2018-06-04 Jakub Jelinek <jakub@redhat.com>
11034 PR tree-optimization/69615
11035 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
11036 maximum or minimum of the type, try to merge it also as if
11037 range1 is + [-, x - 1] or + [x + 1, -].
11040 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
11042 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
11044 PR tree-optimization/86034
11045 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
11046 the unsigned bitfield type in a bit insertion sequence if it does not
11047 have a larger precision than the bitfield size.
11048 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
11050 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
11052 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
11054 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
11056 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
11057 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
11058 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
11059 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
11061 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
11063 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
11064 Disable -fdelete-null-pointer-checks for ELF toolchain.
11066 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
11067 Kito Cheng <kito.cheng@gmail.com>
11069 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
11070 (nds32le-*-*, nds32be-*-*): Integrate checking process.
11071 (nds32*-*-*): Add glibc and uclibc conditions.
11072 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
11073 (TARGET_EXCEPT_UNWIND_INFO): Define.
11074 * config/nds32/elf.h: New file.
11075 * config/nds32/linux.h: New file.
11076 * config/nds32/nds32-elf.opt: New file.
11077 * config/nds32/nds32-linux.opt: New file.
11078 * config/nds32/nds32-fp-as-gp.c
11079 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
11080 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
11082 (nds32_asm_file_end): Ditto.
11083 (nds32_print_operand): Ditto.
11084 (nds32_insert_attributes): Ditto.
11085 (nds32_init_libfuncs): New function.
11086 (TARGET_HAVE_TLS): Define.
11087 (TARGET_INIT_LIBFUNCS): Define.
11088 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
11090 (TARGET_ELF): Apply different mcmodel setting.
11091 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
11092 been migrated into elf.h and linux.h files.
11093 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
11094 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
11095 (mcmodel): The content has been migrated into nds32-elf.opt and
11096 nds32-linux.opt files.
11097 * config/nds32/t-elf: New file.
11098 * config/nds32/t-linux: New file.
11100 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
11101 Shiva Chen <shiva0217@gmail.com>
11103 * config/nds32/constants.md (unspec_volatile_element): Add
11104 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
11105 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
11107 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
11108 (make_pass_nds32_fp_as_gp): Declare.
11109 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
11111 (nds32_asm_function_end_prologue): Remove unused asm output.
11112 (nds32_asm_function_begin_epilogue): Remove unused asm output.
11113 (nds32_asm_file_start): Output necessary fp_as_gp information.
11114 (nds32_option_override): Adjust register usage.
11115 (nds32_expand_prologue): Consider fp_as_gp situation.
11116 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
11117 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
11120 (simple_return): Ditto.
11121 (omit_fp_begin): Output special directive for fp_as_gp.
11122 (omit_fp_end): Output special directive for fp_as_gp.
11123 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
11124 mforbid-fp-as-gp): New options.
11126 2018-06-01 Mark Wielaard <mark@klomp.org>
11128 * dwarf2out.c (dwarf2out_finish): Remove generation of
11129 DW_AT_loclists_base.
11131 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
11133 * gimple-ssa-store-merging.c: Include gimple-fold.h.
11134 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
11135 (struct merged_store_group): Add bit_insertion field.
11136 (dump_char_array): Use standard hexadecimal format.
11137 (merged_store_group::merged_store_group): Set bit_insertion to false.
11138 (merged_store_group::apply_stores): Use optimal buffer size. Deal
11139 with BIT_INSERT_EXPR stores. Move up code updating the mask and
11140 also print the mask in the dump file.
11141 (pass_store_merging::gate): Minor tweak.
11142 (imm_store_chain_info::coalesce_immediate): Fix wrong association
11143 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
11144 stores with INTEGER_CST stores.
11145 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
11146 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
11147 and use it throughout. Generate bit insertion sequences if need be.
11148 (pass_store_merging::process_store): Remove redundant condition.
11149 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
11151 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
11153 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
11154 the 128-bit floating point types. Fix function comment.
11156 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11158 * config/aarch64/aarch64-simd.md
11159 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
11161 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
11164 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
11166 PR tree-optimization/85989
11167 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
11169 (backprop::intersect_uses): Check it when deciding whether this
11170 is a backedge reference.
11171 (backprop::process_block): Add each phi to m_visited_phis
11172 after visiting it, then clear it at the end.
11174 2018-06-01 Richard Biener <rguenther@suse.de>
11176 * tree-vectorizer.h (vect_dr_stmt): New function.
11177 (vect_get_load_cost): Adjust.
11178 (vect_get_store_cost): Likewise.
11179 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
11180 Use vect_dr_stmt instead of DR_SMTT.
11181 (vect_record_base_alignments): Likewise.
11182 (vect_calculate_target_alignment): Likewise.
11183 (vect_compute_data_ref_alignment): Likewise and make static.
11184 (vect_update_misalignment_for_peel): Likewise.
11185 (vect_verify_datarefs_alignment): Likewise.
11186 (vector_alignment_reachable_p): Likewise.
11187 (vect_get_data_access_cost): Likewise. Pass down
11188 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
11189 (vect_get_peeling_costs_all_drs): Likewise.
11190 (vect_peeling_hash_get_lowest_cost): Likewise.
11191 (vect_enhance_data_refs_alignment): Likewise.
11192 (vect_find_same_alignment_drs): Likewise.
11193 (vect_analyze_data_refs_alignment): Likewise.
11194 (vect_analyze_group_access_1): Likewise.
11195 (vect_analyze_group_access): Likewise.
11196 (vect_analyze_data_ref_access): Likewise.
11197 (vect_analyze_data_ref_accesses): Likewise.
11198 (vect_vfa_segment_size): Likewise.
11199 (vect_small_gap_p): Likewise.
11200 (vectorizable_with_step_bound_p): Likewise.
11201 (vect_prune_runtime_alias_test_list): Likewise.
11202 (vect_analyze_data_refs): Likewise.
11203 (vect_supportable_dr_alignment): Likewise.
11204 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
11205 (vect_gen_prolog_loop_niters): Likewise.
11206 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
11207 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
11209 (vect_recog_mask_conversion_pattern): Likewise.
11210 (vect_try_gather_scatter_pattern): Likewise.
11211 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
11212 to vect_get_store_cost.
11213 (vect_get_store_cost): Get stmt_info instead of DR.
11214 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
11215 (vect_get_load_cost): Get stmt_info instead of DR.
11217 2018-06-01 Richard Biener <rguenther@suse.de>
11219 PR middle-end/86017
11220 * gimple-fold.c (var_decl_component_p): Also allow offsetted
11221 vars wrapped in MEM_REFs.
11223 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
11225 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
11226 Fix subreg tests so that we only return a choice between
11227 GENERAL_REGS and FP_REGS if the original classes included both.
11229 2018-06-01 Richard Biener <rguenther@suse.de>
11232 * tree-ssa-structalias.c (get_function_part_constraint):
11233 Handle NULL fi->decl.
11234 (find_func_aliases_for_call): Properly handle indirect
11235 fi from direct call.
11236 (find_func_clobbers): Likewise.
11237 (ipa_pta_execute): Likewise.
11238 (create_variable_info_for): For functions that are ifunc_resolver
11239 resolve to a varinfo that contains the result of the resolver call.
11240 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
11243 2018-05-31 Michael Collison <michael.collison@arm.com>
11245 * config/aarch64/aarch64.md:
11246 (*fix_to_zero_extenddfdi2): New pattern.
11247 * gcc.target/aarch64/fix_extend1.c: New testcase.
11249 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
11251 PR middle-end/78809
11252 PR middle-end/83026
11253 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
11254 and BUILT_IN_STRNCMP_EQ.
11255 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
11256 BUILT_IN_STRNCMP_EQ.
11257 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
11258 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
11259 (gimple_fold_builtin): Likewise.
11260 * tree-ssa-strlen.c (compute_string_length): New function.
11261 (determine_min_obsize): New function.
11262 (handle_builtin_string_cmp): New function to handle calls to
11263 string compare functions.
11264 (strlen_optimize_stmt): Add handling to builtin string compare
11266 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
11267 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
11268 * tree.c (build_common_builtin_nodes): Add new defines of
11269 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
11271 2018-05-31 Jakub Jelinek <jakub@redhat.com>
11274 * bb-reorder.c (pass_partition_blocks::gate): Return false for
11275 functions with naked attribute.
11277 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
11279 * config/i386/sse.md (avx_vec_concat<mode>):
11280 Substitute concat_tg_mode mode attribute with xtg_mode.
11281 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
11282 (concat_tg_mode): Remove mode attribute.
11284 2018-05-31 Martin Sebor <msebor@redhat.com>
11287 * calls.c (alloc_max_size): Correct a logic error/typo.
11288 Treat excessive arguments as infinite. Warn for invalid arguments.
11289 * doc/invoke.texi (-Walloc-size-larger-than): Update.
11291 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
11294 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
11295 and movx for Haswell.
11297 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
11298 Cesar Philippidis <cesar@codesourcery.com>
11300 PR middle-end/85879
11301 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
11302 when emitting error on private/firstprivate reductions.
11303 * omp-low.c (lower_omp_target): Avoid reference-type processing
11304 on pointers for firstprivate clause.
11306 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
11308 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
11311 * config/aarch64/aarch64-simd.md
11312 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
11313 (aarch64_ld1_x3_<mode>): Likewise
11314 (aarch64_st1x2<VALLDIF:mode>): Likewise
11315 (aarch64_st1_x2_<mode>): Likewise
11316 (aarch64_st1x3<VALLDIF:mode>): Likewise
11317 (aarch64_st1_x3_<mode>): Likewise
11318 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
11319 (vld1_s8_x3): Likewise.
11320 (vld1_u16_x3): Likewise.
11321 (vld1_s16_x3): Likewise.
11322 (vld1_u32_x3): Likewise.
11323 (vld1_s32_x3): Likewise.
11324 (vld1_u64_x3): Likewise.
11325 (vld1_s64_x3): Likewise.
11326 (vld1_f16_x3): Likewise.
11327 (vld1_f32_x3): Likewise.
11328 (vld1_f64_x3): Likewise.
11329 (vld1_p8_x3): Likewise.
11330 (vld1_p16_x3): Likewise.
11331 (vld1_p64_x3): Likewise.
11332 (vld1q_u8_x3): Likewise.
11333 (vld1q_s8_x3): Likewise.
11334 (vld1q_u16_x3): Likewise.
11335 (vld1q_s16_x3): Likewise.
11336 (vld1q_u32_x3): Likewise.
11337 (vld1q_s32_x3): Likewise.
11338 (vld1q_u64_x3): Likewise.
11339 (vld1q_s64_x3): Likewise.
11340 (vld1q_f16_x3): Likewise.
11341 (vld1q_f32_x3): Likewise.
11342 (vld1q_f64_x3): Likewise.
11343 (vld1q_p8_x3): Likewise.
11344 (vld1q_p16_x3): Likewise.
11345 (vld1q_p64_x3): Likewise.
11346 (vst1_s64_x2): Likewise.
11347 (vst1_u64_x2): Likewise.
11348 (vst1_f64_x2): Likewise.
11349 (vst1_s8_x2): Likewise.
11350 (vst1_p8_x2): Likewise.
11351 (vst1_s16_x2): Likewise.
11352 (vst1_p16_x2): Likewise.
11353 (vst1_s32_x2): Likewise.
11354 (vst1_u8_x2): Likewise.
11355 (vst1_u16_x2): Likewise.
11356 (vst1_u32_x2): Likewise.
11357 (vst1_f16_x2): Likewise.
11358 (vst1_f32_x2): Likewise.
11359 (vst1_p64_x2): Likewise.
11360 (vst1q_s8_x2): Likewise.
11361 (vst1q_p8_x2): Likewise.
11362 (vst1q_s16_x2): Likewise.
11363 (vst1q_p16_x2): Likewise.
11364 (vst1q_s32_x2): Likewise.
11365 (vst1q_s64_x2): Likewise.
11366 (vst1q_u8_x2): Likewise.
11367 (vst1q_u16_x2): Likewise.
11368 (vst1q_u32_x2): Likewise.
11369 (vst1q_u64_x2): Likewise.
11370 (vst1q_f16_x2): Likewise.
11371 (vst1q_f32_x2): Likewise.
11372 (vst1q_f64_x2): Likewise.
11373 (vst1q_p64_x2): Likewise.
11374 (vst1_s64_x3): Likewise.
11375 (vst1_u64_x3): Likewise.
11376 (vst1_f64_x3): Likewise.
11377 (vst1_s8_x3): Likewise.
11378 (vst1_p8_x3): Likewise.
11379 (vst1_s16_x3): Likewise.
11380 (vst1_p16_x3): Likewise.
11381 (vst1_s32_x3): Likewise.
11382 (vst1_u8_x3): Likewise.
11383 (vst1_u16_x3): Likewise.
11384 (vst1_u32_x3): Likewise.
11385 (vst1_f16_x3): Likewise.
11386 (vst1_f32_x3): Likewise.
11387 (vst1_p64_x3): Likewise.
11388 (vst1q_s8_x3): Likewise.
11389 (vst1q_p8_x3): Likewise.
11390 (vst1q_s16_x3): Likewise.
11391 (vst1q_p16_x3): Likewise.
11392 (vst1q_s32_x3): Likewise.
11393 (vst1q_s64_x3): Likewise.
11394 (vst1q_u8_x3): Likewise.
11395 (vst1q_u16_x3): Likewise.
11396 (vst1q_u32_x3): Likewise.
11397 (vst1q_u64_x3): Likewise.
11398 (vst1q_f16_x3): Likewise.
11399 (vst1q_f32_x3): Likewise.
11400 (vst1q_f64_x3): Likewise.
11401 (vst1q_p64_x3): Likewise.
11403 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
11405 * config/msp430/msp430.c (msp430_output_labelref): Prepend
11406 user_label_prefix to name.
11408 * tree-core.h: Update comment about the format of NAME string
11409 passed to handler in attribute_spec.
11411 * config/msp430/msp430.md: Remove erroneous subreg expression from
11412 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
11413 zero_extend{q,h}isi2.
11415 2018-05-30 Borislav Petkov <bp@suse.de>
11417 * doc/extend.texi: Document some architecture specific
11418 constraints and sort entries.
11420 2018-05-30 Martin Sebor <msebor@redhat.com>
11422 PR middle-end/85369
11423 * builtins.c (expand_builtin_stpcpy_1): New function.
11424 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
11425 only if the former succeeds.
11427 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
11429 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
11432 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11434 * doc/invoke.texi (-flinker-output): Document
11436 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11438 * passes.c (ipa_write_summaries): Only modify statements if body
11440 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
11441 incrementally linking.
11442 (ipa_passes): Likewise.
11443 * lto-cgraph.c (lto_output_node): When incrementally linking do not
11444 pass down resolution info.
11445 * common.opt (flag_incremental_link): Update info.
11446 * gcc.c (plugin specs): Turn flinker-output=* to
11447 -plugin-opt=-linker-output-known
11448 * toplev.c (compile_file): Also cut compilation when doing incremental
11450 * flag-types. (enum lto_partition_model): Add
11451 LTO_LINKER_OUTPUT_NOLTOREL.
11452 (invoke.texi): Add -flinker-output docs.
11453 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
11454 link same way as WPA; do not stream in dead initializers.
11456 * dwarf2out.c (dwarf2out_die_ref_for_decl,
11457 darf2out_register_external_decl): Support incremental link.
11459 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11461 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
11463 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11465 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
11466 it down to simple_object_copy_lto_debug_sections.
11467 (run_gcc): Determine incremental LTO link time and configure
11468 lto1 into non-wpa mode, disable renaming of debug sections.
11470 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
11472 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
11473 descriptions of various incorrectly documented functions.
11475 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
11478 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
11479 address check not strict.
11481 2018-05-30 Richard Biener <rguenther@suse.de>
11483 PR tree-optimization/85964
11484 * tracer.c (better_p): Drop initialized count check, we only
11485 call the function with initialized counts now.
11486 (find_best_successor): Do find a best edge if one
11487 has uninitialized count.
11488 (find_best_predecessor): Likewise. Do BB frequency check only
11489 if count is initialized.
11491 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
11493 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
11494 (aarch64_ldrstr_offset_compare): New.
11495 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
11496 load/store orderings.
11497 (aarch64_gen_adjusted_ldpstp): Likewise.
11499 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
11501 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
11502 Check for subset of GENERAL_REGS and FP_REGS.
11503 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
11506 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
11508 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
11509 and wi::to_poly_offset. Add the current offset and then check
11510 whether the sum fits, rather than using an unchecked addition of
11511 a checked term. Check for a shwi rather than a uhwi.
11512 * expr.c (get_bit_range): Use tree_to_poly_uint64.
11513 (store_constructor): Use poly_int_tree_p.
11514 (expand_expr_real_1): Likewise.
11515 * function.c (assign_temp): Likewise.
11516 * fold-const.c (const_binop): Use poly_int_tree_p and
11517 wi::to_poly_offset.
11518 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
11520 * ipa-icf-gimple.c (func_checker::compare_operand): Use
11521 to_poly_offset for MEM offsets.
11522 * ipa-icf.c (sem_variable::equals): Likewise.
11523 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
11524 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
11525 wi::to_poly_offset for BIT_FIELD_REF offsets.
11526 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
11527 wi::to_poly_offset.
11528 * var-tracking.c (emit_note_insn_var_location): Use
11529 tree_to_poly_uint64.
11531 2018-05-29 Jim Wilson <jimw@sifive.com>
11533 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
11535 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
11538 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
11539 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
11541 (sse4_1_round<mode>2): Use nonimmediate_operand
11542 for operand 1 predicate.
11544 2018-05-29 Martin Sebor <msebor@redhat.com>
11545 Richard Biener <rguenther@suse.de>
11548 * calls.c (get_size_range): Call determine_value_range instead
11549 of get_value_range..
11550 * tree-vrp.h (determine_value_range): Declared new function.
11551 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
11553 2018-05-29 Richard Biener <rguenther@suse.de>
11555 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
11556 sure to use non-pattern stmts for get_earlier_stmt arguments.
11557 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
11558 called on pattern stmts.
11559 (get_later_stmt): Likewise.
11561 2018-05-29 Martin Liska <mliska@suse.cz>
11563 PR gcov-profile/85759
11564 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
11567 2018-05-29 Jakub Jelinek <jakub@redhat.com>
11569 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
11571 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
11575 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
11576 VEC_PACK_FLOAT_EXPR): New tree codes.
11577 * tree-pretty-print.c (op_code_prio): Handle
11578 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
11579 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
11580 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
11581 * tree-inline.c (estimate_operator_cost): Likewise.
11582 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
11583 * fold-const.c (const_binop): Likewise.
11584 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
11585 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
11586 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
11587 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
11588 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
11589 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
11590 * expr.c (expand_expr_real_2): Likewise.
11591 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
11592 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
11593 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
11595 * optabs.c (expand_widen_pattern_expr): For
11596 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
11597 sign from result type rather than operand's type.
11598 (expand_binop_directly): For vec_packu_float_optab and
11599 vec_packs_float_optab allow result type to be different from operand's
11601 * optabs-tree.c (optab_for_tree_code): Handle
11602 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
11603 VEC_PACK_FLOAT_EXPR. Formatting fixes.
11604 * tree-vect-generic.c (expand_vector_operations_1): Handle
11605 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
11606 VEC_PACK_FLOAT_EXPR.
11607 * tree-vect-stmts.c (supportable_widening_operation): Handle
11609 (supportable_narrowing_operation): Handle FLOAT_EXPR.
11610 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
11611 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
11612 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
11613 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
11615 (vec_pack<floatprefix>_float_<mode>): New expander.
11616 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
11618 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
11619 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
11620 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
11621 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
11622 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
11624 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
11625 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
11626 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
11627 VEC_PACK_FLOAT_EXPR): Document.
11629 2018-05-29 Richard Biener <rguenther@suse.de>
11631 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
11633 (stmt_vec_info_vec): Make pointer.
11634 (init_stmt_vec_info_vec): Remove.
11635 (free_stmt_vec_info_vec): Likewise.
11636 (set_stmt_vec_info_vec): New function.
11637 (free_stmt_vec_infos): Likewise.
11638 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
11639 (set_vinfo_for_stmt): Likewise.
11640 (get_earlier_stmt): Likewise.
11641 (get_later_stmt): Likewise.
11642 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
11643 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
11644 (vec_info::~vec_info): Free stmt_vec_infos.
11645 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
11646 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
11647 (pass_slp_vectorize::execute): Likewise.
11648 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
11649 (free_stmt_vec_info_vec): Likewise.
11650 (set_stmt_vec_info_vec): New function.
11651 (free_stmt_vec_infos): Likewise.
11652 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
11653 the global stmt_vec_info_vec.
11654 * tree-parloops.c (gather_scalar_reductions): Use
11655 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
11658 2018-05-29 Richard Biener <rguenther@suse.de>
11660 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
11662 2018-05-29 Martin Liska <mliska@suse.cz>
11663 David Malcolm <dmalcolm@redhat.com>
11665 * vec.c (test_reverse): New.
11666 (vec_c_tests): Add new test.
11667 * vec.h (vl_ptr>::reverse): New function.
11669 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
11671 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
11673 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
11676 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
11678 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
11680 2018-05-28 Richard Biener <rguenther@suse.de>
11682 PR tree-optimization/85933
11683 * tree-vect-data-refs.c (vect_record_base_alignments): Only
11684 look at stmts marked as vectorizable.
11686 2018-05-28 Richard Biener <rguenther@suse.de>
11688 PR tree-optimization/85934
11689 * tree-vect-generic.c (expand_vector_operations_1): Hoist
11690 vector boolean check before scalar optimization.
11692 2018-05-28 Jakub Jelinek <jakub@redhat.com>
11694 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
11697 2018-05-28 Mark Wielaard <mark@klomp.org>
11699 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
11700 if it is an expression containing a minus sign.
11702 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
11704 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
11706 2018-05-27 Paul Koning <ni1d@arrl.net>
11708 * config/pdp11/pdp11.md (truncsihi2): Remove.
11710 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
11711 Chung-Ju Wu <jasonwucj@gmail.com>
11713 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
11715 (unaligned_store_dw): Ditto.
11716 * config/nds32/nds32-memory-manipulation.c
11717 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
11718 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
11719 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
11720 (emit_setmem_word_loop): Rename to ...
11721 (emit_setmem_doubleword_loop): ... this.
11722 (nds32_gen_dup_4_byte_to_word_value): New function.
11723 (nds32_gen_dup_8_byte_to_double_word_value): New function.
11724 (nds32_expand_setmem_loop): Refine implementation.
11725 (nds32_expand_setmem_loop_v3m): Ditto.
11726 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
11729 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
11731 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
11733 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
11735 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
11736 (nds32_init_machine_status): Initialize machine->attr_naked_p and
11737 machine->attr_no_prologue_p.
11738 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
11739 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
11740 (nds32_expand_epilogue): Consider attr_naked_p.
11741 (nds32_expand_epilogue_v3pop): Likewise.
11742 (nds32_can_use_return_insn): Likewise.
11743 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
11744 attr_no_prologue_p fields.
11745 * config/nds32/nds32.opt (mret-in-naked-func): New option.
11747 2018-05-27 Jakub Jelinek <jakub@redhat.com>
11750 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
11752 * config/i386/sse.md
11753 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
11755 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
11757 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
11759 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
11761 (*<floatsuffix>floatv2div2sf2): Rename to ...
11762 (*float<floatunssuffix>v2div2sf2): ... this.
11763 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
11764 (float<floatunssuffix>v2div2sf2_mask): ... this.
11765 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
11766 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
11767 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
11769 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
11771 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
11773 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
11775 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
11777 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
11779 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
11780 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
11781 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
11782 gen_ufix_truncv8dfv8si2.
11783 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
11784 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
11785 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
11786 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
11787 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
11788 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
11789 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
11790 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
11792 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
11796 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
11798 2018-05-25 Jim Wilson <jimw@sifive.com>
11800 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
11801 * config/riscv/riscv.c (struct machine_function): Add
11802 interrupt_handler_p and attribute_checked_p fields.
11803 (riscv_attribute_table): Add interrupt.
11804 (riscv_interrupt_type_p): New.
11805 (riscv_save_reg_p): Save extra regs for interrupt handler.
11806 (riscv_use_save_libcall): Return false for interrupt handler.
11807 (riscv_first_stack_step): Add forward declaration.
11808 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
11809 for interrupt handler with large frame. Use it for saved reg list.
11810 (riscv_expand_prologue): Move flag_stack_usage_info support to
11811 eliminate duplication.
11812 (riscv_expand_epilogue): Generate mret for interrupt handler.
11813 (riscv_epilogue_uses): New.
11814 (riscv_can_use_return_insn): Return false for interrupt handler.
11815 (riscv_function_ok_for_sibcall): Likewise.
11816 (riscv_set_current_function): Add interrupt handler support.
11817 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
11818 * config/riscv/riscv.md (UNSPECV_MRET): New.
11820 (riscv_frflags, riscv_fsflags): Use tab after opcode.
11822 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
11824 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
11826 PR tree-optimization/85712
11827 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
11828 this candidate has already been replaced in-situ by a copy.
11830 2018-05-25 Jason Merrill <jason@redhat.com>
11832 PR c++/80485 - inline function non-zero address.
11833 * symtab.c (nonzero_address): Check DECL_COMDAT.
11835 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
11838 * config/alpha/alpha.md (ashlsi3): New insn pattern.
11839 (*ashlsi_se): Rename from *ashldi_se. Define as sign
11840 extension of SImode operation. Use const123_operand predicate.
11841 (*saddsi_1): Remove.
11842 (*saddl_se_1): Ditto.
11843 (*ssubsi_1): Ditto.
11844 (*ssubl_se_1): Ditto.
11845 * config/alpha/predicates.md (const123_operand): New predicate.
11846 * config/alpha/constraints.md (P): Use IN_RANGE.
11848 2018-05-25 Richard Biener <rguenther@suse.de>
11850 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
11852 (ref_maybe_used_by_stmt_p): Likewise.
11853 (stmt_may_clobber_ref_p): Likewise.
11854 (stmt_may_clobber_ref_p_1): Likewise.
11855 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
11857 (ref_maybe_used_by_stmt_p): Likewise.
11858 (stmt_may_clobber_ref_p): Likewise.
11859 (stmt_may_clobber_ref_p_1): Likewise.
11860 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
11861 the alias oracle to disambiguate DRs with stmts DR analysis
11863 (vect_analyze_data_refs): Do not give up on not analyzable
11864 DRs for BB vectorization. Remove code truncating the dataref
11867 2018-05-25 Jakub Jelinek <jakub@redhat.com>
11870 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
11871 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
11872 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
11874 2018-05-25 Richard Biener <rguenther@suse.de>
11876 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
11877 function, combining stmt data ref gathering and fatal analysis
11879 (vect_analyze_data_refs): Remove now redudnant code and simplify.
11880 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
11881 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
11882 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
11883 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
11885 2018-05-25 Bin Cheng <bin.cheng@arm.com>
11887 PR tree-optimization/85720
11888 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
11889 SCC if all partitions are builtins.
11890 (version_loop_by_alias_check): New parameter. Generate cancelable
11891 runtime alias check if all partitions are builtins.
11892 (distribute_loop): Update call to above function.
11894 2018-05-25 Bin Cheng <bin.cheng@arm.com>
11896 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
11897 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.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 (remove_ssa_form): Refactor call to init_var_map here.
11903 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
11904 computation for loop region.
11905 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
11906 (register_default_def): Delete.
11907 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
11908 (parm_default_def_partition_arg): Ditto.
11909 (set_parm_default_def_partition): Ditto.
11910 (get_parm_default_def_partitions): Ditto and make it static.
11911 (get_undefined_value_partitions): Ditto and make it static.
11912 (coalesce_with_default, coalesce_with_default): Update comment.
11913 (create_coalesce_list_for_region): New func factored out from
11914 create_outofssa_var_map.
11915 (populate_coalesce_list_for_outofssa): New func factored out from
11916 create_outofssa_var_map and coalesce_ssa_name.
11917 (create_outofssa_var_map): Delete.
11918 (coalesce_ssa_name): Refactor to support live range computation.
11919 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
11920 (get_parm_default_def_partitions): Delete.
11921 (get_undefined_value_partitions): Ditto.
11922 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
11923 computation for loop region.
11924 (new_tree_live_info, loe_visit_block): Ditto.
11925 (live_worklist, set_var_live_on_entry): Ditto.
11926 (calculate_live_on_exit, verify_live_on_entry): Ditto.
11927 * tree-ssa-live.h (struct _var_map): New fields.
11928 (init_var_map): Change decl.
11929 (region_contains_p): New.
11931 2018-05-25 Bin Cheng <bin.cheng@arm.com>
11933 * tree-ssa-live.h (live_merge_and_clear): Delete.
11935 2018-05-25 Richard Biener <rguenther@suse.de>
11938 * tree-dump.c (dequeue_and_dump): Remove access to removed
11939 operand 2 of a SWITCH_EXPR.
11941 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11943 * doc/sourcebuild.texi (vect_double_cond_arith): Include
11944 multiplication and division.
11945 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
11946 (cond_udiv@var{m}, cond_umod@var{m}): Document.
11947 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
11948 (cond_udiv_optab, cond_umod_optab): New optabs.
11949 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
11950 (IFN_COND_RDIV): New internal functions.
11951 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
11952 TRUNC_MOD_EXPR and RDIV_EXPR.
11953 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
11954 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
11956 (SVE_INT_BINARY): Include mult.
11957 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
11958 (optab, sve_int_op): Handle mult.
11959 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
11961 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
11962 for SVE_INT_BINARY_SD.
11964 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11966 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
11967 (optab, sve_int_op): Handle div and udiv.
11968 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
11969 for SVE_INT_BINARY_SD.
11970 (*<optab><mode>3): New insn for the same.
11972 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11974 * tree-vect-patterns.c: Include predict.h.
11975 (vect_recog_divmod_pattern): Restrict check for division support
11976 to when optimizing for size.
11978 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11980 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
11981 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
11982 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
11983 (gimple_match_op::set_op): Likewise.
11984 (gimple_resimplify4): Declare.
11985 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
11986 (expr::gen_transform): Likewise.
11987 (decision_tree::gen): Generate a simplification routine for 4 operands.
11988 * gimple-match-head.c (gimple_simplify): Add an overload for
11989 4 operands. In the top-level function, handle up to 4 call
11990 arguments and call gimple_resimplify4.
11991 (gimple_resimplify4): New function.
11992 (build_call_internal): Pass a fourth operand.
11993 (maybe_push_to_seq): Likewise.
11994 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
11995 Fold VEC_COND_EXPRs of an operation and a default value into
11996 an IFN_COND_* function if possible.
11997 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
11999 (SVE_COND_FP_BINARY): Include them.
12000 (optab, sve_fp_op): Handle them.
12001 (SVE_INT_BINARY_REV): New code iterator.
12002 (SVE_COND_FP_BINARY_REV): New int iterator.
12003 (commutative): New int attribute.
12004 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
12006 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
12008 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
12009 (*cond_<optab><mode>): New patterns for reversed operands.
12011 2018-05-25 Richard Biener <rguenther@suse.de>
12013 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
12014 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
12015 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
12016 (STMT_VINFO_GROUPED_ACCESS): Adjust.
12017 * tree-vect-data-refs.c (everywhere): Adjust users.
12018 * tree-vect-loop.c (everywhere): Likewise.
12019 * tree-vect-slp.c (everywhere): Likewise.
12020 * tree-vect-stmts.c (everywhere): Likewise.
12021 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
12023 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12025 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
12027 (gcc_cv_as_section_exclude): ... this.
12028 Try Solaris as #exclude syntax.
12029 * configure: Regenerate.
12030 * config.in: Regenerate.
12031 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
12033 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
12034 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
12036 * varasm.c (default_elf_asm_named_section): Don't check if
12037 HAVE_GAS_SECTION_EXCLUDE is defined.
12039 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
12041 * doc/md.texi: Update the documentation of the cond_* optabs
12042 to mention the new final operand. Fix GET_MODE_NUNITS call.
12043 Describe the scalar case too.
12044 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
12045 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
12047 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
12048 (get_conditional_internal_fn): Update comment.
12049 * tree-vect-loop.c (vectorizable_reduction): Pass the original
12050 accumulator value as a final argument to conditional functions.
12051 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
12052 a define_expand and add an "else" operand. Assert for now that
12053 the else operand is equal to operand 2. Use SVE_INT_BINARY and
12054 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
12055 (*cond_<optab><mode>): New patterns.
12056 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
12057 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
12058 (UNSPEC_COND_EOR): Delete.
12059 (optab): Remove associated mappings.
12060 (SVE_INT_BINARY): New code iterator.
12061 (sve_int_op): Remove int attribute and add "minus" to the code
12063 (SVE_COND_INT_OP): Delete.
12064 (SVE_COND_FP_OP): Rename to...
12065 (SVE_COND_FP_BINARY): ...this.
12067 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
12069 * optabs.c (can_reuse_operands_p): New function.
12070 (maybe_legitimize_operands): Try to reuse the results for
12073 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
12075 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
12076 Add {q} suffix to insn mnemonic.
12078 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
12080 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
12081 (msp430_warn_func_return): New.
12083 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
12085 * fold-const.c (tree_nonzero_bits): New function.
12086 * fold-const.h (tree_nonzero_bits): Likewise.
12087 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
12088 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
12090 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
12094 * varasm.c (assemble_alias): Check ifunc_resolver only on
12097 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
12100 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
12101 when memory input operand is handled.
12103 2018-05-24 Luis Machado <luis.machado@linaro.org>
12105 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
12107 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
12109 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
12111 * match.pd: Delay FMA folds until after vectorization.
12113 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
12116 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
12117 address check not strict.
12119 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
12121 * gimple-match.h (gimple_match_op): New class.
12122 (mprts_hook): Replace parameters with a gimple_match_op *.
12123 (maybe_build_generic_op): Likewise.
12124 (gimple_simplified_result_is_gimple_val): Replace parameters with
12125 a const gimple_match_op *.
12126 (gimple_simplify): Replace code_helper * and tree * parameters with
12127 a gimple_match_op * parameter.
12128 (gimple_resimplify1): Replace code_helper *, tree and tree *
12129 parameters with a gimple_match_op * parameter.
12130 (gimple_resimplify2): Likewise.
12131 (gimple_resimplify3): Likewise.
12132 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
12133 parameters with a gimple_match_op * parameter.
12134 * gimple-match-head.c (gimple_simplify): Change prototypes of
12135 auto-generated functions to take a gimple_match_op * instead of
12136 separate code_helper * and tree * parameters. Make the same
12137 change in the top-level overload and update calls to the
12138 gimple_resimplify routines. Update calls to the auto-generated
12139 functions and to maybe_push_res_to_seq in the publicly-facing
12140 operation-specific gimple_simplify overloads.
12141 (gimple_match_op::MAX_NUM_OPS): Define.
12142 (gimple_resimplify1): Replace rcode and ops with a single res_op
12143 parameter. Update call to gimple_simplify.
12144 (gimple_resimplify2): Likewise.
12145 (gimple_resimplify3): Likewise.
12146 (mprts_hook): Replace parameters with a gimple_match_op *.
12147 (maybe_build_generic_op): Likewise.
12148 (build_call_internal): Replace type, nargs and ops with
12149 a gimple_match_op *.
12150 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
12151 with a single gimple_match_op *. Update calls to mprts_hook,
12152 build_call_internal and gimple_simplified_result_is_gimple_val.
12153 Factor out code that is common to the tree_code and combined_fn cases.
12154 * genmatch.c (expr::gen_transform): Replace tem_code and
12155 tem_ops with a gimple_match_op called tem_op. Update calls
12156 to the gimple_resimplify functions and maybe_push_res_to_seq.
12157 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
12158 res_ops. Update call to the gimple_resimplify functions.
12159 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
12160 (decision_tree::gen): Make the functions take a gimple_match_op *
12161 called res_op instead of separate res_code and res_ops parameters.
12162 Update call accordingly.
12163 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
12164 and ops with a single res_op parameter. Update calls to
12165 maybe_build_generic_op and maybe_push_res_to_seq.
12166 (fold_stmt_1): Update calls to gimple_simplify and
12167 replace_stmt_with_simplification.
12168 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
12169 and gimple_simplified_result_is_gimple_val.
12170 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
12172 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
12173 with a gimple_match_op *.
12174 (vn_nary_build_or_lookup): Likewise. Update call to
12175 vn_nary_build_or_lookup_1.
12176 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
12177 gimple_match_op *. Update calls to the gimple_resimplify routines
12178 and to gimple_simplified_result_is_gimple_val.
12179 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
12180 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
12181 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
12182 (visit_nary_op): Likewise.
12183 (visit_reference_op_load): Likewise.
12185 2018-05-23 Luis Machado <luis.machado@linaro.org>
12187 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
12188 modifier for printing the step amount.
12190 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
12193 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
12196 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
12198 * doc/sourcebuild.texi (Endianness): New subsubsection.
12200 2018-05-23 Luis Machado <luis.machado@linaro.org>
12202 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12203 <prefetch_dynamic_strides>: New const bool field.
12204 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12205 prefetch_dynamic_strides.
12206 (exynosm1_prefetch_tune): Likewise.
12207 (thunderxt88_prefetch_tune): Likewise.
12208 (thunderx_prefetch_tune): Likewise.
12209 (thunderx2t99_prefetch_tune): Likewise.
12210 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
12212 (aarch64_override_options_internal): Update to set
12213 PARAM_PREFETCH_DYNAMIC_STRIDES.
12214 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
12215 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
12216 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
12217 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
12218 prefetch-dynamic-strides setting.
12220 2018-05-23 Luis Machado <luis.machado@linaro.org>
12222 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12223 <minimum_stride>: New const int field.
12224 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12225 minimum_stride field defaulting to -1.
12226 (exynosm1_prefetch_tune): Likewise.
12227 (thunderxt88_prefetch_tune): Likewise.
12228 (thunderx_prefetch_tune): Likewise.
12229 (thunderx2t99_prefetch_tune): Likewise.
12230 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
12231 <default_opt_level>: Set to 3.
12232 (aarch64_override_options_internal): Update to set
12233 PARAM_PREFETCH_MINIMUM_STRIDE.
12234 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
12235 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
12236 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
12237 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
12238 stride is constant and is below the minimum stride threshold.
12240 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12242 * config/arm/arm-cpus.in (mode26): Delete.
12243 (armv4): Delete mode26 reference.
12244 * config/arm/arm.c (arm_configure_build_target): Delete use of
12247 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
12249 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
12251 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
12252 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
12254 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
12255 (floatunsdidf2): Ditto.
12257 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
12259 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
12260 (fixuns_trunc<mode>si2_avx512f): Ditto.
12261 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
12262 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
12263 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
12265 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
12267 PR rtl-optimization/79985
12268 * df-scan.c (df_insn_refs_collect): Remove special case for
12269 global registers and asm statements.
12271 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
12273 * extend.texi (Global Register Variables): Rewrite the bullet list.
12274 Note that the register is available for allocation. Note that access
12275 via inline asm must use constraints. Add note about async-signal
12276 handlers. Remove paragraph about automagic register selection.
12278 2018-05-23 Richard Biener <rguenther@suse.de>
12280 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
12281 of fixed offset from memset VN.
12283 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
12285 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
12286 first_interp field.
12287 (alloc_cand_and_find_basis): Initialize first_interp field.
12288 (slsr_process_mul): Modify first_interp field.
12289 (slsr_process_add): Likewise.
12290 (slsr_process_cast): Modify first_interp field for each new
12292 (slsr_process_copy): Likewise.
12293 (dump_candidate): Dump first_interp field.
12294 (replace_mult_candidate): Process all interpretations, not just
12296 (replace_rhs_if_not_dup): Likewise.
12297 (replace_one_candidate): Likewise.
12299 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
12301 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
12303 (aarch64_needs_frame_chain): New function.
12304 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
12306 2018-05-23 Sudakshina Das <sudi.das@arm.com>
12309 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
12310 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
12311 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
12312 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
12313 as true for strict-align.
12314 (aarch64_can_inline_p): Perform checks even when callee has no
12315 attributes to check for strict alignment.
12316 * doc/extend.texi (AArch64 Function Attributes): Document
12318 * doc/invoke.texi: (AArch64 Options): Likewise.
12320 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
12322 PR tree-optimization/85853
12323 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
12324 the handling of the root of the node to...
12325 (vect_slp_analyze_node_operations_1): ...this new function,
12326 and run the whole thing with the child nodes' def types
12327 set according to their SLP node's def type.
12329 2018-05-23 Richard Biener <rguenther@suse.de>
12331 PR middle-end/85874
12332 * tree-data-ref.c (create_runtime_alias_checks): Defer
12333 and ignore overflow warnings.
12335 2018-05-23 Yury Gribov <tetra2005@gmail.com>
12337 PR tree-optimization/85822
12338 * tree-vrp.c (is_masked_range_test): Fix handling of negative
12341 2018-05-23 Richard Biener <rguenther@suse.de>
12343 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
12344 memset constants via native_interpret_expr.
12346 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
12349 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
12351 (cgraph_node::create_alias): Likewise.
12352 (cgraph_node::get_availability): Check ifunc_resolver instead
12353 of looking up ifunc attribute.
12354 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
12355 * varasm.c (do_assemble_alias): Likewise.
12356 (assemble_alias): Likewise.
12357 (default_binds_local_p_3): Likewise.
12358 * cgraph.h (cgraph_node): Add ifunc_resolver.
12359 (cgraph_node::only_called_directly_or_aliased_p): Return false
12360 for IFUNC resolver.
12361 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
12363 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
12364 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
12365 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
12366 instead of looking up ifunc attribute.
12368 2018-05-22 Luis Machado <luis.machado@linaro.org>
12370 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
12372 2018-05-22 Martin Sebor <msebor@redhat.com>
12374 PR middle-end/85359
12375 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
12376 only when expasion succeeds.
12377 (expand_builtin_strcmp): Same.
12378 (expand_builtin_strncmp): Same.
12380 2018-05-22 Martin Sebor <msebor@redhat.com>
12382 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
12384 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
12385 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12387 * config/aarch64/aarch64-ldpstp.md: Replace uses of
12388 aarch64_mem_pair_operand with memory_operand and delete operand swapping
12390 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
12391 Add check for legitimate_address.
12392 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
12393 (aarch64_swap_ldrstr_operands): New.
12394 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
12397 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
12398 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12400 * config/aarch64/aarch64.md: New patterns to generate stp
12402 (store_pair_sw, store_pair_dw): New patterns to generate stp for
12403 single words and double words.
12404 (load_pair_sw, load_pair_dw): Likewise.
12405 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
12407 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
12409 * config/aarch64/aarch64-ldpstp.md: Modify peephole
12410 for different mode ldpstp and add peephole for merged zero stores.
12411 Likewise for loads.
12412 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
12414 (aarch64_gen_store_pair): Rename calls to match new patterns.
12415 (aarch64_gen_load_pair): Rename calls to match new patterns.
12416 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
12417 (load_pair<DREG:mode><DREG2:mode>): ... This.
12418 (store_pair<mode>): Rename to...
12419 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
12420 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
12421 New mode iterators.
12422 (V_INT_EQUIV): Handle SImode.
12423 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
12426 2018-05-22 Martin Sebor <msebor@redhat.com>
12429 * calls.c (maybe_warn_nonstring_arg): Use string length to set
12430 or ajust the presumed bound on an operation to avoid unnecessary
12433 2018-05-22 Martin Sebor <msebor@redhat.com>
12435 PR tree-optimization/85826
12436 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
12437 assuming that a DECL necesarily has a constant size.
12439 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
12441 PR middle-end/85862
12442 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
12444 2018-05-22 Richard Biener <rguenther@suse.de>
12446 PR tree-optimization/85834
12447 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
12448 non-constant and non-zero memset arguments.
12450 2018-05-22 Martin Liska <mliska@suse.cz>
12453 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
12455 2018-05-22 Richard Biener <rguenther@suse.de>
12457 PR tree-optimization/85863
12458 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
12459 comparisons when vectype is specified.
12460 (vectorizable_condition): Do not specify vectype for
12461 vect_is_simple_cond when SLP vectorizing.
12463 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
12466 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
12467 define __ibm128 as long double.
12468 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
12469 as a distinct type when IEEE 128-bit support is enabled.
12470 (init_float128_ieee): Fix up conversions between IFmode and IEEE
12471 128-bit types to use the correct functions.
12472 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
12473 convert between 128-bit floating point types that have different
12474 modes but the same representation, instead of using gen_lowpart to
12476 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
12478 (IFKF_reg): New attributes to give the register constraints for
12480 (extend<mode>tf2_internal): New insns to mark an explicit
12481 conversion between 128-bit floating point types that have a
12482 different mode but share the same representation.
12484 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
12486 PR tree-optimization/85814
12487 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
12488 a null return from get_strinfo when unsharing the next
12489 strinfo in the chain.
12491 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
12494 * varasm.c (weak_finish): Clean up weak_decls.
12496 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12498 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
12499 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
12500 UNSPEC_UADALP values.
12501 * config/aarch64/iterators.md (ABAL): New int iterator.
12504 (sur): Add mappings for the above.
12505 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
12507 (aarch64_<sur>abal<mode>_4): Likewise.
12508 (aarch64_<sur>adalp<mode>_3): Likewise.
12509 (<sur>sadv16qi): New define_expand.
12511 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
12513 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
12514 (*movdf_internal): Ditto.
12515 (*rcpsf2_sse): Ditto.
12516 (*rsqrtsf2_sse): Ditto.
12517 (*sqrt<mode>2_sse): Ditto.
12519 2018-05-21 Tamar Christina <tamar.christina@arm.com>
12521 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
12523 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
12524 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
12525 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
12526 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
12528 * config/aarch64/arm_neon.h: Likewise.
12529 * config/aarch64/iterators.md (VQ_I): New.
12531 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
12533 * config.gcc: Add arc/t-multilib-linux to tmake_file for
12535 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
12538 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
12540 * config/nds32/constraints.md (S): New constraint.
12541 * config/nds32/nds32.md (call_internal): Use constraint S.
12542 (call_value_internal): Likewise.
12543 (sibcall_internal): Likewise.
12544 (sibcall_value_internal): Likewise.
12546 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
12547 Chung-Ju Wu <jasonwucj@gmail.com>
12549 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
12550 into consideration.
12552 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
12553 Chung-Ju Wu <jasonwucj@gmail.com>
12555 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
12556 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
12557 (nds32_rtx_costs_impl): Simplify.
12558 (nds32_address_cost_impl): Simplify.
12559 (nds32_init_rtx_costs): New function.
12560 (nds32_rtx_costs_speed_prefer): Likewise.
12561 (nds32_rtx_costs_size_prefer): Likewise.
12562 (nds32_address_cost_speed_prefer): Likewise.
12563 (nds32_address_cost_speed_fwprop): Likewise.
12564 (nds32_address_cost_size_prefer): Likewise.
12565 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
12566 * config/nds32/nds32.c (nds32_option_override): Use
12567 nds32_init_rtx_costs function.
12569 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
12571 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
12572 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
12573 (TARGET_PIPELINE_N8): Likewise.
12574 (TARGET_PIPELINE_N10): Likewise.
12575 (TARGET_PIPELINE_N13): Likewise.
12576 (TARGET_PIPELINE_GRAYWOLF): Likewise.
12578 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
12580 * config/nds32/nds32-fpu.md: Update copyright year.
12582 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12584 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
12586 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12588 * config/nds32/nds32.c
12589 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
12590 * config/nds32/nds32.opt (minline-asm-r15): New option.
12592 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12594 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
12596 * config/nds32/nds32.md (abssi2): New pattern.
12598 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
12600 * config/i386/i386.md (rex64namesuffix): New mode attribute.
12601 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
12602 Merge insn pattern from sse_cvtsi2ss<round_name> and
12603 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
12604 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
12605 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
12606 using SWI48 mode iterator.
12607 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
12608 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
12609 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
12610 pattern from sse_cvttss2si<round_saeonly_name>
12611 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
12612 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
12613 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
12614 using SWI48 mode iterator.
12615 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
12616 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
12617 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
12618 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
12619 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
12620 using SWI48 mode iterator.
12621 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
12622 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
12623 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
12624 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
12625 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
12626 SWI48 mode iterator.
12627 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
12628 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
12629 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
12630 pattern from sse_cvttsd2si<round_saeonly_name>
12631 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
12633 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12635 * config/nds32/nds32-md-auxiliary.c
12636 (nds32_valid_smw_lwm_base_p): Refine.
12637 (nds32_output_smw_single_word): Refine.
12638 (nds32_output_smw_double_word): New.
12639 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
12641 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12643 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
12644 (nds32_output_stack_pop): Refine.
12645 (nds32_expand_unaligned_load): Refine.
12646 (nds32_expand_unaligned_store): Refine.
12648 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
12649 Chung-Ju Wu <jasonwucj@gmail.com>
12651 * config/nds32/constants.md: Add TP_REGNUM constant.
12652 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
12653 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
12655 * config/nds32/nds32-doubleword.md: Consider flag_pic.
12656 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
12657 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
12658 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
12659 and PIC code generation.
12660 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
12662 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
12664 * config/nds32/nds32.md: Support TLS and PIC.
12665 * config/nds32/nds32.c: Support TLS and PIC.
12666 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
12667 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
12670 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12672 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
12673 mode with E_ prefix.
12675 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
12676 Chung-Ju Wu <jasonwucj@gmail.com>
12678 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
12679 * config/nds32/nds32-md-auxiliary.c
12680 (symbolic_reference_mentioned_p): New.
12681 (nds32_legitimize_ict_address): New.
12682 (nds32_expand_ict_move): New.
12683 (nds32_indirect_call_referenced_p): New.
12684 (nds32_symbol_binds_local_p): Delete.
12685 (nds32_long_call_p): Modify.
12686 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
12687 * config/nds32/nds32-protos.h
12688 (symbolic_reference_mentioned_p): Declare.
12689 (nds32_legitimize_ict_address): Declare.
12690 (nds32_expand_ict_move): Declare.
12691 (nds32_indirect_call_referenced_p): Declare.
12692 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
12693 (nds32_relax_group): Use nds32_ict_const_p as condition.
12694 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
12695 (nds32_asm_file_start): Output ict_model directive in asm code.
12696 (nds32_legitimate_address_p): Consider indirect call.
12697 (nds32_print_operand): Consider indirect call.
12698 (nds32_print_operand_address): Consider indirect call.
12699 (nds32_insert_attributes): Handle "indirect_call" attribute.
12700 (TARGET_LEGITIMATE_ADDRESS_P): Define.
12701 (TARGET_LEGITIMATE_CONSTANT_P): Define.
12702 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
12703 (TARGET_DELEGITIMIZE_ADDRESS): Define.
12704 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
12705 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
12706 (TARGET_ICT_MODEL_SMALL): Define.
12707 (TARGET_ICT_MODEL_LARGE): Define.
12708 * config/nds32/nds32.md (movsi): Consider ict model.
12709 (call, call_value): Consider ict model.
12710 (sibcall, sibcall_value): Consider ict model.
12711 * config/nds32/nds32.opt (mict-model): New option.
12712 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
12715 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
12716 Monk Chiang <sh.chiang04@gmail.com>
12717 Jim Wilson <jimw@sifive.com>
12719 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
12720 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
12721 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
12722 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
12723 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
12724 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
12725 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
12726 compute save_libcall_adjustment properly.
12727 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
12728 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
12729 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
12730 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
12731 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
12732 (ABI_SPEC): Handle mabi=ilp32e.
12733 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
12734 (RVE): Add RVE mask.
12735 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
12736 <-march>: Add rv32e as an example.
12738 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
12741 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
12742 (intra_create_variable_infos): Handle C++ constructors.
12744 2018-05-18 Martin Liska <mliska@suse.cz>
12746 * passes.def: Remove a redundant pass.
12748 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
12751 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
12753 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12755 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
12757 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
12759 (armv2, armv2a, armv3, armv3m): Delete architectures.
12760 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
12761 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
12762 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
12764 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
12765 (*mulsidi3adddi): Likewise.
12766 (mulsidi3): Likewise.
12767 (*mulsidi3_nov6): Likewise.
12768 (umulsidi3): Likewise.
12769 (umulsidi3_nov6): Likewise.
12770 (umaddsidi4): Likewise.
12771 (*umulsidi3adddi): Likewise.
12772 (smulsi3_highpart): Likewise.
12773 (*smulsi3_highpart_nov6): Likewise.
12774 (umulsi3_highpart): Likewise.
12775 (*umulsi3_highpart_nov6): Likewise.
12776 * config/arm/arm.h (arm_arch3m): Delete.
12777 * config/arm/arm.c (arm_arch3m): Delete.
12778 (arm_option_override_internal): Update armv3-related comment.
12779 (arm_configure_build_target): Delete use of isa_bit_mode32.
12780 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
12781 (arm_rtx_costs_internal): Delete check of arm_arch3m.
12782 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
12783 (mulsa3): Likewise.
12784 (mulusa3): Likewise.
12785 * config/arm/arm-protos.h (arm_arch3m): Delete.
12786 * config/arm/arm-tables.opt: Regenerate.
12787 * config/arm/arm-tune.md: Likewise.
12788 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
12789 deleted architectures.
12791 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12793 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
12794 (armv5t, armv5te): New features.
12795 (ARMv5, ARMv5e): Delete fgroups.
12796 (ARMv5t, ARMv5te): Adjust for above changes.
12797 (ARMv6m): Likewise.
12798 (armv5, armv5e): Delete arches.
12799 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
12801 (*call_reg_arm): Likewise.
12802 (*call_value_reg_armv5): Likewise.
12803 (*call_value_reg_arm): Likewise.
12804 (*call_symbol): Likewise.
12805 (*call_value_symbol): Likewise.
12806 (*sibcall_insn): Likewise.
12807 (*sibcall_value_insn): Likewise.
12808 (clzsi2): Likewise.
12809 (prefetch): Likewise.
12810 (define_split and define_peephole2 dependent on arm_arch5):
12812 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
12814 (TARGET_ARM_QBIT): Likewise.
12815 (TARGET_DSP_MULTIPLY): Likewise.
12816 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
12817 (arm_arch5, arm_arch5e): Delete.
12818 (arm_arch5t, arm_arch5te): Declare.
12819 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
12820 (arm_arch5t): Declare.
12821 (arm_option_reconfigure_globals): Update for the above.
12822 (arm_options_perform_arch_sanity_checks): Update comment, replace
12823 use of arm_arch5 with arm_arch5t.
12824 (use_return_insn): Likewise.
12825 (arm_emit_call_insn): Likewise.
12826 (output_return_instruction): Likewise.
12827 (arm_final_prescan_insn): Likewise.
12828 (arm_coproc_builtin_available): Likewise.
12829 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
12830 arm_arch5e with arm_arch5t and arm_arch5te.
12831 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
12832 (arm_arch5t, arm_arch5te): Declare.
12833 * config/arm/arm-tables.opt: Regenerate.
12834 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
12835 * config/arm/t-multilib: Likewise.
12836 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
12837 instead of arm_arch5.
12838 (*call_reg_thumb1): Likewise.
12839 (*call_value_reg_thumb1_v5): Likewise.
12840 (*call_value_reg_thumb1): Likewise.
12841 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
12843 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
12845 2018-05-18 Martin Liska <mliska@suse.cz>
12847 PR gcov-profile/84846
12848 * doc/gcov.texi: Document -t option of gcov tool.
12850 2018-05-18 Martin Liska <mliska@suse.cz>
12852 PR gcov-profile/84846
12853 * gcov.c (print_usage): Add new -t option.
12854 (process_args): Handle the option.
12855 (generate_results): Use stdout as output when requested by
12858 2018-05-18 Martin Liska <mliska@suse.cz>
12860 PR gcov-profile/84846
12861 * coverage.c (coverage_init): Write PWD to .gcno file.
12862 * doc/gcov.texi: Document how working directory is printed.
12863 * gcov-dump.c (dump_gcov_file): Print PWD.
12864 * gcov.c (output_intermediate_file): Likewise.
12865 (read_graph_file): Read PWD string.
12866 (output_lines): Print PWD.
12868 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12870 PR middle-end/85817
12871 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
12872 for retval and return false if all args to phi are zero.
12874 2018-05-18 Richard Biener <rguenther@suse.de>
12876 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
12878 (evrp_dom_walker::before_dom_children): Call it.
12880 2018-05-18 Richard Biener <rguenther@suse.de>
12882 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
12883 results when processing array refs with variable index.
12885 2018-05-18 Toon Moene <toon@moene.org>
12887 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
12888 directly after that of -floop-interchange. Indicate that both
12889 options are enabled by default when specifying -O3.
12891 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12893 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
12894 iterator. Delete separate integer-mode vec_set<mode> expander.
12895 (aarch64_simd_vec_setv2di): Delete.
12896 (vec_setv2di): Delete.
12897 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
12898 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
12899 the "w, r" alternative.
12901 2018-05-18 Martin Liska <mliska@suse.cz>
12903 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
12904 * tree-pass.h (make_pass_lower_switch_O0): New function.
12905 * tree-switch-conversion.c (node_has_low_bound): Remove.
12906 (node_has_high_bound): Likewise.
12907 (node_is_bounded): Likewise.
12908 (class pass_lower_switch): Make it a template type and create
12910 (pass_lower_switch::execute): Add template argument.
12911 (make_pass_lower_switch): New function.
12912 (make_pass_lower_switch_O0): New function.
12913 (do_jump_if_equal): Remove.
12914 (emit_case_nodes): Simplify to just handle all 3 cases and leave
12915 all the hard work to tree optimization passes.
12917 2018-05-18 Martin Liska <mliska@suse.cz>
12919 * dbgcnt.c (limit_low): Renamed from limit.
12920 (limit_high): New variable.
12921 (dbg_cnt_is_enabled): Check for upper limit.
12922 (dbg_cnt): Adjust dumping.
12923 (dbg_cnt_set_limit_by_index): Add new argument for high
12925 (dbg_cnt_set_limit_by_name): Likewise.
12926 (dbg_cnt_process_single_pair): Parse new format.
12927 (dbg_cnt_process_opt): Use strtok.
12928 (dbg_cnt_list_all_counters): Remove 'value' and add
12930 * doc/invoke.texi: Document changes.
12932 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
12934 * doc/sourcebuild.texi (scalar_all_fma): Document.
12935 * tree.def (FMA_EXPR): Delete.
12936 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
12937 * internal-fn.c (ternary_direct): New macro.
12938 (expand_ternary_optab_fn): Likewise.
12939 (direct_ternary_optab_supported_p): Likewise.
12940 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
12941 * builtins.c (fold_builtin_fma): Delete.
12942 (fold_builtin_3): Don't call it.
12943 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
12944 * expr.c (expand_expr_real_2): Likewise.
12945 * fold-const.c (operand_equal_p): Likewise.
12946 (fold_ternary_loc): Likewise.
12947 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
12948 * gimple.c (DEFTREECODE): Likewise.
12949 * gimplify.c (gimplify_expr): Likewise.
12950 * optabs-tree.c (optab_for_tree_code): Likewise.
12951 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
12952 * tree-eh.c (operation_could_trap_p): Likewise.
12953 (stmt_could_throw_1_p): Likewise.
12954 * tree-inline.c (estimate_operator_cost): Likewise.
12955 * tree-pretty-print.c (dump_generic_node): Likewise.
12956 (op_code_prio): Likewise.
12957 * tree-ssa-loop-im.c (stmt_cost): Likewise.
12958 * tree-ssa-operands.c (get_expr_operands): Likewise.
12959 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
12960 * fold-const-call.h (fold_fma): Delete.
12961 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
12962 CFN_FNMA and CFN_FNMS.
12963 (fold_fma): Delete.
12964 * genmatch.c (combined_fn): New enum.
12965 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
12966 (commutative_op): New function.
12967 (commutate): Use it. Handle more than 2 operands.
12968 (dt_operand::gen_gimple_expr): Use commutative_op.
12969 (parser::parse_expr): Allow :c to be used with non-binary
12970 operators if the commutative operand is known.
12971 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
12972 CFN_FMS, CFN_FNMA and CFN_FNMS.
12973 (backprop::process_assign_use): Remove FMA_EXPR handling.
12974 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
12975 (gen_hsa_fma): New function.
12976 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
12977 IFN_FNMA and IFN_FNMS.
12978 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
12979 * gimple-fold.h (follow_all_ssa_edges): Declare.
12980 * gimple-fold.c (follow_all_ssa_edges): New function.
12981 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
12982 gimple_build interface and use follow_all_ssa_edges to fold the result.
12983 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
12984 instead of checking for optabs directly.
12985 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
12986 rather than FMA_EXPRs.
12987 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
12988 call to IFN_FMA instead of an FMA_EXPR.
12990 2018-05-17 Jim Wilson <jimw@sifive.com>
12992 * expr.c (do_tablejump): When converting index to Pmode, if we have a
12993 sign extended promoted subreg, and the range does not have the sign bit
12994 set, then do a sign extend.
12996 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
12997 test, check for sign extended subreg and/or constant operands, and
12998 do a sign extend in that case.
13000 2018-05-17 Steve Ellcey <sellcey@cavium.com>
13002 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
13003 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
13005 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
13006 Change logics_shift_reg to logics_shift_imm.
13007 (thunderx2t99_fp_loadpair_basic): Delete.
13008 (thunderx2t99_fp_storepair_basic): Delete.
13009 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
13010 (thunderx2t99_asimd_polynomial): Delete.
13011 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
13012 and neon_fp_mul_d_scalar_q.
13013 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
13014 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
13015 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
13016 (thunderx2t99_asimd_lut): Add missing tbl types.
13017 (thunderx2t99_asimd_ext): Delete.
13018 (thunderx2t99_asimd_load1_1_mult): Delete.
13019 (thunderx2t99_asimd_load1_2_mult): Delete.
13020 (thunderx2t99_asimd_load1_ldp): New.
13021 (thunderx2t99_asimd_load1): New.
13022 (thunderx2t99_asimd_load2): Add missing *load2* types.
13023 (thunderx2t99_asimd_load3): New.
13024 (thunderx2t99_asimd_load4): New.
13025 (thunderx2t99_asimd_store1_1_mult): Delete.
13026 (thunderx2t99_asimd_store1_2_mult): Delete.
13027 (thunderx2t99_asimd_store2_mult): Delete.
13028 (thunderx2t99_asimd_store2_onelane): Delete.
13029 (thunderx2t99_asimd_store_stp): New.
13030 (thunderx2t99_asimd_store1): New.
13031 (thunderx2t99_asimd_store2): New.
13032 (thunderx2t99_asimd_store3): New.
13033 (thunderx2t99_asimd_store4): New.
13035 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
13037 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
13038 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
13040 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
13041 Segher Boessenkool <segher@kernel.crashing.org>
13044 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
13047 2018-05-17 Richard Biener <rguenther@suse.de>
13049 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
13050 for pruning loop and prune defs feeding only already visited PHIs.
13052 2018-05-17 Richard Biener <rguenther@suse.de>
13054 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
13056 2018-05-17 Bin Cheng <bin.cheng@arm.com>
13057 Richard Biener <rguenther@suse.de>
13059 PR tree-optimization/85793
13060 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
13061 for VMAT_ELEMENTWISE.
13063 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
13065 * internal-fn.h (lookup_internal_fn): Declare
13066 * internal-fn.c (lookup_internal_fn): New function.
13067 * gimple.c (gimple_build_call_from_tree): Handle calls to
13068 internal functions.
13069 * gimple-pretty-print.c (dump_gimple_call): Print "." before
13070 internal function names.
13071 * tree-pretty-print.c (dump_generic_node): Likewise.
13072 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
13074 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
13076 * gimple-fold.h (gimple_build): Make the function forms take
13077 combined_fn rather than built_in_function.
13078 (gimple_simplify): Likewise.
13079 * gimple-match-head.c (gimple_simplify): Likewise.
13080 * gimple-fold.c (gimple_build): Likewise.
13081 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
13082 rather than gimple_build_call_internal.
13083 (get_initial_defs_for_reduction): Likewise.
13084 (vect_create_epilog_for_reduction): Likewise.
13085 (vectorizable_live_operation): Likewise.
13087 2018-05-17 Martin Liska <mliska@suse.cz>
13089 * gimple-ssa-sprintf.c (format_directive): Do not use
13090 space in between 'G_' and '('.
13092 2018-05-17 Jakub Jelinek <jakub@redhat.com>
13095 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
13096 even if the mask is not all ones.
13099 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
13101 (ix86_gimple_fold_builtin): Likewise.
13104 * config/i386/i386.c: Include tree-vector-builder.h.
13105 (ix86_vector_shift_count): New function.
13106 (ix86_fold_builtin): Fold shift builtins by scalar count.
13107 (ix86_gimple_fold_builtin): Likewise.
13109 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
13110 _mm512_setzero): New intrinsics.
13112 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
13113 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13115 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
13116 code generation for cases where splatting a value is not useful.
13117 * simplify-rtx.c (simplify_ternary_operation): Simplify
13118 vec_merge across a vec_duplicate and a paradoxical subreg forming
13119 a vector mode to a vec_concat.
13121 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
13123 * config.gcc: Support "goldmont-plus".
13124 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
13126 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13127 PROCESSOR_GOLDMONT_PLUS.
13128 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
13129 (processor_target_table): Add "goldmont-plus".
13130 (PTA_GOLDMONT_PLUS): Define.
13131 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
13132 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
13133 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
13134 (fold_builtin_cpu): Add "goldmont-plus".
13135 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
13136 (ix86_option_override_internal): Add "goldmont-plus".
13137 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
13138 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
13139 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
13140 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
13142 2018-05-17 Richard Biener <rguenther@suse.de>
13144 PR tree-optimization/85757
13145 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
13146 remove defs that only feed that PHI from further processing.
13148 2018-05-16 Jim Wilson <jimw@sifive.com>
13150 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
13152 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
13154 2018-05-16 Mark Wielaard <mark@klomp.org>
13156 * dwarf2out.c (count_index_strings): New function.
13157 (output_indirect_strings): Call count_index_strings and generate
13158 header for dwarf_version >= 5.
13160 2018-05-16 Mark Wielaard <mark@klomp.org>
13162 * dwarf2out.c (dwarf_FORM): New function.
13163 (set_indirect_string): Use dwarf_FORM.
13164 (reset_indirect_string): Likewise.
13165 (size_of_die): Likewise.
13166 (value_format): Likewise.
13167 (output_die): Likewise.
13168 (add_skeleton_AT_string): Likewise.
13169 (output_macinfo_op): Likewise.
13170 (index_string): Likewise.
13171 (output_index_string_offset): Likewise.
13172 (output_index_string): Likewise.
13173 (count_index_strings): Likewise.
13175 2018-05-16 Carl Love <cel@us.ibm.com>
13177 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
13178 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
13180 2018-05-16 Martin Jambor <mjambor@suse.cz>
13182 * ipa-prop.c (ipa_free_all_edge_args): Remove.
13183 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
13185 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
13187 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
13188 (fnma<mode>4): Likewise.
13189 (fms<mode>4): Likewise.
13190 (fnms<mode>4): Likewise.
13191 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
13192 (aarch64_fnma<mode>4): Likewise.
13193 (aarch64_fms<mode>4): Likewise.
13194 (aarch64_fnms<mode>4): Likewise.
13195 (aarch64_fnmadd<mode>4): Likewise.
13197 2018-05-16 Jason Merrill <jason@redhat.com>
13199 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
13201 2018-05-16 Richard Biener <rguenther@suse.de>
13203 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
13204 (dump_stmt_cost): Declare.
13205 (add_stmt_cost): Dump cost we add.
13206 (add_stmt_costs): New function.
13207 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
13208 No longer exported.
13209 (vect_analyze_stmt): Adjust prototype.
13210 (vectorizable_condition): Likewise.
13211 (vectorizable_live_operation): Likewise.
13212 (vectorizable_reduction): Likewise.
13213 (vectorizable_induction): Likewise.
13214 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
13215 cost vector to pass to vectorizable_ and record afterwards.
13216 (vect_model_reduction_cost): Take cost vector argument and adjust.
13217 (vect_model_induction_cost): Likewise.
13218 (vectorizable_reduction): Likewise.
13219 (vectorizable_induction): Likewise.
13220 (vectorizable_live_operation): Likewise.
13221 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
13222 SLP_TREE_NUMBER_OF_VEC_STMTS.
13223 (vect_analyze_slp_cost_1): Remove.
13224 (vect_analyze_slp_cost): Likewise.
13225 (vect_slp_analyze_node_operations): Take visited args and
13226 a target cost vector. Avoid processing already visited stmt sets.
13227 (vect_slp_analyze_operations): Use a local cost vector to gather
13228 costs and register those of non-discarded instances.
13229 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
13230 (vect_schedule_slp_instance): Remove copying of
13231 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
13233 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
13234 adding cost. Record cost entry location.
13235 (vect_prologue_cost_for_slp_op): Function to compute cost of
13236 a constant or invariant generated for SLP vect in the prologue,
13237 split out from vect_analyze_slp_cost_1.
13238 (vect_model_simple_cost): Make static. Adjust for SLP costing.
13239 (vect_model_promotion_demotion_cost): Likewise.
13240 (vect_model_store_cost): Likewise, make static.
13241 (vect_model_load_cost): Likewise.
13242 (vectorizable_bswap): Add cost vector arg and adjust.
13243 (vectorizable_call): Likewise.
13244 (vectorizable_simd_clone_call): Likewise.
13245 (vectorizable_conversion): Likewise.
13246 (vectorizable_assignment): Likewise.
13247 (vectorizable_shift): Likewise.
13248 (vectorizable_operation): Likewise.
13249 (vectorizable_store): Likewise.
13250 (vectorizable_load): Likewise.
13251 (vectorizable_condition): Likewise.
13252 (vectorizable_comparison): Likewise.
13253 (can_vectorize_live_stmts): Likewise.
13254 (vect_analyze_stmt): Likewise.
13255 (vect_transform_stmt): Adjust calls to vectorizable_*.
13256 * tree-vectorizer.c: Include gimple-pretty-print.h.
13257 (dump_stmt_cost): New function.
13259 2018-05-16 Richard Biener <rguenther@suse.de>
13261 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
13262 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
13263 * tree-ssa-dse.c: Include tree-ssa-loop.h.
13264 (check_name): New callback.
13265 (dse_classify_store): Track cycles via a visited bitmap of PHI
13266 defs and simplify handling of in-loop and across loop dead stores
13267 and properly fail for loop-variant refs. Handle byte-tracking with
13268 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
13271 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
13273 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
13274 (vect_get_mask_type_for_stmt): Likewise.
13275 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
13277 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
13278 to determine the statement's vector type and the vector type that
13279 should be used for calculating nunits. Deal with cases in which
13280 the type has to be deferred.
13281 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
13282 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
13283 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
13284 (vect_determine_vf_for_stmt): New functions, split out from...
13285 (vect_determine_vectorization_factor): ...here.
13286 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
13287 (vect_get_mask_type_for_stmt): New functions, split out from
13288 vect_determine_vectorization_factor.
13290 2018-05-16 Richard Biener <rguenther@suse.de>
13292 * tree-cfg.c (verify_gimple_assign_ternary): Properly
13293 verify the [VEC_]COND_EXPR embedded comparison.
13295 2018-05-15 Martin Sebor <msebor@redhat.com>
13297 PR tree-optimization/85753
13298 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
13299 RECORD_TYPE in addition to ARRAY_TYPE.
13301 2018-05-15 Martin Sebor <msebor@redhat.com>
13303 PR middle-end/85643
13304 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
13306 2018-05-15 Richard Biener <rguenther@suse.de>
13308 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
13309 add by_clobber_p one. Change algorithm to collect all defs
13310 representing uses we need to walk and try reducing them to
13311 a single one before failing.
13312 (dse_dom_walker::dse_optimize_stmt): Adjust.
13314 2018-05-13 Mark Wielaard <mark@klomp.org>
13316 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
13317 (size_of_loc_descr): Likewise.
13318 (output_loc_operands): Likewise.
13319 (output_loc_operands_raw): Likewise.
13320 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
13321 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
13322 (hash_loc_operands): Likewise.
13323 (compare_loc_operands): Likewise.
13325 2018-05-14 Mark Wielaard <mark@klomp.org>
13327 * dwarf2out.c (count_index_addrs): New function.
13328 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
13330 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13332 PR tree-optimization/83648
13333 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
13334 return value as malloc candidate.
13336 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13339 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
13340 param as true in call to suggest_attribute.
13342 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
13344 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
13345 -mreadonly-in-sdata.
13347 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13349 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
13351 (aarch64_crypto_aesd_fused): Likewise.
13353 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
13355 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
13356 (movsi_aarch64): Likewise.
13357 (load_pairsi): Likewise.
13358 (load_pairdi): Likewise.
13359 (store_pairsi): Likewise.
13360 (store_pairdi): Likewise.
13361 (load_pairsf): Likewise.
13362 (load_pairdf): Likewise.
13363 (store_pairsf): Likewise.
13364 (store_pairdf): Likewise.
13365 (zero_extend): Likewise.
13366 (trunc): Swap alternatives.
13367 (fcvt_target): Add '?' to prefer w over r.
13369 2018-05-14 Jakub Jelinek <jakub@redhat.com>
13372 * config/i386/i386.md: Disallow non-commutative arithmetics in
13373 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
13374 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
13375 in the peephole2 before it.
13377 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
13379 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
13380 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
13381 (ix86_handle_option): Handle -mcldemote.
13382 * config.gcc: New header.
13383 * config/i386/cldemoteintrin.h: New file.
13384 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
13385 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
13387 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13388 OPTION_MASK_ISA_CLDEMOTE.
13389 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
13390 (ix86_valid_target_attribute_inner_p): Ditto.
13391 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
13392 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
13393 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
13394 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
13395 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
13397 * config/i386/i386.opt: Add -mcldemote.
13398 * config/i386/x86intrin.h: New header.
13399 * doc/invoke.texi: Add -mcldemote.
13401 2018-05-14 Richard Biener <rguenther@suse.de>
13403 * doc/match-and-simplify.texi: Adjust :s documentation.
13405 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
13407 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
13408 intended memcpy size.
13409 (REORDER_45): Likewise.
13411 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
13413 * sort.cc: New file.
13414 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
13415 * vec.c (qsort_chk): Use gcc_qsort.
13416 * Makefile.in (OBJS-libcommon): Add sort.o.
13417 (build/sort.o): New target. Use it...
13418 (BUILD_RTL): ... here, and...
13419 (build/gencfn-macros): ... here, and...
13420 (build/genmatch): ... here.
13422 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
13423 Chung-Ju Wu <jasonwucj@gmail.com>
13425 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
13426 * config/nds32/nds32-graywolf.md: New file.
13427 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
13428 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
13430 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
13431 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
13432 * config/nds32/nds32.md (pipeline_model): Add graywolf.
13433 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
13434 * config/nds32/pipelines.md: Include n15 settings.
13436 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
13437 Chung-Ju Wu <jasonwucj@gmail.com>
13439 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
13440 * config/nds32/nds32-n13.md: New file.
13441 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
13442 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
13444 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
13445 * config/nds32/nds32.md (pipeline_model): Add n13.
13446 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
13447 * config/nds32/pipelines.md: Include n13 settings.
13449 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
13450 Chung-Ju Wu <jasonwucj@gmail.com>
13452 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
13453 * config/nds32/nds32-n10.md: New file.
13454 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
13455 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
13457 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
13458 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
13459 * config/nds32/nds32.md (pipeline_model): Add n10.
13460 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
13461 * config/nds32/pipelines.md: Include n10 settings.
13463 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
13464 Kito Cheng <kito.cheng@gmail.com>
13465 Chung-Ju Wu <jasonwucj@gmail.com>
13467 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
13468 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
13469 Add enum values for DSP extension instructions.
13470 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
13472 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
13473 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
13474 New code iterators.
13475 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
13476 * config/nds32/nds32-dspext.md: New file for DSP implementation.
13477 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
13478 * config/nds32/nds32-intrinsic.md: Likewise.
13479 * config/nds32/nds32_intrinsic.h: Likewise.
13480 * config/nds32/nds32-md-auxiliary.c: Likewise.
13481 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
13482 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
13483 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
13484 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
13485 * config/nds32/nds32-protos.h: New declarations for DSP extension.
13486 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
13487 TYPE_DMAC in switch statement.
13488 * config/nds32/nds32.c: New checking and implementation for DSP
13489 extension instructions.
13490 * config/nds32/nds32.h: Likewise.
13491 * config/nds32/nds32.md: Likewise.
13492 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
13493 * config/nds32/predicates.md: Implement new predicates for DSP
13496 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
13498 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
13499 Reformat alternatives and attributes so it is easier to identify
13500 which constraints/attributes go with which instruction.
13501 (mov<mode>_hardfloat32, FMOVE64): Likewise.
13502 (mov<mode>_softfloat32, FMOVE64): Likewise.
13503 (mov<mode>_hardfloat64, FMOVE64): Likewise.
13504 (mov<mode>_softfloat64, FMOVE64): Likewise.
13506 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
13508 * doc/extend.texi (PowerPC Built-in Functions): Rename this
13510 (Basic PowerPC Built-in Functions): The new name of the
13511 subsection previously known as "PowerPC Built-in Functions".
13512 (Basic PowerPC Built-in Functions Available on all Configurations):
13514 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
13515 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
13516 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
13517 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
13519 2018-05-11 Martin Jambor <mjambor@suse.cz>
13522 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
13525 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
13528 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
13530 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
13532 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
13533 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
13534 (ix86_handle_option): Handle -mwaitpkg.
13535 * config.gcc: New header.
13536 * config/i386/cpuid.h (bit_WAITPKG): New bit.
13537 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
13538 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
13540 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13541 OPTION_MASK_ISA_WAITPKG.
13542 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
13543 (ix86_option_override_internal): Add PTA_WAITPKG.
13544 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
13545 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
13546 IX86_BUILTIN_TPAUSE.
13547 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
13548 __builtin_ia32_umwait and __builtin_ia32_tpause.
13549 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
13550 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
13551 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
13552 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
13553 UNSPECV_TPAUSE): New.
13554 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
13555 * config/i386/i386.opt: Add -mwaitpkg.
13556 * config/i386/waitpkgintrin.h: New file.
13557 * config/i386/x86intrin.h: New header.
13558 * doc/invoke.texi: Add -mwaitpkg.
13560 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
13563 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
13565 (cortex-m0): Use armv6s-m isa.
13566 (cortex-m0plus): Likewise.
13567 (cortex-m1): Likewise.
13568 (cortex-m0.small-multiply): Likewise.
13569 (cortex-m0plus.small-multiply): Likewise.
13570 (cortex-m1.small-multiply): Likewise.
13572 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
13573 Jakub Jelinek <jakub@redhat.com>
13575 PR tree-optimization/85692
13576 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
13577 source permute as well.
13579 2018-05-11 Martin Liska <mliska@suse.cz>
13582 * doc/extend.texi: Document LLVM style format for no_sanitize
13585 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
13587 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
13588 mode_supports_vsx_dform_quad to mode_supports_dq_form.
13589 (mode_supports_vsx_dform_quad): Likewise.
13590 (mode_supports_vmx_dform): Move these functions to be next to the
13591 other mode_supports functions.
13592 (mode_supports_dq_form): Likewise.
13593 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
13594 mode_supports_dq_form.
13595 (reg_offset_addressing_ok_p): Likewise.
13596 (offsettable_ok_by_alignment): Likewise.
13597 (rs6000_legitimate_offset_address_p): Likewise.
13598 (legitimate_lo_sum_address_p): Likewise.
13599 (rs6000_legitimize_address): Likewise.
13600 (rs6000_legitimize_reload_address): Likewise.
13601 (rs6000_secondary_reload_inner): Likewise.
13602 (rs6000_preferred_reload_class): Likewise.
13603 (rs6000_output_move_128bit): Likewise.
13605 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
13607 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
13608 Generate SImode target register for null target.
13609 <case IX86_BUILTIN_XGETBV>: Ditto.
13610 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
13611 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
13613 2018-05-10 Carl Love <cel@us.ibm.com>
13615 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
13616 dcbtt and dcbtstt if operands[2] is 0.
13618 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
13621 * config/i386/sse.md (usadv64qi): New expander.
13623 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
13625 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
13626 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
13627 -maltivec=be support.
13628 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
13629 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
13630 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
13631 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
13632 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
13633 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
13634 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
13635 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
13636 altivec_vsumsws): Adjust.
13637 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
13638 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
13639 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
13641 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
13642 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
13643 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
13644 (altivec_lve<VI_char>x): Delete expand.
13645 (*altivec_lve<VI_char>x_internal): Rename to...
13646 (altivec_lve<VI_char>x): ... this.
13647 (altivec_lvxl_<mode>): Delete expand.
13648 (*altivec_lvxl_<mode>_internal): Rename to ...
13649 (altivec_lvxl_<mode>): ... this.
13650 (altivec_stvxl_<mode>): Delete expand.
13651 (*altivec_stvxl_<mode>_internal): Rename to ...
13652 (altivec_stvxl_<mode>): ... this.
13653 (altivec_stve<VI_char>x): Delete expand.
13654 (*altivec_stve<VI_char>x_internal): Rename to ...
13655 (altivec_stve<VI_char>x): ... this.
13656 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
13657 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
13658 reduc_plus_scal_<mode>): Adjust.
13659 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
13661 (rs6000_cpu_cpp_builtins): Adjust.
13662 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
13663 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
13664 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
13665 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
13666 -maltivec=be support.
13667 (rs6000_split_vec_extract_var): Adjust.
13668 (rs6000_split_v4si_init): Adjust.
13669 (swap_selector_for_mode): Delete.
13670 (altivec_expand_lvx_be, altivec_expand_stvx_be,
13671 altivec_expand_stvex_be): Delete.
13672 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
13673 -maltivec=be support.
13674 (rs6000_gimple_fold_builtin): Ditto.
13675 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
13677 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
13678 (TARGET_DIRECT_MOVE_64BIT): Adjust.
13679 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
13680 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
13681 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
13682 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
13683 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
13684 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
13685 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
13686 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
13687 anonymous split): Adjust.
13688 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
13689 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
13691 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
13693 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
13694 when --with-gxx-include-dir is also specified.
13695 * configure: Regenerate.
13697 2018-05-09 Jim Wilson <jimw@sifive.com>
13700 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
13701 * config/riscv/t-withmultilib: New.
13702 * config/riscv/withmultilib.h: New.
13703 * doc/install.texi: Document RISC-V --with-multilib-list support.
13705 2018-05-09 Richard Biener <rguenther@suse.de>
13707 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
13709 (vect_bb_vectorization_profitable_p): Adjust. Compute
13710 actual scalar cost using the cost vector and the add_stmt_cost
13713 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13715 PR rtl-optimization/85645
13716 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
13717 in the REG_CFA_REGISTER note for LR, don't leave it empty.
13719 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13721 PR rtl-optimization/85645
13722 * shrink-wrap.c (spread_components): Return a boolean saying if
13723 anything was changed.
13724 (try_shrink_wrapping_separate): Iterate spread_components until
13725 nothing changes anymore.
13727 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13729 PR rtl-optimization/85645
13730 * regrename.c (build_def_use): Also kill the chains that include the
13731 destination of a REG_CFA_REGISTER note.
13733 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13735 PR rtl-optimization/85645
13736 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
13737 insn that has a REG_CFA_REGISTER note.
13739 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
13741 * cfgexpand.c (expand_clobber): New function.
13742 (expand_gimple_stmt_1): Use it.
13743 * tree-vect-stmts.c (vect_clobber_variable): New function,
13745 (vectorizable_simd_clone_call): ...here.
13746 (vectorizable_store): Emit a clobber either side of an
13747 IFN_STORE_LANES sequence.
13748 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
13750 2018-05-09 Tom de Vries <tom@codesourcery.com>
13753 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
13754 (define_insn "trap_if_false"): Add exit after trap.
13756 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
13758 PR rtl-optimization/85638
13759 * bb-reorder.c: Include common/common-target.h.
13760 (create_forwarder_block): New function extracted from...
13761 (fix_up_crossing_landing_pad): ...here. Rename into...
13762 (dw2_fix_up_crossing_landing_pad): ...this.
13763 (sjlj_fix_up_crossing_landing_pad): New function.
13764 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
13765 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
13766 from both partitions and exit the loop after one iteration.
13768 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
13771 * doc/extend.texi (PowerPC Built-in Functions): Rename this
13773 (Basic PowerPC Built-in Functions): The new name of the
13774 subsection previously known as "PowerPC Built-in Functions".
13775 (Basic PowerPC Built-in Functions Available on all Configurations):
13777 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
13779 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
13781 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
13783 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
13786 2018-05-08 Jim Wilson <jimw@sifive.com>
13788 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
13789 (LD_EMUL_SUFFIX): New.
13790 (LINK_SPEC): Use it.
13792 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
13794 * doc/extend.texi (PowerPC Built-in Functions): Rename this
13796 (Basic PowerPC Built-in Functions): The new name of the
13797 subsection previously known as "PowerPC Built-in Functions".
13798 (Basic PowerPC Built-in Functions Available on all Configurations):
13800 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
13802 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
13804 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
13806 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
13809 2018-05-08 Jakub Jelinek <jakub@redhat.com>
13812 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
13813 after cmpelim optimization.
13815 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
13817 * config.gcc: Support "goldmont".
13818 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
13819 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13820 PROCESSOR_GOLDMONT.
13821 * config/i386/i386.c (m_GOLDMONT): Define.
13822 (processor_target_table): Add "goldmont".
13823 (PTA_GOLDMONT): Define.
13824 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
13825 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
13826 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
13827 (fold_builtin_cpu): Add "goldmont".
13828 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
13829 (ix86_option_override_internal): Add "goldmont".
13830 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
13831 (processor_type): Add PROCESSOR_GOLDMONT.
13832 * config/i386/i386.md: Add CPU "glm".
13833 * config/i386/glm.md: New file.
13834 * config/i386/x86-tune.def: Add m_GOLDMONT.
13835 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
13837 2018-05-08 Jakub Jelinek <jakub@redhat.com>
13840 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
13842 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
13843 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
13844 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
13847 * config/i386/i386.c (ix86_fold_builtin): Handle
13848 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
13851 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
13852 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
13854 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
13857 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
13858 (check_arch): Likewise.
13859 (check_fpu): Return the result rather than printing it.
13860 (end arch): Fix operator precedence.
13861 (end cpu): Likewise.
13862 (END): Print the result from check_fpu.
13864 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
13865 Alan Hayward <alan.hayward@arm.com>
13866 David Sherwood <david.sherwood@arm.com>
13868 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
13869 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
13870 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
13871 (*fcmuo<mode>_and): New patterns.
13873 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
13875 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
13876 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
13877 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
13878 (cmp_op, sve_imm_con): New code attributes.
13879 (SVE_COND_INT_CMP, imm_con): Delete.
13880 (cmp_op): Remove above unspecs from int attribute.
13881 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
13883 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
13884 comparison-specific unspecs.
13885 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
13886 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
13887 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
13888 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
13889 (*vec_fcm<cmp_op><mode>): Rename to...
13890 (*fcm<cmp_op><mode>): ...this and adjust likewise.
13891 (*vec_fcmuo<mode>): Rename to...
13892 (*fcmuo<mode>): ...this and adjust likewise.
13893 (*pred_fcm<cmp_op><mode>): New pattern.
13894 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
13895 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
13897 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
13899 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
13900 (aarch64_emit_sve_predicated_cond): New function.
13901 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
13902 (aarch64_emit_unspec_cond_or): Replace with...
13903 (aarch64_emit_sve_or_conds): ...this new function. Use
13904 aarch64_emit_sve_ptrue_op for the individual comparisons and
13905 aarch64_emit_binop to OR them together.
13906 (aarch64_emit_inverted_unspec_cond): Replace with...
13907 (aarch64_emit_sve_inverted_cond): ...this new function. Use
13908 aarch64_emit_sve_ptrue_op for the comparison and
13909 aarch64_emit_unop to invert the result.
13910 (aarch64_expand_sve_vec_cmp_float): Update after the above
13911 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
13913 2018-05-07 Nathan Sidwell <nathan@acm.org>
13915 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
13916 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
13917 (Backwards Compatibility): Likewise.
13919 2018-05-07 Luis Machado <luis.machado@linaro.org>
13923 2018-05-07 Luis Machado <luis.machado@linaro.org>
13925 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13926 <prefetch_dynamic_strides>: New const bool field.
13927 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13928 prefetch_dynamic_strides.
13929 (exynosm1_prefetch_tune): Likewise.
13930 (thunderxt88_prefetch_tune): Likewise.
13931 (thunderx_prefetch_tune): Likewise.
13932 (thunderx2t99_prefetch_tune): Likewise.
13933 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
13935 (aarch64_override_options_internal): Update to set
13936 PARAM_PREFETCH_DYNAMIC_STRIDES.
13937 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
13938 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
13939 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
13940 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
13941 prefetch-dynamic-strides setting.
13943 2018-05-07 Luis Machado <luis.machado@linaro.org>
13945 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13946 <minimum_stride>: New const int field.
13947 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13948 minimum_stride field.
13949 (exynosm1_prefetch_tune): Likewise.
13950 (thunderxt88_prefetch_tune): Likewise.
13951 (thunderx_prefetch_tune): Likewise.
13952 (thunderx2t99_prefetch_tune): Likewise.
13953 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
13954 (aarch64_override_options_internal): Update to set
13955 PARAM_PREFETCH_MINIMUM_STRIDE.
13956 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
13957 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
13958 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
13959 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
13960 stride is constant and is below the minimum stride threshold.
13962 2018-05-07 Luis Machado <luis.machado@linaro.org>
13964 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
13967 2018-05-07 Luis Machado <luis.machado@linaro.org>
13969 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13970 <prefetch_dynamic_strides>: New const bool field.
13971 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13972 prefetch_dynamic_strides.
13973 (exynosm1_prefetch_tune): Likewise.
13974 (thunderxt88_prefetch_tune): Likewise.
13975 (thunderx_prefetch_tune): Likewise.
13976 (thunderx2t99_prefetch_tune): Likewise.
13977 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
13979 (aarch64_override_options_internal): Update to set
13980 PARAM_PREFETCH_DYNAMIC_STRIDES.
13981 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
13982 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
13983 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
13984 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
13985 prefetch-dynamic-strides setting.
13987 2018-05-07 Luis Machado <luis.machado@linaro.org>
13989 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13990 <minimum_stride>: New const int field.
13991 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13992 minimum_stride field.
13993 (exynosm1_prefetch_tune): Likewise.
13994 (thunderxt88_prefetch_tune): Likewise.
13995 (thunderx_prefetch_tune): Likewise.
13996 (thunderx2t99_prefetch_tune): Likewise.
13997 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
13998 (aarch64_override_options_internal): Update to set
13999 PARAM_PREFETCH_MINIMUM_STRIDE.
14000 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
14001 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
14002 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
14003 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
14004 stride is constant and is below the minimum stride threshold.
14006 2018-05-06 Jakub Jelinek <jakub@redhat.com>
14009 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
14010 the type is addressable. Don't force op into register if it has
14013 2018-05-05 Roland McGrath <mcgrathr@google.com>
14016 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
14017 any section for which we don't know a specific type it should have,
14018 regardless of name. Previously this was done only for the exact
14019 names ".init_array", ".fini_array", and ".preinit_array".
14020 (default_elf_asm_named_section): Add comment about
14021 relationship with default_section_type_flags and SECTION_NOTYPE.
14022 (get_section): Don't consider it a type conflict if one side has
14023 SECTION_NOTYPE and the other doesn't, as long as neither has the
14024 SECTION_BSS et al used in the default_section_type_flags logic.
14026 2018-05-05 Tom de Vries <tom@codesourcery.com>
14029 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
14030 (workaround_barsyncs): New function.
14031 (nvptx_reorg): Use workaround_barsyncs.
14032 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
14033 (define_expand "nvptx_membar_cta"): New define_expand.
14034 (define_insn "*nvptx_membar_cta"): New insn.
14036 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
14038 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
14039 To improve optimization opportunities.
14040 * builtin-types.def: The new needed builtin types for the above.
14042 2018-05-04 Richard Biener <rguenther@suse.de>
14044 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
14045 * gimple-ssa-store-merging.c
14046 (imm_store_chain_info::output_merged_store): Remove redundant create,
14047 release split_store vector contents on failure.
14048 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
14049 scalar stmt vector on cache hit.
14051 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
14053 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
14055 * config.gcc (powerpc-xilinx-eabi*): Remove.
14056 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
14058 (fusion_addis_mem_combo_load): Ditto.
14059 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
14061 (rs6000_cpu_cpp_builtins): Ditto.
14062 * config/rs6000/rs6000-linux.c
14063 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
14064 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
14065 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
14067 (rs6000_setup_reg_addr_masks): Ditto.
14068 (rs6000_init_hard_regno_mode_ok): Ditto.
14069 (rs6000_option_override_internal): Ditto.
14070 (legitimate_lo_sum_address_p): Ditto.
14071 (rs6000_legitimize_address): Ditto.
14072 (rs6000_legitimize_reload_address): Ditto.
14073 (rs6000_legitimate_address_p): Ditto.
14074 (abi_v4_pass_in_fpr): Ditto.
14075 (setup_incoming_varargs): Ditto.
14076 (rs6000_gimplify_va_arg): Ditto.
14077 (rs6000_split_multireg_move): Ditto.
14078 (rs6000_savres_strategy): Ditto.
14079 (rs6000_emit_prologue_components): Ditto.
14080 (rs6000_emit_epilogue_components): Ditto.
14081 (rs6000_emit_prologue): Ditto.
14082 (rs6000_emit_epilogue): Ditto.
14083 (rs6000_elf_file_end): Ditto.
14084 (rs6000_function_value): Ditto.
14085 (rs6000_libcall_value): Ditto.
14086 * config/rs6000/rs6000.h: Ditto.
14087 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
14088 (TARGET_MINMAX): ... this. New.
14089 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
14090 * config/rs6000/rs6000.md: Remove Xilinx FP support.
14091 (*movsi_internal1_single): Delete.
14092 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
14093 mfpu=, mxilinx-fpu): Delete.
14094 * config/rs6000/singlefp.h: Delete.
14095 * config/rs6000/sysv4.h: Remove Xilinx FP support.
14096 * config/rs6000/t-rs6000: Ditto.
14097 * config/rs6000/t-xilinx: Delete.
14098 * config/rs6000/titan.md: Adjust for fp_type removal.
14099 * config/rs6000/vsx.md: Remove Xilinx FP support.
14100 (VStype_simple): Delete.
14101 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
14102 * config/rs6000/xfpu.h: Delete.
14103 * config/rs6000/xfpu.md: Delete.
14104 * config/rs6000/xilinx.h: Delete.
14105 * config/rs6000/xilinx.opt: Delete.
14106 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
14107 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
14109 2018-05-04 Tom de Vries <tom@codesourcery.com>
14112 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
14115 2018-05-04 Richard Biener <rguenther@suse.de>
14117 PR middle-end/85627
14118 * tree-complex.c (update_complex_assignment): We are always in SSA form.
14119 (expand_complex_div_wide): Likewise.
14120 (expand_complex_operations_1): Likewise.
14121 (expand_complex_libcall): Preserve EH info of the original stmt.
14122 (tree_lower_complex): Handle removed blocks.
14123 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
14124 on complex multiplication and division libcall builtins.
14126 2018-05-04 Richard Biener <rguenther@suse.de>
14128 PR middle-end/85574
14129 * fold-const.c (negate_expr_p): Restrict negation of operand
14130 zero of a division to when we know that can happen without
14132 (fold_negate_expr_1): Likewise.
14134 2018-05-04 Jakub Jelinek <jakub@redhat.com>
14137 * real.h (real_nextafter): Declare.
14138 * real.c (real_nextafter): New function.
14139 * fold-const-call.c (fold_const_nextafter): New function.
14140 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
14141 CASE_CFN_NEXTTOWARD.
14142 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
14143 even when arg1_mode is different from arg0_mode.
14145 2018-05-03 Nathan Sidwell <nathan@acm.org>
14147 * doc/extend.texi (Deprecated Features): Remove
14148 -ffriend-injection.
14149 (Backwards Compatibility): Likewise.
14150 * doc/invoke.texi (C++ Language Options): Likewise.
14151 (C++ Dialect Options): Likewise.
14153 2018-05-03 Jakub Jelinek <jakub@redhat.com>
14156 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
14157 _mm512_mask_mullox_epi64): New intrinsics.
14159 2018-05-03 Tom de Vries <tom@codesourcery.com>
14162 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
14163 dump files): Add offload-tree.
14165 2018-05-03 Richard Biener <rguenther@suse.de>
14167 PR tree-optimization/85615
14168 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
14169 to loops not nested in BBs loop father to avoid creating multi-entry
14172 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14174 PR tree-optimization/70291
14175 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
14176 arguments. Change return type to tree. Emit libcall as a new
14177 statement rather than replacing existing one when inplace_p is true.
14178 (expand_complex_multiplication_components): New function.
14179 (expand_complex_multiplication): Expand floating-point complex
14180 multiplication using the above.
14181 (expand_complex_division): Rename inner_type parameter to type.
14182 Update expand_complex_libcall call-site.
14183 (expand_complex_operations_1): Update expand_complex_multiplication
14184 and expand_complex_division call-sites.
14186 2018-05-02 Jakub Jelinek <jakub@redhat.com>
14189 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
14190 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
14191 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
14192 the highest significant bit of the shift count mask is clear. In
14193 check whether and[sq]i3 is needed verify that all significant bits
14194 of the shift count other than the highest are set.
14196 2018-05-02 Tom de Vries <tom@codesourcery.com>
14199 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
14200 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
14201 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
14202 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
14203 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
14204 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
14205 __builtin_goacc_parlevel_size.
14207 2018-05-02 Richard Biener <rguenther@suse.de>
14209 PR tree-optimization/85597
14210 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
14211 do not use split vect_get_vec_defs call but call vect_get_slp_defs
14214 2018-05-02 Tom de Vries <tom@codesourcery.com>
14217 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
14218 dump files): Add ltrans-tree.
14220 2018-05-02 Tom de Vries <tom@codesourcery.com>
14223 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
14224 dump files): Add wpa-ipa.
14226 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
14228 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
14229 powerpc*-*-linux*paired* target.
14230 * config/rs6000/750cl.h: Delete.
14231 * config/rs6000/paired.h: Delete.
14232 * config/rs6000/paired.md: Delete.
14233 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
14235 * config/rs6000/rs6000-builtin.def: Remove paired float support.
14236 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
14237 comment. Remove paired float support.
14238 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
14239 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
14241 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
14242 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
14244 * config/rs6000/rs6000.c: Remove paired float support.
14245 (paired_expand_vector_init, paired_expand_vector_move,
14246 paired_emit_vector_compare, paired_emit_vector_cond_expr,
14247 (paired_expand_lv_builtin, paired_expand_stv_builtin,
14248 paired_expand_builtin, paired_expand_predicate_builtin,
14249 paired_init_builtins): Delete.
14250 * config/rs6000/rs6000.h: Remove paired float support.
14251 * config/rs6000/rs6000.md: Remove paired float support.
14252 (move_from_CR_ov_bit): Delete.
14253 * config/rs6000/rs6000.opt (mpaired): Delete.
14254 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
14255 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
14257 2018-05-02 Richard Biener <rguenther@suse.de>
14259 PR middle-end/85567
14260 * gimplify.c (gimplify_save_expr): When in SSA form allow
14261 SAVE_EXPRs to compute to SSA vars.
14263 2018-05-02 Jakub Jelinek <jakub@redhat.com>
14266 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
14267 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
14268 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
14269 clobber operands[2], instead use a new pseudo. Formatting fixes.
14271 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
14273 PR tree-optimization/85586
14274 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
14275 exit early for statements in the same group if the accesses are
14278 2018-05-02 Tom de Vries <tom@codesourcery.com>
14281 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
14284 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
14286 PR tree-optimization/85143
14287 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
14289 2018-05-01 Tom de Vries <tom@codesourcery.com>
14292 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
14293 not found" error message.
14295 2018-05-01 Tom de Vries <tom@codesourcery.com>
14298 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
14299 * vec.c (test_ordered_remove_if): New function.
14300 (vec_c_tests): Call test_ordered_remove_if.
14301 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
14302 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
14303 * tree-vect-patterns.c (vect_pattern_recog_1): Use
14304 VEC_ORDERED_REMOVE_IF.
14306 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14308 PR tree-optimization/82665
14309 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
14310 pointer subtraction where arguments come from a memchr call.
14312 2018-05-01 Jakub Jelinek <jakub@redhat.com>
14314 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
14315 --push-state --as-needed and --pop-state instead of --as-needed and
14316 --no-as-needed if ld supports it.
14317 * configure: Regenerated.
14320 * doc/install.texi2html: Replace _002d with - and _002a with * in
14321 generated html files using sed.
14323 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14326 * gcc-rich-location.c (blank_line_before_p): New function.
14327 (use_new_line): New function.
14328 (gcc_rich_location::add_fixit_insert_formatted): New function.
14329 * gcc-rich-location.h
14330 (gcc_rich_location::add_fixit_insert_formatted): New function.
14332 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14334 * selftest.c (assert_streq): Rename "expected" and "actual" to
14335 "val1" and "val2". Extend NULL-handling to cover both inputs
14336 symmetrically, while still requiring both to be non-NULL for a pass.
14337 * selftest.h (assert_streq): Rename "expected" and "actual" to
14339 (ASSERT_EQ): Likewise.
14340 (ASSERT_EQ_AT): Likewise.
14341 (ASSERT_KNOWN_EQ): Likewise.
14342 (ASSERT_KNOWN_EQ_AT): Likewise.
14343 (ASSERT_NE): Likewise.
14344 (ASSERT_MAYBE_NE): Likewise.
14345 (ASSERT_MAYBE_NE_AT): Likewise.
14346 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
14347 the assertion to pass.
14348 (ASSERT_STREQ_AT): Likewise.
14350 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
14352 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
14353 interaction with -pie.
14355 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14357 * selftest.h: Fix alphabetization of per-source-file selftest
14360 2018-04-30 Jason Merrill <jason@redhat.com>
14362 PR c++/61982 - dead stores to destroyed objects.
14363 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
14366 2018-04-30 Jason Merrill <jason@redhat.com>
14368 * tree.c (build_clobber): New.
14369 * tree.h: Declare it.
14370 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
14372 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14374 * diagnostic-show-locus.c (layout::layout): Update for
14375 location_get_source_line returning a char_span.
14376 (struct char_span): Move to input.h.
14377 (struct correction): Update for fields in char_span becoming
14379 (struct source_line): Update for location_get_source_line
14380 returning a char_span.
14381 (layout::print_line): Likewise.
14382 * edit-context.c (edited_file::print_content): Likewise.
14383 (edited_file::print_diff_hunk): Likewise.
14384 (edited_file::print_run_of_changed_lines): Likewise.
14385 (edited_file::get_num_lines): Likewise.
14386 (edited_line::edited_line): Likewise.
14387 * final.c (asm_show_source): Likewise.
14388 * input.c (location_get_source_line): Convert return type
14389 from const char * to char_span, losing the final "line_len"
14391 (dump_location_info): Update for the above.
14392 (get_substring_ranges_for_loc): Likewise. Use a char_span
14393 when handling the literal within the line.
14394 (test_reading_source_line): Update for location_get_source_line
14395 returning a char_span.
14396 * input.h (class char_span): Move here from
14397 diagnostic-show-locus.c, converting from a struct to a class.
14398 Make data members private.
14399 (char_span::operator bool): New.
14400 (char_span::length): New.
14401 (char_span::get_buffer): New.
14402 (char_span::operator[]): New.
14403 (char_span::subspan): Make const.
14404 (char_span::xstrdup): New.
14405 (location_get_source_line): Convert return type from const char *
14406 to char_span, losing the final "line_size" param.
14408 2018-04-30 Jan Hubicka <jh@suse.cz>
14410 * lto-wrapper.c (ltrans_priorities): New static var.
14411 (cmp_priority): New.
14412 (run_gcc): Read priorities and if doing parallel build order
14413 the Makefile by them.
14415 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14417 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
14419 2018-04-30 Richard Biener <rguenther@suse.de>
14421 * tree-cfg.c (verify_address): Remove base argument, add
14422 flag whether to check TREE_ADDRESSABLE and do that.
14423 (verify_expr): Remove.
14424 (verify_types_in_gimple_reference): Add pieces from verify_expr.
14425 (verify_gimple_assign_single): Likewise.
14426 (verify_gimple_switch): Likewise.
14427 (verify_expr_location_1): Dereference tp once. Add (disabled)
14428 piece from verify_expr.
14429 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
14431 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
14433 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
14435 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
14437 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
14438 (small_data_pattern): Likewise.
14439 (arc_rewrite_small_data): Likewise.
14440 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
14441 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
14442 (get_symbol_alignment): New function.
14443 (legitimate_small_data_address_p): Likewise.
14444 (legitimate_scaled_address): Update, call
14445 legitimate_small_data_address_p.
14446 (output_sdata): New static variable.
14447 (arc_print_operand): Update how we handle small data operands.
14448 (arc_print_operand_address): Likewise.
14449 (arc_legitimate_address_p): Update, use
14450 legitimate_small_data_address_p.
14451 (arc_rewrite_small_data_p): Remove.
14452 (arc_rewrite_small_data_1): Likewise.
14453 (arc_rewrite_small_data): Likewise.
14454 (small_data_pattern): Likewise.
14455 (compact_sda_memory_operand): Update to use
14456 legitimate_small_data_address_p and get_symbol_alignment.
14457 (prepare_move_operands): Don't rewite sdata pattern.
14458 (prepare_extend_operands): Remove.
14459 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
14461 (zero_extendqisi2): Likewise.
14462 (zero_extendhisi2): Likewise.
14463 (extendqihi2): Likewise.
14464 (extendqisi2): Likewise.
14465 (extendhisi2): Likewise.
14466 (addsi3): Likewise.
14467 (subsi3): Likewise.
14468 (andsi3): Likewise.
14469 * config/arc/constraints.md (Usd): Change it to memory constraint.
14471 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
14473 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
14474 as source of std instructions.
14475 * config/arc/arc.md (movsi_insn): Update pattern predicate to
14476 allow 6-bit constants as source for store instructions.
14477 (movdi_insn): Update instruction pattern to allow 6-bit constants
14478 as source for store instructions.
14480 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
14482 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
14484 2018-04-30 Nathan Sidwell <nathan@acm.org>
14485 Sandra Loosemore <sandra@codesourcery.com>
14487 * dumpfile.c (dump_open): Allow '-' for stdout.
14488 * doc/invoke.texi (Developer Options): Document dump filename
14489 determination early. Document stdin/stdout selection.
14491 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
14493 Microblaze Target: PIC data text relative
14495 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
14496 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
14498 * config/microblaze/microblaze.h (microblaze_constant_address_p):
14499 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
14500 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
14501 New addressing mode for data-text relative position indepenedent code.
14502 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
14503 'ADDRESS_SYMBOLIC_TXT_REL'.
14504 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
14505 (microblaze_legitimate_pic_operand): Exclude function calls from
14506 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
14507 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
14509 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
14510 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
14511 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
14512 for 'address + offset'.
14513 (microblaze_expand_prologue): Add new function prologue call for
14515 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
14516 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
14517 table in case of TARGET_PIC_DATA_TEXT_REL.
14518 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
14519 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
14520 Add new macros 'UNSPEC_TEXT',
14521 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
14522 + exclude function calls from 'UNSPEC_PLT' in case of data text
14524 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
14525 new target hook for generating address diff vector tables in case of
14527 * doc/tm.texi : Regenerate.
14528 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
14529 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
14530 of addr diff vector generation.
14531 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
14532 target hook definition.
14533 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
14534 Add default function for generate_pic_addr_diff_vec -> flag_pic.
14535 * doc/invoke.texi (Add new pic option): Add new microblaze pic
14536 option for data text relative.
14538 2018-04-30 Richard Biener <rguenther@suse.de>
14540 * tree-chrec.h (evolution_function_is_constant_p): Remove
14542 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
14544 2018-04-30 Richard Biener <rguenther@suse.de>
14547 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
14549 2018-04-30 Richard Biener <rguenther@suse.de>
14551 PR tree-optimization/28364
14552 PR tree-optimization/85275
14553 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
14554 copying first exit test.
14556 2018-04-28 Mark Wielaard <mark@klomp.org>
14558 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
14559 dwarf_version >= 5.
14560 (dwarf_AT): Handle DW_AT_addr_base.
14561 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
14563 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
14566 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
14567 (*ashl<dwi>3_doubleword_mask_1): Ditto.
14568 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
14569 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
14571 2018-04-28 Richard Biener <rguenther@suse.de>
14573 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
14574 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
14575 to reflect use. Only add interesting stmts.
14577 2018-04-27 Martin Jambor <mjambor@suse.cz>
14580 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
14581 the jump function allows for passing through aggregate values.
14583 2018-04-27 David Malcolm <dmalcolm@redhat.com>
14585 * input.h (in_system_header_at): Convert from macro to inline
14587 (from_macro_expansion_at): Likewise.
14588 (from_macro_definition_at): Likewise.
14590 2018-04-27 Jeff Law <law@redhat.com>
14592 * config.gcc: Mark tile* targets as deprecated/obsolete.
14594 2018-04-27 Richard Biener <rguenther@suse.de>
14596 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
14599 2018-04-27 Richard Biener <rguenther@suse.de>
14601 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
14603 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
14605 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
14606 with Yd constraint. Set "preferred_for_speed" attribute from
14607 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
14608 with Yd constraint.
14609 (*movdi_internal): Ditto.
14610 (movti_interunit splitters): Remove
14611 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
14612 (movdi_interunit splitters): Ditto.
14613 * config/i386/constraints.md (Ye): Remove.
14614 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
14616 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14619 * config/aarch64/constraints.md (Usg): Limit to 31.
14620 (Usj): Limit to 63.
14622 2018-04-27 Jakub Jelinek <jakub@redhat.com>
14624 PR tree-optimization/85529
14625 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
14626 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
14627 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
14628 zero extension or masking of the MSB bit.
14629 (optimize_range_tests): Add FIRST_BB argument, pass it through
14630 to optimize_range_tests_var_bound.
14631 (maybe_optimize_range_tests, reassociate_bb): Adjust
14632 optimize_range_tests callers.
14634 2018-04-26 Richard Biener <rguenther@suse.de>
14635 Jakub Jelinek <jakub@redhat.com>
14637 * cgraph.h (symbol_table): Just declare debug method here.
14638 * symtab.c (symbol_table::debug): Define.
14640 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
14642 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
14644 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
14646 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
14647 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
14648 (*movdi_internal): Substitute Yi and Yj constraint with x
14649 and Ym and Yn constraint with y constraint. Update "isa"
14650 attribute and set "preferred_for_speed" attribute from
14651 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
14652 (*movsi_internal): Ditto.
14653 (*movdf_internal): Ditto.
14654 (*movsf_internal): Ditto.
14655 (*zero_extendsidi2): Ditto.
14656 * config/i386/sse.md (vec_set<mode>_0): Ditto.
14657 (sse2_loadld): Ditto.
14658 (*vec_extract<ssevecmodelower>_0): Ditto.
14659 (*vec_extractv4si_0_zext_sse4): Ditto.
14660 (vec_concatv2di): Ditto.
14661 (*vec_dup<mode>): Ditto.
14662 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
14663 * config/i386/constraints.md (Yi): Remove.
14668 2018-04-26 Nathan Sidwell <nathan@acm.org>
14670 * dumpfile.c (dump_open): New.
14671 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
14672 (dump_finish): Detect stdio/stderr by value not name.
14674 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
14676 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
14678 2018-04-26 Tom de Vries <tom@codesourcery.com>
14681 * config/nvptx/nvptx.c (verify_neutering_jumps)
14682 (verify_neutering_labels): New function
14683 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
14685 2018-04-26 Tom de Vries <tom@codesourcery.com>
14688 * config/nvptx/nvptx.c (needs_neutering_p): New function.
14689 (nvptx_single): Use needs_neutering_p to skip over insns that do not
14692 2018-04-26 Richard Biener <rguenther@suse.de>
14693 Tom de Vries <tom@codesourcery.com>
14696 * lto-streamer-out.c (output_function): Fixup loops if required to match
14697 discovery done in the reader.
14699 2018-04-26 Richard Biener <rguenther@suse.de>
14701 PR tree-optimization/85116
14702 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
14703 have a loop exit from the single latch predecessor. Remove
14704 case of header with just condition.
14705 (ch_base::copy_headers): Exclude infinite loops from any
14707 (pass_ch::execute): Record exits.
14709 2018-04-26 Richard Biener <rguenther@suse.de>
14711 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
14712 prologue cost vector and pass it to vect_get_load_cost.
14713 (vect_get_peeling_costs_all_drs): Likewise.
14714 (vect_peeling_hash_get_lowest_cost): Likewise.
14715 (vect_enhance_data_refs_alignment): Likewise.
14717 2018-04-26 Richard Biener <rguenther@suse.de>
14719 PR middle-end/85450
14720 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
14721 checking of integer<->pointer conversions.
14722 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
14723 sign-/zero-extending pointer types.
14724 (expand_omp_for_static_chunk): Likewise.
14726 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
14727 Jean Lee <xiaoyur347@gmail.com>
14729 * config/mips/mips.c (mips_asan_shadow_offset): New function.
14730 (TARGET_ASAN_SHADOW_OFFSET): Define.
14731 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
14732 true for -fsanitize=address.
14734 2018-04-25 Mark Wielaard <mark@klomp.org>
14736 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
14739 2018-04-25 Jakub Jelinek <jakub@redhat.com>
14741 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
14742 than "alu", remove explicit "memory" and "imm_disp" attributes.
14743 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
14745 PR middle-end/85414
14746 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
14747 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
14748 gen_lowpart_no_emit.
14750 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
14753 * config/i386/i386.c (ix86_expand_builtin): Change memory
14754 operand to XI, extend p0 to Pmode.
14755 * config/i386/i386.md: Change unspec volatile and operand
14756 1 mode to XI, change operand 0 mode to P.
14758 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14760 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
14761 GET_MODE_MASK before any checking.
14762 (nds32_can_use_bset_p): Likewise.
14763 (nds32_can_use_btgl_p): Likewise.
14765 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14767 * config/nds32/nds32-doubleword.md: New define_split pattern for
14768 illegal register number.
14770 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14772 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
14774 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14776 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
14778 2018-04-25 Richard Biener <rguenther@suse.de>
14780 * lto-streamer.h (LTO_major_version): Bump to 8.
14782 2018-04-25 Jakub Jelinek <jakub@redhat.com>
14784 * BASE-VER: Set to 9.0.0.
14786 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
14788 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
14789 in __abskf2 and __powikf2.
14791 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14794 * config/aarch64/constraints.md (Usg, Usj): New constraints.
14795 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
14796 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
14797 Use the above on operand 2. Reindent.
14798 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
14800 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
14803 * common/config/i386/i386-common.c (ix86_handle_option): Don't
14805 * config/i386/i386.opt (mcet): Removed.
14806 * doc/install.texi: Remove -mcet documentation.
14807 * doc/invoke.texi: Likewise.
14809 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
14812 * doc/install.texi: Remove -mcet from bootstrap-cet.
14814 2018-04-24 Jakub Jelinek <jakub@redhat.com>
14817 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
14818 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
14822 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
14823 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
14824 containing a CONST_VECTOR.
14826 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
14828 * doc/install.texi: Update newlib dependency for nvptx.
14830 2018-04-24 Jakub Jelinek <jakub@redhat.com>
14833 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
14834 instead of INTVAL when shifting x left.
14836 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
14838 PR tree-optimization/85478
14839 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
14840 vect_grouped_store_supported for single element vectors.
14842 2018-04-24 Richard Biener <rguenther@suse.de>
14845 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
14846 load cost increase to the case of non-constant step.
14848 2018-04-24 Jakub Jelinek <jakub@redhat.com>
14851 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
14852 destination if any_malformed_asm.
14854 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
14856 PR middle-end/85496
14857 * expr.c (store_field): In the bitfield case, if the value comes from
14858 a function call and is returned in registers by means of a PARALLEL,
14859 do not change the mode of the temporary unless BLKmode and VOIDmode.
14861 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
14863 PR rtl-optimization/85423
14864 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
14865 dependencies to debug insns when the previous insn is non-debug.
14867 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
14869 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
14870 enums into a single definition.
14871 (fls): Fix predicates and printing.
14874 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
14876 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
14877 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
14878 and short u6 immediate.
14879 (check_if_valid_sleep_operand): Remove.
14880 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
14882 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14884 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
14885 flag_always_save_lp condition.
14886 * config/nds32/nds32.opt (malways-save-lp): New option.
14888 2018-04-22 Shiva Chen <shiva0217@gmail.com>
14890 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
14891 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
14892 * config/nds32/nds32.h
14893 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
14894 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
14896 2018-04-22 Shiva Chen <shiva0217@gmail.com>
14898 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
14899 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
14901 2018-04-22 Shiva Chen <shiva0217@gmail.com>
14902 Chung-Ju Wu <jasonwucj@gmail.com>
14904 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
14906 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
14907 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
14909 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14911 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
14913 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14915 * config/nds32/nds32-protos.h (nds32_data_alignment,
14916 nds32_local_alignment): Declare.
14917 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
14918 nds32_local_alignment): New functions.
14919 (TARGET_CONSTANT_ALIGNMENT): Define.
14920 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
14922 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14924 * config/nds32/nds32.c
14925 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
14926 (TARGET_MODES_TIEABLE_P): Likewise.
14928 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14930 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
14931 level Ofast and Og.
14933 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
14934 Chung-Ju Wu <jasonwucj@gmail.com>
14936 * config/nds32/constants.md (unspec_volatile_element): Add enum values
14937 for unaligned access.
14938 * config/nds32/nds32-intrinsic.c: Implementation of expanding
14940 * config/nds32/nds32-intrinsic.md: Likewise.
14941 * config/nds32/nds32_intrinsic.h: Likewise.
14942 * config/nds32/nds32.h (nds32_builtins): Likewise.
14943 * config/nds32/nds32.opt (munaligned-access): New option.
14944 * config/nds32/nds32.c (nds32_asm_file_start): Display
14945 flag_unaligned_access status.
14947 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
14949 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
14950 -mno-relax is present.
14951 * config/riscv/linux.h (LINK_SPEC): Ditto.
14953 2018-04-20 Martin Sebor <msebor@redhat.com>
14956 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
14958 (gimple_fold_builtin_stxcpy_chk): Same.
14959 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
14961 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
14964 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
14965 __powikf2 when long double is IEEE 128-bit.
14967 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
14969 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
14970 step to make sure stack always aligned.
14972 2018-04-20 Carl Love <cel@us.ibm.com>
14975 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
14976 size check for arg0.
14978 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
14979 Tom de Vries <tom@codesourcery.com>
14982 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
14983 Emit insns for calls too.
14984 (nvptx_find_par): Always look for worker-level predecessor insn.
14985 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
14987 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
14988 (nvptx_process_pars): Propagate frames for calls.
14990 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
14993 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
14995 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
14996 (ix86_handle_option): Don't handle OPT_mibt.
14997 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
14999 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
15001 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
15002 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
15003 (ix86_target_macros): Define __CET__ with flag_cf_protection
15004 for -fcf-protection.
15005 * config/i386/i386.c (isa2_opts): Remove -mibt.
15006 * config/i386/i386.h (TARGET_IBT): Removed.
15007 (TARGET_IBT_P): Likewise.
15008 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
15009 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
15010 * config/i386/i386.opt (mcet): Update help message.
15011 (mshstk): Likewise.
15013 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
15014 -mcet as an alias for -mshstk.
15016 2018-04-20 Richard Biener <rguenther@suse.de>
15018 PR middle-end/85475
15019 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
15020 complexity by forcing a single use of the multiply operand.
15022 2018-04-20 Martin Jambor <mjambor@suse.cz>
15025 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
15026 recursion dependency to only apply to non-clones.
15028 2018-04-20 Martin Jambor <mjambor@suse.cz>
15031 * ipa-cp.c (create_specialized_node): Check that clones of
15032 self-recursive edges exist during IPA-CP.
15034 2018-04-19 Toon Moene <toon@moene.org>
15036 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
15039 2018-04-19 Jakub Jelinek <jakub@redhat.com>
15041 PR tree-optimization/85467
15042 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
15043 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
15044 VECTOR_CST element to type.
15046 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
15049 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
15050 * config/i386/i386.md (builtin_setjmp_setup): Removed.
15051 (builtin_longjmp): Likewise.
15052 (save_stack_nonlocal): New pattern.
15053 (restore_stack_nonlocal): Likewise.
15055 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
15058 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
15059 Replace ASM_OUTPUT_LABEL with fprintf.
15061 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
15064 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
15065 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
15066 * config/i386/i386-c.c (ix86_target_macros_internal): Also
15067 define __IBT__ and __SHSTK__ for -fcf-protection.
15068 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
15070 (ix86_trampoline_init): Likewise.
15071 (x86_output_mi_thunk): Likewise.
15072 (ix86_notrack_prefixed_insn_p): Likewise.
15073 (ix86_option_override_internal): Don't disallow -fcf-protection.
15074 * config/i386/i386.md (rdssp<mode>): Also enable for
15076 (incssp<mode>): Likewise.
15077 (nop_endbr): Likewise.
15078 * config/i386/i386.opt (mcet): Change help message to built-in
15081 (mshstk): Likewise.
15082 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
15083 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
15084 enable CET built-in functions.
15086 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
15088 * common/config/i386/i386-common.c
15089 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
15090 OPTION_MASK_ISA_MOVDIRI_UNSET,
15091 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
15092 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
15093 * config.gcc (movdirintrin.h): New header.
15094 * config/i386/cpuid.h (bit_MOVDIRI,
15095 bit_MOVDIR64B): New bits.
15096 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
15098 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
15099 (VOID, PVOID, PCVOID)): New function types.
15100 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
15101 __builtin_ia32_directstoreu_u64,
15102 __builtin_ia32_movdir64b): New builtins.
15103 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
15104 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
15106 (ix86_valid_target_attribute_inner_p): Ditto.
15107 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
15108 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
15109 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
15110 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
15111 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
15112 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
15113 (movdiri<mode>, movdir64b_<mode>): New.
15114 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
15115 * config/i386/immintrin.h: Include movdirintrin.h.
15116 * config/i386/movdirintrin.h: New file.
15117 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
15119 2018-04-19 Richard Biener <rguenther@suse.de>
15121 PR middle-end/85455
15122 * cfg.c (clear_bb_flags): When loop state says we have
15123 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
15125 2018-04-19 Richard Biener <rguenther@suse.de>
15127 PR tree-optimization/84737
15128 * tree-vect-data-refs.c (vect_copy_ref_info): New function
15129 copying restrict info.
15130 (vect_setup_realignment): Use it.
15131 * tree-vectorizer.h (vect_copy_ref_info): Declare.
15132 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
15133 the first DR to all generated stores.
15134 (vectorizable_load): Likewise for loads.
15136 2018-04-19 Jakub Jelinek <jakub@redhat.com>
15138 PR tree-optimization/85446
15139 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
15140 the integral and pointer types to have the same precision.
15142 * doc/install.texi: Document --disable-cet being the default and
15145 2018-04-18 Martin Liska <mliska@suse.cz>
15147 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
15150 2018-04-18 Martin Liska <mliska@suse.cz>
15153 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
15156 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
15157 arguments if they are comparable.
15159 2018-04-18 Martin Liska <mliska@suse.cz>
15162 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
15165 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
15168 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
15171 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
15172 ENDBR after calling __morestack.
15174 2018-04-18 David Malcolm <dmalcolm@redhat.com>
15177 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
15178 by using gcc_base_ver to generate a gcc_driver_version, and use
15179 it when generating GCC_DRIVER_NAME.
15180 * configure: Regenerate.
15182 2018-04-18 Jakub Jelinek <jakub@redhat.com>
15185 * config.gcc: Obsolete powerpc*-*-*spe*.
15187 2018-04-17 Jakub Jelinek <jakub@redhat.com>
15190 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
15191 (stabstr_D): Change type of unum from unsigned int to
15192 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
15195 2018-04-17 Jim Wilson <jimw@sifive.com>
15198 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
15199 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
15200 Set arg_pointer_offset after using pretend_args_size.
15202 2018-04-17 Jakub Jelinek <jakub@redhat.com>
15204 PR rtl-optimization/85431
15205 * dse.c (record_store): Ignore zero width stores.
15208 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
15209 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
15210 __builtin_stack_restore rather than after it.
15211 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
15212 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
15213 argument instead of virtual_dynamic_stack_rtx.
15215 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
15217 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
15219 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
15220 Add note to error message to explain internal mapping of overloaded
15221 built-in function name to non-overloaded built-in function name.
15222 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
15225 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
15228 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
15229 where the inputs overlap with the output.
15231 2018-04-17 Jakub Jelinek <jakub@redhat.com>
15233 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
15234 (=v, v) alternative and explicit "memory" attribute.
15235 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
15236 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
15238 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
15239 "sselog1" type instead of "sselog".
15240 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
15241 "sselog". Remove explicit "memory" attribute.
15242 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
15243 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
15245 (vec_extract_hi_v32hi): Merge all alternatives into one, use
15246 "sselog1" type instead of "sselog". Remove explicit "memory"
15248 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
15249 use "sselog1" type instead of "sselog". Remove explicit "memory"
15251 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
15252 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
15254 (vec_extract_hi_v64qi): Merge all alternatives into one, use
15255 "sselog1" type instead of "sselog". Remove explicit "memory"
15257 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
15258 use "sselog1" type instead of "sselog". Remove explicit "memory"
15262 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
15264 PR middle-end/85414
15265 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
15268 2018-04-17 Martin Jambor <mjambor@suse.cz>
15271 * ipa-cp.c (create_specialized_node): Call
15272 expand_all_artificial_thunks if necessary.
15274 2018-04-17 Martin Liska <mliska@suse.cz>
15277 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
15278 in message, remote space in between '_G' and '('.
15280 2018-04-17 Jakub Jelinek <jakub@redhat.com>
15283 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
15284 avx512f_vmcmp<mode>3<round_saeonly_name>,
15285 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
15286 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
15287 avx512f_rndscale<mode><round_saeonly_name>,
15288 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
15289 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
15290 Use %<iptr>2 instead of %2 for -masm=intel.
15291 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
15292 avx512f_vcvttss2usi<round_saeonly_name>,
15293 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
15295 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
15296 avx512f_vcvttsd2usi<round_saeonly_name>,
15297 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
15298 Use %q1 instead of %1 for -masm=intel.
15299 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
15300 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
15301 of %3 for -masm=intel.
15302 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
15304 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
15306 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
15308 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
15309 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
15311 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
15313 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
15314 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
15315 %g1 and one with %0 and %1.
15316 (avx512er_vmrcp28<mode><round_saeonly_name>,
15317 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
15318 %1 for -masm=intel.
15319 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
15320 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
15321 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
15322 of %0 and %{%4%} for -masm=intel.
15323 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
15324 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
15325 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
15326 order of %0 and %{%5%}%{z%} for -masm=intel.
15328 2018-04-17 Jan Hubicka <jh@suse.cz>
15331 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
15333 2018-04-17 Martin Liska <mliska@suse.cz>
15336 * multiple_target.c (create_dispatcher_calls): Set apostrophes
15337 for target_clone error message. Make default implementation
15338 clone to be a local declaration.
15339 (separate_attrs): Add new argument and check for an empty
15341 (expand_target_clones): Handle it.
15342 (ipa_target_clone): Make redirection just for target_clones
15345 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
15346 Tom de Vries <tom@codesourcery.com>
15348 PR middle-end/84955
15349 * omp-expand.c (expand_oacc_for): Add dummy false branch for
15350 tiled basic blocks without omp continue statements.
15352 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
15355 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
15356 vec_extract expression as having side effects to make sure it gets
15359 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
15362 * config/i386/i386.c (get_builtin_code_for_version): Check
15365 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
15368 * config.gcc: Support "skylake".
15369 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15371 * config/i386/i386.c (m_SKYLAKE): Define.
15372 (processor_target_table): Add "skylake".
15373 (ix86_option_override_internal): Add "skylake".
15374 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
15375 PROCESSOR_CANNONLAKE.
15376 (get_builtin_code_for_version): Fix priority for
15377 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
15378 PROCESSOR_SKYLAKE-AVX512.
15379 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
15380 (processor_type): Add PROCESSOR_SKYLAKE.
15382 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
15383 Jason Merrill <jason@redhat.com>
15386 * convert.c (convert_to_integer_1): Use direct recursion for
15387 enumeral types and types with a precision less than the number
15388 of bits in their mode.
15390 2018-04-16 Julia Koval <julia.koval@intel.com>
15393 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
15394 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
15396 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
15399 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
15400 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
15401 and -mno-direct-move.
15403 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
15406 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
15407 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
15408 Ensure negative shifts result in {0}.
15410 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
15412 PR rtl-optimization/79916
15413 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
15414 regs (if any) to define how to gnerate SD moves when LRA is in
15417 2018-04-13 Jakub Jelinek <jakub@redhat.com>
15419 PR rtl-optimization/85393
15420 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
15421 * except.c (expand_dw2_landing_pad_for_region): Make static.
15422 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
15423 a label and unconditional jump to old_bb, rather than
15424 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
15427 PR rtl-optimization/85376
15428 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
15429 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
15430 instead of a specific value.
15432 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
15433 Bin Cheng <bin.cheng@arm.com>
15435 PR tree-optimization/82965
15436 PR tree-optimization/83991
15437 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
15438 by_profile_only parameter.
15439 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
15440 information if the loop was predicted to iterate too many times.
15441 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
15443 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
15446 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
15449 2018-04-13 Martin Liska <mliska@suse.cz>
15450 Jakub Jelinek <jakub@redhat.com>
15452 PR middle-end/81657
15453 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
15454 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
15455 * builtins.c (expand_builtin_memory_copy_args): Use
15456 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
15457 handle dest_addr == pc_rtx.
15459 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
15462 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
15463 asked to not generate direct moves.
15464 (fix_trunc<mode>si2_stfiwx): Similar.
15465 (fix_trunc<mode>si2_internal): Similar.
15467 2018-04-12 Jakub Jelinek <jakub@redhat.com>
15470 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
15471 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
15472 lookup if dest in some wider mode is known to be const0_rtx and
15473 if so, record permanent equivalence for it to be ZERO_EXTEND of
15474 the narrower mode destination.
15476 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
15478 * lto-streamer-out.c (output_function): Revert 259346.
15479 * omp-expand.c (expand_oacc_for): Likewise.
15481 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
15483 PR rtl-optimization/85354
15484 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
15485 * sel-sched.c (sel_global_init): ... here.
15487 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
15490 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
15491 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
15492 mode for PE-COFF targets.
15493 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
15494 (i386_pe_asm_lto_end): Likewise.
15495 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
15496 (TARGET_ASM_LTO_END): Likewise.
15497 * config/i386/winnt.c (saved_debug_info_level): New static variable.
15498 (i386_pe_asm_lto_start): New function.
15499 (i386_pe_asm_lto_end): Likewise.
15501 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
15502 Richard Biener <rguenther@suse.de>
15504 PR middle-end/84955
15505 * lto-streamer-out.c (output_function): Fix CFG loop state before
15507 * omp-expand.c (expand_oacc_for): Handle calls to internal
15508 functions like regular functions.
15510 2018-04-12 Richard Biener <rguenther@suse.de>
15513 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
15514 for the early LTO debug to properly generate references to it
15515 during DIE emission. Do not re-use that for the skeleton for
15517 (dwarf2out_early_finish): Likewise.
15519 2018-04-12 Jakub Jelinek <jakub@redhat.com>
15522 * config/i386/sse.md
15523 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
15524 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
15525 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
15526 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
15527 and output is a reg, avoid creating invalid lowpart subreg, but
15528 instead split into a 512-bit move. Don't split if not AVX512VL,
15529 input is xmm16+ reg and output is a mem.
15530 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
15531 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
15532 xmm16+ reg and output is a mem.
15534 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15536 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
15537 also for flag_dwarf2_cfi_asm.
15539 2018-04-12 Jakub Jelinek <jakub@redhat.com>
15541 PR rtl-optimization/85342
15542 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
15543 a bool scalar var inside of the loop instead. Don't try to update
15544 recog_data.operand after failed apply_change_group.
15546 2018-04-12 Tom de Vries <tom@codesourcery.com>
15549 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
15550 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
15551 array with flexible array member as array without given dimension.
15552 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
15553 argument for undefined param to true.
15555 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
15558 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
15559 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
15560 from PowerPC section.
15561 * config/rs6000/sysv4.opt (mcall-): Improve help text.
15562 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
15563 help text that is too long.
15564 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
15565 help text that is too long.
15566 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
15567 help text that is too long.
15569 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
15571 * config/alpha/alpha.md (stack_probe_internal): Rename
15572 from "probe_stack". Update all callers.
15574 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
15576 PR rtl-optimization/84566
15577 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
15578 sched_macro_fuse_insns.
15580 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
15583 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
15584 (compute_block_dependences): ... from here.
15586 2018-04-11 Jakub Jelinek <jakub@redhat.com>
15588 PR tree-optimization/85331
15589 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
15590 from int to HOST_WIDE_INT.
15592 2018-04-11 Martin Jambor <mjambor@suse.cz>
15595 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
15596 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
15597 not the same as the source val.
15598 (cgraph_edge_brings_value_p): New parameter.
15599 (gather_edges_for_value): Pass destination value to
15600 cgraph_edge_brings_value_p.
15601 (perhaps_add_new_callers): Likewise.
15602 (get_info_about_necessary_edges): Likewise and exclude values brought
15603 only by self-recursive edges.
15604 (create_specialized_node): Redirect only clones of self-calling edges.
15605 (+self_recursive_pass_through_p): New function.
15606 (find_more_scalar_values_for_callers_subset): Use it.
15607 (find_aggregate_values_for_callers_subset): Likewise.
15608 (known_aggs_to_agg_replacement_list): Removed.
15609 (decide_whether_version_node): Re-calculate known constants for all
15610 remaining context clones.
15612 2018-04-11 Richard Biener <rguenther@suse.de>
15615 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
15616 from early DWARF output.
15617 (dwarf2out_early_finish): Output line info unconditionally into
15618 early DWARF and add reference to it.
15620 2018-04-11 Jakub Jelinek <jakub@redhat.com>
15623 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
15624 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
15625 other than V2DFmode using iptr mode attribute.
15626 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
15628 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
15630 PR rtl-optimization/84659
15631 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
15633 2018-04-11 Jakub Jelinek <jakub@redhat.com>
15636 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
15638 (output_loc_list): Pass address of a dummy size variable even in the
15639 locview handling loop.
15640 (index_location_lists): Add comment on why skip_loc_list_entry can't
15643 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
15646 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
15649 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
15652 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
15653 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
15654 and -mstring-compare-inline-limit.
15656 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
15659 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
15660 for stack clash protection in a register whenever we need it to be in
15663 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
15665 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
15666 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
15668 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
15671 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
15673 (mlong-double-): Ditto.
15674 * config/rs6000/sysv4.opt (msdata=): Ditto.
15675 (mtls-size=): Ditto.
15677 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
15679 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
15680 erroneous entries for
15681 "vector int vec_ldl (int, long int *)", and
15682 "vector unsigned int vec_ldl (int, unsigned long int *)".
15683 Add comments and entries for
15684 "vector bool char vec_ldl (int, bool char *)",
15685 "vector bool short vec_ldl (int, bool short *)",
15686 "vector bool int vec_ldl (int, bool int *)",
15687 "vector bool long long vec_ldl (int, bool long long *)",
15688 "vector pixel vec_ldl (int, pixel *)",
15689 "vector long long vec_ldl (int, long long *)",
15690 "vector unsigned long long vec_ldl (int, unsigned long long *)".
15691 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
15692 type tree bool_long_long_type_node and correct definition of
15693 bool_V2DI_type_node to make reference to this new type tree.
15694 (rs6000_mangle_type): Replace erroneous reference to
15695 bool_long_type_node with bool_long_long_type_node.
15696 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
15697 comments to emphasize sign distinctions for char and int types and
15698 replace RS6000_BTI_bool_long constant with
15699 RS6000_BTI_bool_long_long constant. Also add comment to restrict
15700 use of RS6000_BTI_pixel.
15701 (bool_long_type_node): Remove this macro definition.
15702 (bool_long_long_type_node): New macro definition
15704 2018-04-10 Jakub Jelinek <jakub@redhat.com>
15706 PR rtl-optimization/85300
15707 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
15708 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
15709 simplify_unary_operation fails.
15711 2018-04-10 Martin Liska <mliska@suse.cz>
15713 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
15714 cgraph_edge and ipa_ref.
15716 2018-04-10 Jakub Jelinek <jakub@redhat.com>
15720 * config/i386/sse.md
15721 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
15722 computation of the VEC_MERGE selector from mask.
15723 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
15724 Fix decoding of the VEC_MERGE selector into mask.
15726 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
15728 PR tree-optimization/85286
15729 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
15731 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
15733 * final.c (final_1): Set insn_last_address as well as
15734 insn_current_address.
15736 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15739 * explow.c (emit_stack_probe): Call validize_mem on memory location
15740 before passing it to gen_probe_stack. Create address operand and
15741 legitimize it for the probe_stack_address case.
15743 2018-04-09 Jan Hubicka <jh@suse.cz>
15746 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
15747 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
15748 * tree.c (free_lang_data_in_type): Fix handling of binfos;
15750 (free_lang_data): Rebuild type inheritance graph.
15752 2018-04-09 Martin Sebor <msebor@redhat.com>
15754 * invoke.texi (-finline-small-functions): Mention other optimization
15756 (-findirect-inlining, -fpartial-inlining): Same.
15757 (-finline-functions-called-once): Same.
15758 (-freorder-blocks-and-partition): Same.
15760 2018-04-09 Jan Hubicka <jh@suse.cz>
15763 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
15764 jumps; choose last target that matches the criteria (i.e.
15765 no partition changes for non-crossing jumps).
15766 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
15767 support for redirecting crossing jumps to non-crossing.
15769 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
15771 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
15772 also for naked functions.
15774 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
15776 * config/arc/arc.md (add_shift): New pattern.
15777 (add_shift2): Likewise.
15778 (sub_shift): Likewise.
15779 (sub_shift_cmp0_noout): Likewise.
15780 (compare_si_ashiftsi): Likewise.
15781 (xbfu_cmp0_noout): New combine pattern.
15782 (xbfu_cmp0"): Likewise.
15783 (movsi_set_cc_insn): Place the predicable variant first.
15784 (commutative_binary_cmp0_noout): Remove clobber.
15785 (commutative_binary_cmp0): New pattern.
15786 (noncommutative_binary_cmp0): Likewise.
15787 (noncommutative_binary_cmp0_noout): Likewise.
15788 (noncommutative_binary_comparison_result_used): Removed.
15789 (rsub_cmp0): New pattern.
15790 (rsub_cmp0_noout): Likewise.
15791 (extzvsi): Changed, keep only meaningful variants.
15792 (SQH, SEZ): New iterators.
15793 (SQH_postfix): New mode attribute.
15794 (SEZ_prefix): New code attribute.
15795 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
15796 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
15797 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
15798 of numerical value.
15799 (noncommutative_operator): Check the availability of barrel
15802 2018-04-09 Richard Biener <rguenther@suse.de>
15804 PR tree-optimization/85284
15805 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
15806 Only use the niter constraining form of simple_iv when the exit
15807 is always executed.
15809 2018-04-09 Tom de Vries <tom@codesourcery.com>
15812 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
15813 (define_expand "*memory_barrier"): New define_expand.
15814 (define_insn "memory_barrier"): New insn.
15816 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15818 PR rtl-optimization/80463
15819 PR rtl-optimization/83972
15820 PR rtl-optimization/83480
15822 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
15823 correct producer for the insn.
15824 (tidy_control_flow): Fixup seqnos in case of debug insns.
15826 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15828 PR rtl-optimization/83913
15830 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
15831 different sched-times when merging exprs.
15833 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15835 PR rtl-optimization/83962
15837 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
15838 tidy_fallthru_edge and tidy_control_flow.
15840 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15842 PR rtl-optimization/83530
15844 * sel-sched.c (force_next_insn): New global variable.
15845 (remove_insn_for_debug): When force_next_insn is true, also leave only
15846 next insn in the ready list.
15847 (sel_sched_region): When the region wasn't scheduled, make another pass
15848 over it with force_next_insn set to 1.
15850 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
15852 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
15854 * config/nds32/constants.md (unspec_volatile_element): Add enum values
15855 for interrupt control.
15856 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
15857 functions for interrupt control.
15858 * config/nds32/nds32-intrinsic.md: Likewise.
15859 * config/nds32/nds32_intrinsic.h: Likewise.
15860 * config/nds32/nds32.h (nds32_builtins): Likewise.
15862 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
15864 * config/nds32/nds32.c (nds32_init_machine_status,
15865 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
15866 strict_aligned_p field.
15867 (nds32_expand_to_rtl_hook): New function.
15868 (TARGET_EXPAND_TO_RTL_HOOK): Define.
15869 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
15871 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15872 Chung-Ju Wu <jasonwucj@gmail.com>
15874 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
15875 * config/nds32/nds32-n7.md: New file.
15876 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
15877 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
15879 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
15880 * config/nds32/nds32.md (pipeline_model): Add n7.
15881 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
15882 * config/nds32/pipelines.md: Include n7 settings.
15884 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15885 Chung-Ju Wu <jasonwucj@gmail.com>
15887 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
15888 * config/nds32/nds32-e8.md: New file.
15889 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
15890 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
15892 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
15893 * config/nds32/nds32.md (pipeline_model): Add e8.
15894 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
15895 * config/nds32/pipelines.md: Include e8 settings.
15897 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15898 Chung-Ju Wu <jasonwucj@gmail.com>
15900 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
15901 * config/nds32/nds32-n8.md: New file.
15902 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
15903 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
15905 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
15906 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
15907 * config/nds32/nds32.md (pipeline_model): Add n8.
15908 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
15909 * config/nds32/pipelines.md: Include n8 settings.
15911 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15912 Chung-Ju Wu <jasonwucj@gmail.com>
15914 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
15915 * config/nds32/nds32-n9-2r1w.md: New file.
15916 * config/nds32/nds32-n9-3r2w.md: New file.
15917 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
15918 nds32_register_ports): New or modify for cpu n9.
15919 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
15921 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
15922 * config/nds32/nds32-utils.c: New file.
15923 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
15924 TARGET_MUL_SLOW): Define.
15925 * config/nds32/nds32.md (pipeline_model): New attribute.
15926 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
15927 New options that support cpu n9.
15928 * config/nds32/pipelines.md: Include n9 settings.
15929 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
15931 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
15933 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
15934 information if necessary.
15935 (output_cond_branch_compare_zero): Likewise.
15936 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
15937 (nds32_target_alignment): Refine for alignment.
15938 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
15939 (FUNCTION_BOUNDARY): Modify.
15940 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
15942 * config/nds32/nds32.opt (malways-align, malign-functions): New.
15944 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
15946 * config/nds32/constants.md (unspec_volatile_element): Add values for
15947 TLB operation and data prefetch.
15948 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
15949 functions for TLB operation and data prefetch.
15950 * config/nds32/nds32-intrinsic.md: Likewise.
15951 * config/nds32/nds32_intrinsic.h: Likewise.
15952 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
15953 (nds32_print_operand): Likewise.
15954 * config/nds32/nds32.h (nds32_builtins): Likewise.
15956 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
15957 Andrew Pinski <pinsika@gcc.gnu.org>
15959 PR middle-end/82976
15960 * match.pd: Use constant_boolean_node of correct type instead of
15961 boolean_true_node or boolean_false_node for simplifying
15962 pointer comparisons to zero.
15964 2018-04-07 Jakub Jelinek <jakub@redhat.com>
15966 PR tree-optimization/80021
15967 * tree.c (verify_type_variant): Make error call in verify_variant_match
15968 translatable and remove final full stop.
15970 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
15972 * config/nds32/constants.md (unspec_volatile_element): Add
15973 UNSPEC_VOLATILE_EH_RETURN.
15974 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
15975 nds32_output_stack_pop): Support dwarf exception handling process.
15976 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
15977 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
15978 exception handling process.
15979 (nds32_compute_stack_frame): Likewise.
15980 (nds32_return_addr_rtx): Likewise.
15981 (nds32_initial_elimination_offset): Likewise.
15982 (nds32_expand_prologue): Likewise.
15983 (nds32_expand_epilogue): Likewise.
15984 (nds32_dynamic_chain_address): New function.
15985 * config/nds32/nds32.h (machine_function): Add fields for dwarf
15986 exception handling.
15987 (DYNAMIC_CHAIN_ADDRESS): Define.
15988 (EH_RETURN_DATA_REGNO): Define.
15989 (EH_RETURN_STACKADJ_RTX): Define.
15990 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
15991 patterns for dwarf exception handling.
15993 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
15995 * config/nds32/nds32.h: Clean up obsolete macros.
15997 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15999 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
16000 Add enum values for particular instructions.
16001 * config/nds32/nds32-intrinsic.c: Implementation of expanding
16002 particular intrinsic functions.
16003 * config/nds32/nds32-intrinsic.md: Likewise.
16004 * config/nds32/nds32_intrinsic.h: Likewise.
16005 * config/nds32/nds32.h (nds32_builtins): Likewise.
16006 * config/nds32/nds32.md (type): Add pbsad and pbsada.
16007 (btst, ave): New patterns for particular instructions.
16009 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
16011 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
16012 Add enum values for atomic load/store and memory sync.
16013 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
16015 * config/nds32/nds32-intrinsic.md: Likewise.
16016 * config/nds32/nds32_intrinsic.h: Likewise.
16017 * config/nds32/nds32.h (nds32_builtins): Likewise.
16019 2018-04-07 Jakub Jelinek <jakub@redhat.com>
16021 PR tree-optimization/85257
16022 * fold-const.c (native_encode_vector): If not all elts could fit
16023 and off is -1, return 0 rather than offset.
16024 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
16025 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
16026 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
16027 adjust buffer in native_interpret_expr call.
16029 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
16031 * config/nds32/constants.md (unspec_volatile_element): Add cache
16032 control enum values.
16033 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
16034 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
16035 * config/nds32/nds32.c (nds32_cctl_names): New.
16036 (nds32_print_operand): Handle cache control register names.
16037 * config/nds32/nds32.h (nds32_builtins): New enum values.
16038 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
16040 * config/nds32/nds32.md (type): Add mmu.
16041 * config/nds32/pipelines.md (simple_insn): Add mmu.
16043 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
16045 * config/nds32/nds32.md (type): Remove call.
16046 * config/nds32/pipelines.md (simple_insn): Likewise.
16048 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
16050 * config/nds32/constants.md (unspec_volatile_element): Add
16051 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
16052 UNSPEC_VOLATILE_FMFCFG.
16053 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
16054 description for fmfcfg and fmfcsr.
16055 (bdesc_1arg): Add fmtcsr.
16056 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
16057 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
16058 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
16059 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
16060 unspec_fmfcfg): New patterns.
16061 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
16062 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
16063 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
16064 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
16065 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
16066 __nds32__fmfcfg): Define.
16068 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
16070 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
16071 intrinsic register names.
16072 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
16073 intrinsic register enum values and macros.
16075 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
16077 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
16078 for load/store addressing form.
16079 (nds32_print_operand_address): Likewise.
16081 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
16084 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
16085 based on LABEL_REF. Remove useless assertion.
16086 (pic_address_needs_scratch): Fix formatting.
16087 (sparc_legitimize_pic_address): Minor tweaks.
16088 (sparc_delegitimize_address): Adjust assertion accordingly.
16089 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
16090 into symbolic_operand.
16091 (movsi_high_pic_label_ref): Likewise.
16092 (movsi_lo_sum_pic_label_ref): Likewise.
16093 (movdi_pic_label_ref): Likewise.
16094 (movdi_high_pic_label_ref): Likewise.
16095 (movdi_lo_sum_pic_label_ref): Likewise.
16097 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
16099 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
16100 custom LIB_SPEC setup.
16102 2018-04-06 Ruslan Bukin <br@bsdpad.com>
16103 Kito Cheng <kito.cheng@gmail.com>
16105 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
16106 * config/riscv/freebsd.h: New.
16108 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
16110 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
16111 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
16114 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
16115 Kito Cheng <kito.cheng@gmail.com>
16117 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
16118 nds32_output_call, nds32_symbol_binds_local_p): New functions.
16119 * config/nds32/nds32-protos.h (nds32_output_call,
16120 nds32_output_return): Declare.
16121 * config/nds32/nds32.md: Refine all the call and return patterns.
16123 2018-04-06 Jakub Jelinek <jakub@redhat.com>
16126 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
16127 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
16129 PR rtl-optimization/84872
16130 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
16131 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
16132 EDGE_CROSSING edge.
16134 2018-04-06 Tamar Christina <tamar.christina@arm.com>
16136 * expr.c (copy_blkmode_to_reg): Revert 254862.
16137 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
16139 2018-04-06 Richard Biener <rguenther@suse.de>
16141 PR middle-end/85244
16142 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
16143 after seeing a component reference with an adjacent field. Treat
16144 refs to arrays at struct end of external decls similar to
16145 refs to unconstrained commons.
16147 2018-04-06 Jakub Jelinek <jakub@redhat.com>
16150 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
16151 look through SAVE_EXPRs with non-side-effects argument. Adjust
16153 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
16156 2018-04-06 Richard Biener <rguenther@suse.de>
16158 PR middle-end/85180
16159 * alias.c (find_base_term): New wrapper around find_base_term
16160 unwinding CSELIB_VAL_PTR changes.
16161 (find_base_term): Do not restore CSELIB_VAL_PTR during the
16164 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
16166 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
16168 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
16169 constant definitions.
16170 ("nop"): lr 0,0 -> nopr r0
16171 ("nop_lr0", "nop_lr1"): New insn definitions.
16173 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
16175 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
16176 NDS32_V3PUSH_AVAILABLE_P macro.
16178 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
16179 Chung-Ju Wu <jasonwucj@gmail.com>
16181 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
16182 (nds32*-*-*): Add float and fpu_config into supported_defaults.
16183 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
16184 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
16185 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
16186 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
16187 * config/nds32/constraints.md: New constraints and checking for hard
16188 float configuration.
16189 * config/nds32/iterators.md: New mode iterator and attribute for hard
16190 float configuration.
16191 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
16193 * config/nds32/nds32-fpu.md: New file.
16194 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
16195 deal with hard float code generation.
16196 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
16198 (abi_type, float_reg_number): New enum type.
16199 * config/nds32/nds32-predicates.c: New predicates for hard float.
16200 * config/nds32/nds32-protos.h: Declare functions for hard float.
16201 * config/nds32/nds32.c: Implementation for hard float configuration.
16202 * config/nds32/nds32.h: Definitions for hard float configuration.
16203 * config/nds32/nds32.md: Include hard float machine description and
16204 modify patterns for hard float configuration.
16205 * config/nds32/nds32.opt: New options for hard float configuration.
16206 * config/nds32/predicates.md: New predicates for hard float
16209 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
16211 * common/config/nds32/nds32-common.c
16212 (nds32_option_optimization_table): Enable -mreleax-hint by default.
16214 2018-04-05 Jakub Jelinek <jakub@redhat.com>
16216 PR middle-end/85195
16217 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
16218 CONSTRUCTOR_ELT (ctor, ...)->value.
16220 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
16223 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
16225 2018-04-05 Tom de Vries <tom@codesourcery.com>
16228 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
16231 2018-04-05 Shiva Chen <shiva0217@gmail.com>
16232 Kito Cheng <kito.cheng@gmail.com>
16234 * config/nds32/constraints.md (U33): Fine-tune checking condition.
16235 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
16236 * config/nds32/nds32.h (nds32_16bit_address_type): Add
16237 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
16239 2018-04-05 Shiva Chen <shiva0217@gmail.com>
16240 Kito Cheng <kito.cheng@gmail.com>
16242 * config/nds32/constraints.md (Ufe): New memory constraint.
16243 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
16244 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
16245 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
16247 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
16248 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
16250 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16252 * config/nds32/nds32.md: Use optimize_size in the condition for
16253 alu-shift instructions.
16255 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16257 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
16259 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16261 * config/nds32/nds32.md (negsi2): Refine pattern.
16263 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
16264 Chung-Ju Wu <jasonwucj@gmail.com>
16266 * config/nds32/iterators.md (shift_rotate): New code iterator.
16267 (shift): New code attribute.
16268 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
16269 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
16270 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
16271 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
16272 bit-wise operations.
16273 (andsi3, *andsi3): Ditto.
16274 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
16275 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
16276 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
16277 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
16278 nds32_ior_operand, nds32_xor_operand): New predicates.
16280 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16282 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
16283 (addsi3, subsi3): ... this.
16285 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16287 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
16289 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16291 * config/nds32/nds32.md: Adjust indention.
16293 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
16295 * config/nds32/nds32.md (feature): New attribute.
16297 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16299 * config/nds32/nds32.md (subtype): New attribute.
16301 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
16304 * config/arm/arm-builtins.c (arm_expand_builtin): Change
16305 expansion to perform a bitwise AND of the argument followed by a
16306 boolean negation of the result.
16308 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
16310 PR rtl-optimization/84878
16311 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
16312 the basic block. Assert the use reference is not artificial and that
16313 it has an associated insn.
16315 2018-04-04 Michael Matz <matz@suse.de>
16317 * builtins.c (compute_objsize): Pass correct operand
16318 to array_at_struct_end_p.
16320 2018-04-04 Richard Biener <rguenther@suse.de>
16323 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
16324 from contexts for DINFO_LEVEL_TERSE and below.
16326 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
16328 * config/nds32/nds32-doubleword.md (move_<mode>): Require
16329 resiter_operand condition.
16330 * config/nds32/nds32.md (*move<mode>): Ditto.
16332 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
16333 Monk Chiang <sh.chiang04@gmail.com>
16335 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
16337 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
16339 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
16341 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
16342 Kito Cheng <kito.cheng@gmail.com>
16344 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
16345 nds32_cond_code_str, output_cond_branch,
16346 output_cond_branch_compare_zero, nds32_expand_cbranch,
16347 nds32_expand_cstore, nds32_expand_movcc,
16348 nds32_output_cbranchsi4_equality_zero,
16349 nds32_output_cbranchsi4_equality_reg,
16350 nds32_output_cbranchsi4_equality_reg_or_const_int,
16351 nds32_output_cbranchsi4_greater_less_zero: New functions.
16352 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
16353 nds32_expand_cstore, nds32_expand_movcc,
16354 nds32_output_cbranchsi4_equality_zero,
16355 nds32_output_cbranchsi4_equality_reg,
16356 nds32_output_cbranchsi4_equality_reg_or_const_int,
16357 nds32_output_cbranchsi4_greater_less_zero): Declare.
16358 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
16359 nds32_rimm11s_operand): New predicates.
16360 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
16361 * config/nds32/nds32.md: Rewrite all the branch and conditional move
16364 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
16366 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
16367 * config/nds32/nds32.md: Ditto.
16368 * config/nds32/pipelines.md: Ditto.
16370 2018-04-04 Richard Biener <rguenther@suse.de>
16372 PR tree-optimization/85168
16373 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
16374 propagating abnormals.
16376 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
16378 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
16380 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
16381 Kito Cheng <kito.cheng@gmail.com>
16383 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
16384 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
16385 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
16386 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
16387 * config/nds32/nds32.md (sibcall_internal): New.
16388 (sibcall_register): Remove.
16389 (sibcall_immediate): Remove.
16390 (sibcall_value_internal): New.
16391 (sibcall_value_register): Remove.
16392 (sibcall_value_immediate): Remove.
16393 * config/nds32/predicates.md (nds32_general_register_operand): New.
16394 (nds32_call_address_operand): New.
16396 2018-04-03 Jakub Jelinek <jakub@redhat.com>
16398 PR rtl-optimization/85167
16399 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
16400 bb_defs if *split_p, instead preinitialize it to NULL.
16402 PR tree-optimization/85156
16403 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
16404 evaluating the argument multiple times.
16406 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
16408 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
16410 (_mm_cvtpd_ps): Likewise.
16411 (_mm_cvttpd_epi32): Likewise.
16412 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
16413 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
16414 vector, pixel, and bool following altivec.h include.
16416 2018-04-03 Martin Sebor <msebor@redhat.com>
16418 * doc/extend.texi (Common Function Attributes): Clarify.
16419 (const attribute): Likewise.
16420 (pure attribute): Likewise.
16422 2018-04-03 Jakub Jelinek <jakub@redhat.com>
16425 * config/i386/i386.c (ix86_expand_vector_set): Use
16426 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
16428 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
16430 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
16431 instructions when changing rounding bits to preserve precision bits
16432 in the x87 control word.
16434 2018-04-03 Martin Liska <mliska@suse.cz>
16436 PR tree-optimization/82491
16437 * rtl.h (strip_offset_and_add): Replace += suboffset with
16438 poly_uint64 () + suboffset.
16440 2018-03-29 Martin Liska <mliska@suse.cz>
16441 Martin Jambor <mjambor@suse.cz>
16444 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
16445 param_type is not an integral or pointer type.
16447 2018-04-03 Richard Biener <rguenther@suse.de>
16449 * sese.h (recompute_all_dominators): Remove.
16451 2018-04-02 Martin Sebor <msebor@redhat.com>
16453 * doc/invoke.texi (-Wrestrict): Fix typos.
16455 2018-04-02 Jim Wilson <jimw@sifive.com>
16457 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
16458 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
16459 (<optab>di3, <optab>si3_extend): Likewise.
16460 (<optab>si3_mask, <optab>si3_mask_1): New.
16461 (<optab>di3_mask, <optab>di3_mask_1): New.
16462 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
16463 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
16464 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
16466 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
16468 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
16471 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
16473 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
16474 (nds32_canonicalize_comparison): New function.
16476 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
16477 Kito Cheng <kito.cheng@gmail.com>
16478 Kuan-Lin Chen <kuanlinchentw@gmail.com>
16480 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
16481 * config/nds32/constants.md (unspec_volatile_element): Add
16482 UNSPEC_VOLATILE_RELAX_GROUP.
16483 * config/nds32/nds32-relax-opt.c: New file.
16484 * config/nds32/nds32-predicates.c
16485 (nds32_symbol_load_store_p): New function.
16486 * config/nds32/nds32-protos.h
16487 (nds32_symbol_load_store_p): Declare function.
16488 (make_pass_nds32_relax_opt): Declare new rtl pass function.
16489 * config/nds32/nds32.c
16490 (nds32_register_pass): New function to register pass.
16491 (nds32_register_passes): New function to register passes.
16492 * config/nds32/nds32.md (relax_group): New pattern.
16493 * config/nds32/nds32.opt (mrelax-hint): New option.
16494 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
16496 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
16498 * config/nds32/t-nds32: Modify files dependency.
16500 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
16502 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
16503 (PROFILE_HOOK): Define its implementation.
16505 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
16507 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
16508 type and 32-bit size.
16510 2018-04-01 Jakub Jelinek <jakub@redhat.com>
16512 PR middle-end/85090
16513 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
16514 (V_128_256): New mode iterator.
16515 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
16516 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
16517 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
16519 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
16520 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
16522 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
16525 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
16526 NaN inputs correctly.
16528 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
16531 * config/rs6000/vsx.md (??r): New mode attribute.
16532 (*vsx_mov<mode>_64bit): Use it.
16533 (*vsx_mov<mode>_32bit): Likewise.
16535 2018-03-30 Martin Sebor <msebor@redhat.com>
16537 PR tree-optimization/84818
16538 * builtins.c (check_access): Use warning_n.
16540 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16543 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
16545 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
16548 2018-03-30 Julia Koval <julia.koval@intel.com>
16551 * x86-tune.def (movx, partial_reg_dependency): Enable for
16554 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
16556 PR inline-asm/84985
16557 * lra-constraints.c (process_alt_operands): Move setting
16558 this_alternative_matches below.
16560 2018-03-29 Martin Liska <mliska@suse.cz>
16563 * doc/invoke.texi: Document how LTO works with debug info.
16564 Describe auto-load support of binutils. Mention 'x86-64'
16565 as valid option value of -march option.
16567 2018-03-29 Jakub Jelinek <jakub@redhat.com>
16569 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
16572 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
16573 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
16574 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
16577 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
16580 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
16581 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
16582 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
16583 for RS6000_BTM_POWERPC64.
16584 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
16585 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
16586 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
16589 (DIVDEU): Likewise.
16591 2018-03-28 Carl Love <cel@us.ibm.com>
16594 2017-09-27 Carl Love <cel@us.ibm.com>
16596 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
16597 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
16598 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
16601 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16603 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
16604 instead of __vector bool.
16605 (_mm_max_pu8): Likewise.
16606 (_mm_min_pi16): Likewise.
16608 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
16611 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
16612 (DIVWEUO): Likewise.
16613 (DIVDEO): Likewise.
16614 (DIVDEUO): Likewise.
16615 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
16616 DIVWEUO and DIVDEUO.
16617 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
16618 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
16619 (div_extend): Likewise.
16620 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
16622 (__builtin_divweuo): Likewise.
16623 (__builtin_divdeo): Likewise.
16624 (__builtin_divdeuo): Likewise.
16626 2018-03-28 Jakub Jelinek <jakub@redhat.com>
16629 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
16630 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
16632 PR tree-optimization/82004
16633 * gimple-match-head.c (optimize_pow_to_exp): New function.
16634 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
16635 Don't fold to exp if optimize_pow_to_exp is false.
16637 2018-03-28 Martin Liska <mliska@suse.cz>
16640 * calls.c (initialize_argument_information): Fix trailing space.
16641 * common.opt: Fix typo and provide better explanation for
16642 -fsanitize-coverage option.
16643 * config/i386/i386.opt: Fix typo.
16645 2018-03-28 Jakub Jelinek <jakub@redhat.com>
16646 Martin Liska <mliska@suse.cz>
16649 * gimplify.c (asan_poison_variable): Don't do the check for
16650 gimplify_omp_ctxp here.
16651 (gimplify_decl_expr): Do it here.
16652 (gimplify_target_expr): Likewise.
16654 2018-03-28 Martin Liska <mliska@suse.cz>
16657 * config/i386/i386.c (ix86_function_arg_advance): Do not call
16658 chkp_type_bounds_count if MPX is not enabled.
16660 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
16662 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
16664 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
16667 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
16668 function to create the function decl for complex long double
16669 multiply and divide for -mabi=ieeelongdouble.
16670 (init_float128_ieee): Call it.
16672 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
16675 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
16676 -fcf-protection=branch -mibt.
16677 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
16679 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
16682 * config/arm/arm.c (arm_valid_symbolic_address): Handle
16683 arm_word_relocations.
16685 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
16688 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
16689 extern array declarations.
16691 2018-03-27 Richard Biener <rguenther@suse.de>
16693 PR middle-end/84067
16694 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
16695 explicit single_use checks.
16697 2018-03-27 Richard Biener <rguenther@suse.de>
16699 PR tree-optimization/85082
16700 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
16703 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
16705 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
16706 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
16707 Turn on fasynchronous-unwind-tables and funwind-tables.
16709 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
16712 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
16713 (*bmi_blsr_<mode>_ccz): Ditto.
16715 2018-03-26 Tom de Vries <tom@codesourcery.com>
16717 PR tree-optimization/85063
16718 * omp-general.c (offloading_function_p): New function. Factor out
16720 * omp-offload.c (pass_omp_target_link::gate): ... here.
16721 * omp-general.h (offloading_function_p): Declare.
16722 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
16723 with attribute omp declare target for offloading functions.
16725 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
16727 PR tree-optimization/84005
16728 * tree-data-ref.h (get_base_for_alignment): Declare.
16729 * tree-data-ref.c (get_base_for_alignment_1): New function.
16730 (get_base_for_alignment): Likewise.
16731 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
16732 get_base_for_alignment to find a suitable base object, instead
16733 of always using drb->base_address.
16735 2018-03-23 Jakub Jelinek <jakub@redhat.com>
16737 PR inline-asm/85022
16738 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
16739 known size by default.
16741 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
16743 PR inline-asm/85030
16744 * lra-constraints.c (process_alt_operands): Don't match BLKmode
16745 and non BLKmode operands.
16747 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16750 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
16751 Clean up attributes.
16753 2018-03-23 Richard Biener <rguenther@suse.de>
16756 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
16757 we are going to emit early debug for LTO.
16759 2018-03-23 Jakub Jelinek <jakub@redhat.com>
16761 PR inline-asm/85034
16762 * function.c (match_asm_constraints_1): Don't optimize if input
16763 doesn't satisfy general_operand predicate for output's mode.
16765 PR inline-asm/85022
16766 * alias.c (write_dependence_p): Don't require for x_canonicalized
16767 non-VOIDmode if x has VOIDmode.
16770 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
16771 just don't try to optimize it rather than assert it never happens.
16773 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
16775 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
16776 macro expansions for definition of ST_INTERNAL_<mode> and
16777 LD_INTERNAL_<mode> builtins.
16778 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
16780 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
16782 (altivec_expand_st_builtin): Likewise.
16783 (altivec_expand_builtin): Remove calls to deleted functions.
16784 (rs6000_address_for_altivec): Delete this function.
16785 * config/rs6000/vector.md: Remove expands for
16786 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
16788 2018-03-22 Sudakshina Das <sudi.das@arm.com>
16791 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
16792 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
16793 re-computing once computed.
16794 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
16795 (arm_init_machine_status): Initialize
16796 machine->static_chain_stack_bytes.
16798 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
16801 * doc/extend.texi: Add four new prototypes for vec_ld.
16802 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
16803 definitions for more logical presentation.
16804 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
16805 entries for V1TI variants of __builtin_altivec_ld builtin.
16806 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
16807 handling of V1TI variant of LVX icode pattern.
16808 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
16809 (rs6000_gimple_fold_builtin): Likewise.
16810 (altivec_init_builtins): Add code to define
16811 __builtin_altivec_lvx_v1ti function.
16813 2018-03-22 Jakub Jelinek <jakub@redhat.com>
16815 PR inline-asm/84941
16816 * function.c (match_asm_constraints_1): Don't do the optimization
16817 if input isn't a REG, SUBREG, MEM or constant.
16819 2018-03-22 Tom de Vries <tom@codesourcery.com>
16821 PR tree-optimization/84956
16822 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
16823 bb_has_abnormal_pred.
16825 2018-03-22 Jakub Jelinek <jakub@redhat.com>
16828 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
16829 DECL_INITIAL (decl) to decl at the end.
16830 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
16831 adjust the comment.
16833 2018-03-21 Joseph Myers <joseph@codesourcery.com>
16835 * doc/extend.texi (__builtin_tgmath): Document when complex
16836 integer types are treated as _Complex _Float64.
16838 2018-03-21 Tom de Vries <tom@codesourcery.com>
16840 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
16842 2018-03-21 Jakub Jelinek <jakub@redhat.com>
16844 PR tree-optimization/84960
16845 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
16846 if it is ENTRY block, move them into single succ of ENTRY in that case.
16848 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
16850 PR tree-optimization/84811
16851 * poly-int.h (poly_span_traits): Remove the T3 parameter and
16852 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
16853 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
16854 (known_subrange_p): Update accordingly. Cast each value involved
16855 in the size comparison, rather than casting the result of the
16858 2018-03-21 Jakub Jelinek <jakub@redhat.com>
16860 PR tree-optimization/84982
16861 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
16862 by flipping the least significant bit rather than all bits from
16863 bitpos to bitpos + bitsize - 1.
16865 2018-03-21 Nathan Sidwell <nathan@acm.org>
16867 * doc/extend.texi (Deprecated Features): Remove mention of
16868 long-deleted deprecations.
16870 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16873 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
16874 * configure: Regenerate.
16876 2018-03-21 Tom de Vries <tom@codesourcery.com>
16878 PR tree-optimization/83126
16879 * tree-parloops.c (num_phis): New function.
16880 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
16882 2018-03-21 Nathan Sidwell <nathan@acm.org>
16884 * doc/extend.texi (Deprecated Features): Update deprecated flags,
16885 mention anon-struct/union members and trailing attributes.
16887 2018-03-21 Bin Cheng <bin.cheng@arm.com>
16889 PR tree-optimization/84969
16890 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
16891 builtin memset partitions if they set different rhs values.
16893 2018-03-21 Jakub Jelinek <jakub@redhat.com>
16895 PR rtl-optimization/84989
16896 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
16897 VEC_DUPLICATE with scalar result mode.
16899 2018-03-21 Martin Liska <mliska@suse.cz>
16902 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
16903 not intended return statement.
16905 2018-03-21 Martin Liska <mliska@suse.cz>
16908 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
16909 (chkp_find_bound_slots_1): Limit number of iterations.
16911 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
16914 * Minor grammar fixes for x86 options.
16916 2018-03-20 Jakub Jelinek <jakub@redhat.com>
16919 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
16920 holding REG_CFA_RESTORE notes, instead turn them into a USE.
16922 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
16925 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
16926 (altivec_lvx_<mode>_1op): Likewise.
16927 (altivec_stvx_<mode>_2op): Likewise.
16928 (altivec_stvx_<mode>_1op): Likewise.
16929 (altivec_lvx_<VM2:mode>): New define_expand.
16930 (altivec_stvx_<VM2:mode>): Likewise.
16931 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
16932 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
16933 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
16934 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
16935 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
16936 (rs6000_gen_lvx): Likewise.
16937 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
16938 (altivec_expand_stv_builtin): Likewise.
16939 (altivec_expand_builtin): Likewise.
16940 * config/rs6000/vector.md: Likewise.
16942 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16945 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
16948 2018-03-20 Richard Biener <rguenther@suse.de>
16951 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
16952 sign-conversions as zero, fall back to standard scalar_stmt
16955 2018-03-20 Martin Liska <mliska@suse.cz>
16958 * predict.c (rebuild_frequencies): Handle case when we have
16959 PROFILE_ABSENT, but flag_guess_branch_prob is false.
16961 2018-03-20 Jakub Jelinek <jakub@redhat.com>
16964 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
16965 flag_section_anchors.
16966 * varasm.c (use_blocks_for_decl_p): Remove hack for
16967 dw2_force_const_mem.
16970 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
16972 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
16973 be created, use lowpart_subreg of operands[0] rather than operands[0]
16975 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
16976 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
16977 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
16978 and n constraint instead of aarch64_shift_imm_di and Usd.
16979 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
16980 (*aarch64_<optab>_reg_minus<mode>3): ... this.
16982 2018-03-20 Sudakshina Das <sudi.das@arm.com>
16985 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
16986 to favor GPR over NEON registers.
16987 (<shift>di3_neon): Likewise.
16989 2018-03-20 Tom de Vries <tom@codesourcery.com>
16992 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
16993 (nvptx_process_pars): Emit bar.sync asap and alap.
16995 2018-03-20 Tom de Vries <tom@codesourcery.com>
16998 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
16999 seen_label if seen_label is already set.
17001 2018-03-20 Jakub Jelinek <jakub@redhat.com>
17004 * config/i386/i386.c (fold_builtin_cpu): For features above 31
17005 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
17006 Use 1U instead of 1. Formatting fixes.
17009 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
17010 instead of TREE_TYPE (s1) for the return value.
17012 2018-03-19 Jakub Jelinek <jakub@redhat.com>
17014 PR tree-optimization/84946
17015 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
17016 bitsize + bitsize in poly_uint64 rather than poly_int64.
17019 * dwarf2asm.c: Include fold-const.c.
17020 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
17021 of decl rather than decl itself.
17023 PR rtl-optimization/84643
17024 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
17026 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
17029 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
17030 calling assemble_variable.
17032 2018-03-19 Sudakshina Das <sudi.das@arm.com>
17035 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
17036 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
17038 2018-03-19 Jim Wilson <jimw@sifive.com>
17041 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
17042 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
17043 (riscv_first_stack_step): Likewise.
17044 (riscv_option_override): Use STACK_BOUNDARY instead of
17045 MIN_STACK_BOUNDARY.
17046 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
17047 MIN_STACK_BOUNDARY.
17048 (BIGGEST_ALIGNMENT): Set to 128.
17049 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
17050 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
17053 2018-03-19 Richard Biener <rguenther@suse.de>
17055 PR tree-optimization/84933
17056 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
17057 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
17059 2018-03-19 Richard Biener <rguenther@suse.de>
17061 PR tree-optimization/84859
17062 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
17063 (cond_if_else_store_replacement): Perform sinking operation on
17064 single-store BBs regardless of MAX_STORES_TO_SINK setting.
17065 Generalize what a BB with a single eligible store is.
17067 2018-03-19 Richard Biener <rguenther@suse.de>
17069 PR tree-optimization/84929
17070 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
17071 chrec_is_positive against non-chrec arg.
17073 2018-03-19 Tamar Christina <tamar.christina@arm.com>
17076 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
17078 2018-03-18 Martin Liska <mliska@suse.cz>
17080 PR rtl-optimization/84635
17081 * regrename.c (build_def_use): Use matches_mode only when
17084 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
17086 PR tree-optimization/84913
17087 * tree-vect-loop.c (vectorizable_reduction): Don't try to
17088 vectorize chains of COND_EXPRs.
17090 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
17092 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
17094 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
17096 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
17098 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
17100 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
17102 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
17103 Kito Cheng <kito.cheng@gmail.com>
17105 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
17106 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
17107 (nds32_adjust_reg_alloc_order): New function.
17108 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
17110 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
17112 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
17113 nds32_print_operand, nds32_print_operand_address): Use
17114 HOST_WIDE_INT_PRINT_DEC instead.
17116 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
17118 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
17120 2018-03-17 Jakub Jelinek <jakub@redhat.com>
17123 * config/i386/i386.c (initial_ix86_tune_features,
17124 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
17125 unsigned long long.
17126 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
17127 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
17128 rather than 1u << ix86_tune. Formatting fix.
17129 (ix86_option_override_internal): Change ix86_arch_mask from
17130 unsigned int to unsigned HOST_WIDE_INT, initialize to
17131 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
17132 (ix86_function_specific_restore): Likewise.
17134 2018-03-16 Jakub Jelinek <jakub@redhat.com>
17137 * postreload.c (reload_combine_recognize_pattern): Perform
17138 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
17139 truncate_int_for_mode the result for the destination's mode.
17142 * hsa-gen.c (mem_type_for_type): Fix comment typo.
17143 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
17145 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
17148 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
17151 * lra-assigns.c (lra_split_hard_reg_for): Don't use
17152 regno_allocno_class_array and sorted_pseudos.
17153 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
17154 insns where regno is used.
17156 2018-03-16 Martin Liska <mliska@suse.cz>
17159 * multiple_target.c (create_dispatcher_calls): Redirect
17160 reference in the symbol table.
17162 2018-03-16 Martin Liska <mliska@suse.cz>
17165 * multiple_target.c (create_dispatcher_calls): Redirect also
17168 2018-03-16 Jakub Jelinek <jakub@redhat.com>
17172 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
17173 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
17174 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
17176 2018-03-16 Julia Koval <julia.koval@intel.com>
17178 * doc/invoke.texi (Skylake Server): Add CLWB.
17179 Cannonlake): Remove CLWB.
17181 2018-03-16 Jakub Jelinek <jakub@redhat.com>
17183 PR tree-optimization/84841
17184 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
17186 (FLOAT_ONE_CONST_TYPE): Define.
17187 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
17188 (sort_by_operand_rank): Put entries with higher constant_type last
17189 rather than first to match comments.
17191 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
17193 * config/nios2/nios2.md (movsi_internal): Fix thinko in
17196 2018-03-15 Jakub Jelinek <jakub@redhat.com>
17199 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
17200 check and use address of target always.
17202 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
17205 * config/i386/i386.c (indirect_thunk_needed): Update comments.
17206 (indirect_thunk_bnd_needed): Likewise.
17207 (indirect_thunks_used): Likewise.
17208 (indirect_thunks_bnd_used): Likewise.
17209 (indirect_return_needed): New.
17210 (indirect_return_bnd_needed): Likewise.
17211 (output_indirect_thunk_function): Add a bool argument for
17213 (output_indirect_thunk_function): Don't generate alias for
17214 function return thunk.
17215 (ix86_code_end): Call output_indirect_thunk_function to generate
17216 function return thunks.
17217 (ix86_output_function_return): Set indirect_return_bnd_needed
17218 and indirect_return_needed instead of indirect_thunk_bnd_needed
17219 and indirect_thunk_needed.
17221 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
17223 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
17224 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
17225 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
17227 2018-03-15 David Malcolm <dmalcolm@redhat.com>
17228 Paul Hua <paul.hua.gm@gmail.com>
17231 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
17233 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
17235 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
17236 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
17237 resp. SFmode cases.
17239 2018-03-15 Tamar Christina <tamar.christina@arm.com>
17242 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
17243 instead of GET_MODE_SIZE when comparing Units.
17245 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
17248 * varasm.c (hash_section): Return an unchangeble hash value
17249 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
17250 Return !aarch64_can_use_per_function_literal_pools_p ().
17252 2018-03-15 Jakub Jelinek <jakub@redhat.com>
17255 * optabs.c (emit_conditional_move): Pass address of cmode's copy
17256 rather than address of cmode as last argument to prepare_cmp_insn.
17258 2018-03-15 Julia Koval <julia.koval@intel.com>
17260 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
17261 F_AVX512VNNI, F_AVX512BITALG): New.
17263 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
17266 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
17267 insn for floating-point loads and stores.
17269 2018-03-14 Carl Love <cel@us.ibm.com>
17271 * config/rs6000/rs6000-c.c: Add macro definitions for
17272 ALTIVEC_BUILTIN_VEC_PERMXOR.
17273 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
17274 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
17275 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
17276 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
17278 * config/doc/extend.texi: Add prototypes for vec_permxor.
17280 2018-03-14 David Malcolm <dmalcolm@redhat.com>
17283 * diagnostic-show-locus.c (class layout_point): Convert m_line
17284 from int to linenum_type.
17285 (line_span::comparator): Use linenum "compare" function when
17286 comparing line numbers.
17287 (test_line_span): New function.
17288 (layout_range::contains_point): Convert param "row" from int to
17290 (layout_range::intersects_line_p): Likewise.
17291 (layout::will_show_line_p): Likewise.
17292 (layout::print_source_line): Likewise.
17293 (layout::should_print_annotation_line_p): Likewise.
17294 (layout::print_annotation_line): Likewise.
17295 (layout::print_leading_fixits): Likewise.
17296 (layout::annotation_line_showed_range_p): Likewise.
17297 (struct line_corrections): Likewise for field m_row.
17298 (line_corrections::line_corrections): Likewise for param "row".
17299 (layout::print_trailing_fixits): Likewise.
17300 (layout::get_state_at_point): Likewise.
17301 (layout::get_x_bound_for_row): Likewise.
17302 (layout::print_line): Likewise.
17303 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
17304 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
17305 * input.c (selftest::test_linenum_comparisons): New function.
17306 (selftest::input_c_tests): Call it.
17307 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
17308 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
17309 * selftest.h (ASSERT_GT): New macro.
17310 (ASSERT_GT_AT): New macro.
17311 (ASSERT_LT): New macro.
17312 (ASSERT_LT_AT): New macro.
17314 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
17316 PR rtl-optimization/84780
17317 * combine.c (distribute_links): Don't make a link based on pc_rtx.
17319 2018-03-14 Martin Liska <mliska@suse.cz>
17321 * tree.c (record_node_allocation_statistics): Use
17322 get_stats_node_kind.
17323 (get_stats_node_kind): New function extracted from
17324 record_node_allocation_statistics.
17325 (free_node): Use get_stats_node_kind.
17327 2018-03-14 Richard Biener <rguenther@suse.de>
17329 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
17330 that the value-set of ANTIC_IN doesn't grow.
17333 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
17335 (BB_VISITED_WITH_VISITED_SUCCS): New define.
17336 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
17338 2018-03-14 Julia Koval <julia.koval@intel.com>
17340 * config.gcc (icelake-client, icelake-server): New.
17342 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
17343 (initial_ix86_arch_features): Ditto.
17344 (PTA_SKYLAKE): Add SGX.
17345 (PTA_ICELAKE): Remove.
17346 (PTA_ICELAKE_CLIENT): New.
17347 (PTA_ICELAKE_SERVER): New.
17348 (ix86_option_override_internal): Split up icelake on icelake client and
17350 (get_builtin_code_for_version): Ditto.
17351 (fold_builtin_cpu): Ditto.
17352 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
17353 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
17354 * config/i386/i386.h (processor_type): Ditto.
17355 * doc/invoke.texi: Ditto.
17357 2018-03-14 Jakub Jelinek <jakub@redhat.com>
17360 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
17361 INTEGER_CST offset, add it together with bitpos / 8 and
17362 sign extend based on POINTER_SIZE.
17366 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
17369 * config/i386/constraints.md (Yc): New register constraint.
17370 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
17371 Use Yc constraint for alternative 2 of operand 0. Remove
17372 preferred_for_speed attribute.
17374 2018-03-14 Richard Biener <rguenther@suse.de>
17376 PR tree-optimization/84830
17377 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
17378 with the old one to avoid oscillations.
17380 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
17383 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
17385 (assign_by_spills): Return a flag of reload assignment failure.
17386 Do not process the reload assignment failures. Do not spill other
17387 reload pseudos if they has the same reg class. Update n if
17389 (lra_assign): Add a return arg. Set up from the result of
17390 assign_by_spills call.
17391 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17392 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
17393 usage_insns if it is not NULL.
17394 (spill_hard_reg_in_range): New function.
17395 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17396 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17397 function prototypes.
17398 (lra_assign): Change prototype.
17399 * lra.c (lra): Add code to deal with fails by splitting hard reg
17402 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
17404 * config/riscv/riscv.opt (mrelax): New option.
17405 * config/riscv/riscv.c (riscv_file_start): Emit ".option
17406 "norelax" when riscv_mrelax is disabled.
17407 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
17409 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17412 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
17413 reassociation for int modes.
17415 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
17417 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
17418 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
17420 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
17421 * config/aarch64/aarch64-sve.md
17422 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
17423 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
17424 (*extend<mode><Vwide>2): Rename to...
17425 (aarch64_sve_extend<mode><Vwide>2): ...this.
17426 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
17427 renaming the old pattern to...
17428 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
17430 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
17431 define_expand, renaming the old pattern to...
17432 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
17433 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
17434 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
17435 account when deciding which SVE instruction the optab should use.
17436 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
17438 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
17440 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
17441 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
17442 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
17443 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
17444 (tlsdesc_small_<mode>): Turn a define_expand and use
17445 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
17446 (tlsdesc_small_advsimd_<mode>): ...this.
17447 (tlsdesc_small_sve_<mode>): New pattern.
17449 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
17451 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
17452 (UNSPEC_UMUL_HIGHPART): New constants.
17453 (MUL_HIGHPART): New int iteraor.
17454 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
17455 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
17457 (*<su>mul<mode>3_highpart): New define_insn.
17459 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
17462 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
17465 2018-03-13 Martin Liska <mliska@suse.cz>
17468 * (sem_item_optimizer::sem_item_optimizer): Initialize new
17470 (sem_item_optimizer::~sem_item_optimizer): Release it.
17471 (sem_item_optimizer::merge_classes): Register variable aliases.
17472 (sem_item_optimizer::fixup_pt_set): New function.
17473 (sem_item_optimizer::fixup_points_to_sets): Likewise.
17474 * ipa-icf.h: Declare new variables and functions.
17476 2018-03-13 Jakub Jelinek <jakub@redhat.com>
17478 PR middle-end/84834
17479 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
17480 integer_pow2p@2 and test integer_pow2p in condition.
17481 (A < 0 ? C : 0): Similarly for @1.
17483 PR middle-end/84831
17484 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
17485 characters starting at p contain '\0' character, don't look beyond
17489 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
17490 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
17493 * reg-stack.c (change_stack): Change update_end var from int to
17494 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
17495 also call set_block_for_insn on the newly added insns and rescan.
17498 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
17499 on the last operand.
17502 * tree.c (stabilize_reference_1): Return save_expr (e) for
17503 STATEMENT_LIST even if it doesn't have side-effects.
17505 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
17507 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
17509 2018-03-12 Renlin Li <renlin.li@arm.com>
17511 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
17512 aarch64_output_scalar_simd_mov_immediate.
17514 2018-03-12 Martin Sebor <msebor@redhat.com>
17516 PR tree-optimization/83456
17517 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
17518 for perfectly overlapping calls to memcpy.
17519 (gimple_fold_builtin_memory_chk): Same.
17520 (gimple_fold_builtin_strcpy): Handle no-warning.
17521 (gimple_fold_builtin_stxcpy_chk): Same.
17522 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
17524 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
17526 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
17527 parameter. Use it for SFmode.
17528 (rs6000_function_arg_advance_1): Adjust.
17529 (rs6000_function_arg): Adjust.
17530 (rs6000_gimplify_va_arg): Pass false for that new parameter.
17532 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
17534 PR rtl-optimization/84169
17535 PR rtl-optimization/84780
17536 * combine.c (can_combine_p): Check for a 2-insn combination whether
17537 the destination register is used between the two insns, too.
17539 2018-03-12 Richard Biener <rguenther@suse.de>
17541 PR tree-optimization/84803
17542 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
17543 for refs DR analysis didn't process.
17545 2018-03-12 Richard Biener <rguenther@suse.de>
17547 PR tree-optimization/84777
17548 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
17549 force-vectorize loops ignore whether we are optimizing for size.
17551 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
17553 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
17554 (TARGET_MD_ASM_ADJUST): Define.
17556 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
17557 Kito Cheng <kito.cheng@gmail.com>
17558 Chung-Ju Wu <jasonwucj@gmail.com>
17560 * config/nds32/nds32.c (nds32_compute_stack_frame,
17561 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
17562 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
17563 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
17564 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
17565 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
17566 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
17567 * config/nds32/nds32.md (prologue, epilogue): Use macro
17568 NDS32_V3PUSH_AVAILABLE_P to do checking.
17570 2018-03-11 Jakub Jelinek <jakub@redhat.com>
17573 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
17574 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
17575 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
17576 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
17577 addition of most attributes on !orig_type_die or the attribute not
17578 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
17580 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
17581 Chung-Ju Wu <jasonwucj@gmail.com>
17583 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
17584 __NDS32_VH__ macro.
17585 * config/nds32/nds32.opt (mvh): New option.
17587 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
17588 Chung-Ju Wu <jasonwucj@gmail.com>
17590 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
17592 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
17593 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
17596 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
17597 Chung-Ju Wu <jasonwucj@gmail.com>
17599 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
17601 * config/nds32/nds32-multiple.md (strlensi): New pattern.
17602 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
17604 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
17605 Kito Cheng <kito.cheng@gmail.com>
17606 Chung-Ju Wu <jasonwucj@gmail.com>
17608 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
17609 UNSPEC_FFMISM and UNSPEC_FLMISM.
17610 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
17611 for ffb, ffmism and flmism.
17612 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
17613 (unspec_ffmism): Ditto.
17614 (unspec_flmism): Ditto.
17615 (nds32_expand_builtin_impl): Check if string extension is available.
17616 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
17617 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
17619 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
17622 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
17625 * lra-assigns.c (assign_by_spills): Return a flag of reload
17626 assignment failure. Do not process the reload assignment
17627 failures. Do not spill other reload pseudos if they has the same
17629 (lra_assign): Add a return arg. Set up from the result of
17630 assign_by_spills call.
17631 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17632 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
17633 usage_insns if it is not NULL.
17634 (spill_hard_reg_in_range): New function.
17635 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17636 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17637 function prototypes.
17638 (lra_assign): Change prototype.
17639 * lra.c (lra): Add code to deal with fails by splitting hard reg
17642 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
17645 * config/i386/i386.opt: Replace Enforcment with Enforcement.
17647 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
17650 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
17651 (dw_val_node): Add val_symbolic_view.
17652 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
17653 (symview_upper_bound): New.
17654 (new_line_info_table): Initialize symviews_since_reset.
17655 (dwarf2out_source_line): Count symviews_since_reset and set
17656 symview_upper_bound.
17657 (dw_val_equal_p): Handle symview.
17658 (add_AT_symview): New.
17659 (print_dw_val): Handle symview.
17660 (attr_checksum, attr_checksum_ordered): Likewise.
17661 (same_dw_val_p, size_of_die): Likewise.
17662 (value_format, output_die): Likewise.
17663 (add_high_low_attributes): Use add_AT_symview for entry_view.
17664 (dwarf2out_finish): Reset symview_upper_bound, clear
17667 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
17670 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
17671 Add strict argument and use it.
17672 (rs6000_split_multireg_move): Update for new strict argument.
17673 (mem_operand_gpr): Disallow all non-offsettable addresses.
17674 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
17676 2018-03-09 Jakub Jelinek <jakub@redhat.com>
17679 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
17680 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
17681 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
17684 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
17685 decl, use remap_type if we want to use the type.
17687 2018-03-09 Martin Sebor <msebor@redhat.com>
17689 PR tree-optimization/84526
17690 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
17692 (builtin_access::generic_overlap): Be prepared to handle non-array
17695 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
17697 PR rtl-optimization/84682
17698 * lra-constraints.c (process_address_1): Check is_address flag
17699 for address constraints.
17700 (process_alt_operands): Likewise.
17701 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
17702 preprocess_constraints.
17703 * recog.h (preprocess_constraints): Add oploc parameter.
17705 * recog.c (preprocess_constraints): Test address_operand for
17706 CT_ADDRESS constraints.
17708 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
17711 * lra-assigns.c (assign_by_spills): Return a flag of reload
17712 assignment failure. Do not process the reload assignment
17713 failures. Do not spill other reload pseudos if they has the same
17715 (lra_assign): Add a return arg. Set up from the result of
17716 assign_by_spills call.
17717 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17718 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
17719 usage_insns if it is not NULL.
17720 (spill_hard_reg_in_range): New function.
17721 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17722 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17723 function prototypes.
17724 (lra_assign): Change prototype.
17725 * lra.c (lra): Add code to deal with fails by splitting hard reg
17728 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17731 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
17732 Accept complain bool parameter. Only emit errors if it is true.
17733 (arm_parse_cpu_option_name): Likewise.
17734 (arm_target_thumb_only): Adjust callers of the above.
17735 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
17736 prototype to take a default true bool parameter.
17737 (arm_parse_arch_option_name): Likewise.
17739 2018-03-09 David Malcolm <dmalcolm@redhat.com>
17740 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
17744 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
17745 * configure: Regenerate.
17746 * configure.ac ("linker --version-script option"): New.
17747 ("linker soname option"): New.
17749 2018-03-09 Richard Biener <rguenther@suse.de>
17751 PR tree-optimization/84775
17752 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
17753 immediate uses of predicate stmts and mark them modified.
17756 PR tree-optimization/84178
17757 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
17759 (version_loop_for_if_conversion): Delay update_ssa call.
17760 (tree_if_conversion): Delay update_ssa until after predicate
17763 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
17766 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
17767 when the function accesses prior frames.
17769 2018-03-08 Jakub Jelinek <jakub@redhat.com>
17772 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
17773 gen_llsym, otherwise call maybe_gen_llsym.
17775 PR inline-asm/84742
17776 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
17777 has ',' character inside of it.
17779 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17782 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
17783 as clobbering CC_REGNUM.
17785 2018-03-08 Richard Biener <rguenther@suse.de>
17787 PR middle-end/84552
17788 * tree-scalar-evolution.c: Include tree-into-ssa.h.
17789 (follow_copies_to_constant): Do not follow SSA names registered
17792 2018-03-08 Richard Biener <rguenther@suse.de>
17794 PR tree-optimization/84178
17795 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
17797 (version_loop_for_if_conversion): Delay update_ssa call.
17798 (tree_if_conversion): Delay update_ssa until after predicate
17801 2018-03-08 David Malcolm <dmalcolm@redhat.com>
17803 PR tree-optimization/84178
17804 * tree-if-conv.c (release_bb_predicate): Remove the
17805 the assertion that the stmts have NULL use_ops.
17806 Discard the statements, asserting that they haven't
17807 yet been added to a BB.
17809 2018-03-08 Richard Biener <rguenther@suse.de>
17811 PR tree-optimization/84746
17812 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
17813 (phi_translate): Pass in destination ANTIC_OUT set.
17814 (phi_translate_1): Likewise. For a simplified result lookup
17815 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
17816 (phi_translate_set): Adjust.
17817 (do_pre_regular_insertion): Likewise.
17818 (do_pre_partial_partial_insertion): Likewise.
17820 2018-03-08 Martin Liska <mliska@suse.cz>
17822 PR gcov-profile/84735
17823 * doc/gcov.texi: Document usage of profile files.
17824 * gcov-io.h: Document changes in the format.
17826 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
17830 * dwarf2out.c (struct dw_line_info_table): Update comments for
17832 (FORCE_RESET_NEXT_VIEW): New.
17833 (FORCE_RESETTING_VIEW_P): New.
17834 (RESETTING_VIEW_P): Check for -1 too.
17835 (ZERO_VIEW_P): Likewise.
17836 (new_line_info_table): Force-reset next view.
17837 (dwarf2out_begin_function): Likewise.
17838 (dwarf2out_source_line): Simplify zero_view_p initialization.
17839 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
17840 view directly. Omit view when omitting .loc at line 0.
17842 2018-03-08 Jakub Jelinek <jakub@redhat.com>
17844 PR tree-optimization/84740
17845 * tree-switch-conversion.c (process_switch): Call build_constructors
17846 only if info.phi_count is non-zero.
17848 PR tree-optimization/84739
17849 * tree-tailcall.c (find_tail_calls): Check call arguments against
17850 DECL_ARGUMENTS (current_function_decl) rather than
17851 DECL_ARGUMENTS (func) when checking for tail recursion.
17853 2018-03-07 Jakub Jelinek <jakub@redhat.com>
17855 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
17856 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
17857 Volker Reichelt's entry and add entries for people that perform
17858 GCC fuzzy testing and report numerous bugs.
17860 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
17863 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
17864 readonly data in sdata, if that is disabled.
17865 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
17866 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
17867 -mreadonly-in-sdata option.
17869 2018-03-07 Martin Sebor <msebor@redhat.com>
17871 PR tree-optimization/84468
17872 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
17873 basic block when looking for nul assignment.
17875 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
17878 * except.h (output_function_exception_table): Adjust prototype.
17879 * except.c (output_function_exception_table): Remove FNNAME parameter
17880 and add SECTION parameter. Ouput one part of the table at a time.
17881 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
17882 the first part of the exception table and emit unwind directives.
17883 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
17884 (i386_pe_seh_cold_init): Likewise.
17885 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
17886 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
17887 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
17888 (ix86_output_call_insn): Emit a nop in one more case for SEH.
17889 * config/i386/winnt.c: Include except.h.
17890 (struct seh_frame_state): Add reg_offset, after_prologue and
17891 in_cold_section fields.
17892 (i386_pe_seh_end_prologue): Set seh->after_prologue.
17893 (i386_pe_seh_cold_init): New function.
17894 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
17895 to seh->in_cold_section.
17896 (seh_emit_push): Record the offset of the push.
17897 (seh_emit_save): Record the offet of the save.
17898 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
17899 Test seh->after_prologue to disregard the epilogue.
17900 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
17901 (i386_pe_end_cold_function): New function.
17903 2018-03-07 Jakub Jelinek <jakub@redhat.com>
17906 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
17907 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
17910 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
17912 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
17913 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
17915 PR middle-end/84723
17916 * multiple_target.c: Include tree-inline.h and intl.h.
17917 (expand_target_clones): Diagnose and fail if node->definition and
17918 !tree_versionable_function_p (node->decl).
17920 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
17922 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
17924 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
17925 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
17926 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
17928 2018-03-06 Jakub Jelinek <jakub@redhat.com>
17931 * combine.c (try_combine): Use reg_or_subregno instead of handling
17932 just paradoxical SUBREGs and REGs.
17934 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
17936 * config/arc/arc.c (arc_finalize_pic): Remove function.
17937 (arc_must_save_register): We use single base PIC register, remove
17938 checks to save/restore the PIC register.
17939 (arc_expand_prologue): Likewise.
17940 * config/arc/arc-protos.h (arc_set_default_type_attributes):
17942 (arc_verify_short): Likewise.
17943 (arc_attr_type): Likewise.
17944 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
17945 (walk_stores): Likewise.
17946 (arc_address_cost): Make it static.
17947 (arc_verify_short): Likewise.
17948 (branch_dest): Likewise.
17949 (arc_attr_type): Likewise.
17950 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
17951 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
17952 (arc_final_prescan_insn): Remove inserting the nops due to
17953 hardware hazards. It is done in reorg step.
17954 (insn_length_variant_t): Remove.
17955 (insn_length_parameters_t): Likewise.
17956 (arc_insn_length_parameters): Likewise.
17957 (arc_get_insn_variants): Likewise.
17958 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
17960 2018-03-06 Jakub Jelinek <jakub@redhat.com>
17962 PR inline-asm/84683
17963 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
17966 PR tree-optimization/84687
17967 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
17969 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
17971 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17973 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
17974 Rename to ppc_speculation_barrier.
17975 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
17976 __builtin_ppc_speculation_barrier.
17978 2018-03-05 Jakub Jelinek <jakub@redhat.com>
17981 * combine.c (combine_simplify_rtx): Don't try to simplify if
17982 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
17985 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
17987 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
17988 to 32 bytes when compiling for POWER9.
17990 2018-03-05 Jakub Jelinek <jakub@redhat.com>
17993 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
17994 regparm >= 3 with no arg reg available also for calls with
17995 flag_force_indirect_call. Pass decl to ix86_function_regparm.
17998 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
18000 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
18002 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
18005 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
18007 2018-03-05 Richard Biener <rguenther@suse.de>
18009 PR tree-optimization/84486
18010 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
18011 When inserting a __builtin_assume_aligned call set the LHS
18012 SSA name alignment info accordingly.
18014 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
18016 PR tree-optimization/84114
18017 * config/aarch64/aarch64.c (aarch64_reassociation_width)
18018 Avoid reassociation of FLOAT_MODE addition.
18020 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
18022 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
18023 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
18024 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
18025 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
18026 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
18027 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
18028 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
18030 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
18031 __builtin_ia32_wbinvd): New builtins.
18032 (SPECIAL_ARGS2): New.
18033 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
18034 (SPECIAL_ARGS2): New.
18035 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
18036 (ix86_valid_target_attribute_inner_p): Ditto.
18037 (ix86_init_mmx_sse_builtins): Add special_args2.
18038 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
18039 TARGET_WBNOINVD_P): New.
18040 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
18041 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
18042 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
18043 * config/i386/immintrin.h (_wbinvd): New intrinsic.
18044 * config/i386/pconfigintrin.h: New file.
18045 * config/i386/wbnoinvdintrin.h: Ditto.
18046 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
18048 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
18050 2018-03-05 Richard Biener <rguenther@suse.de>
18052 PR tree-optimization/84670
18053 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
18055 (BB_VISITED_WITH_VISITED_SUCCS): New define.
18056 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
18057 (compute_antic_aux): Only assert the number of values in ANTIC_IN
18058 doesn't grow if all successors (recursively) were visited at least
18061 2018-03-05 Richard Biener <rguenther@suse.de>
18063 PR tree-optimization/84650
18064 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
18065 if executed in the loop pipeline.
18067 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
18069 * doc/configfiles.texi (Configuration Files): Move info about
18070 conditionalizing $target-protos.h to...
18071 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
18072 differs from $target-protos.h.
18074 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
18075 Chung-Ju Wu <jasonwucj@gmail.com>
18077 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
18078 * config/nds32/nds32-multiple.md (setmemsi): Define.
18079 * config/nds32/nds32-memory-manipulation.c
18080 (nds32_gen_dup_4_byte_to_word_value): New.
18081 (emit_setmem_word_loop): New.
18082 (emit_setmem_byte_loop): New.
18083 (nds32_expand_setmem_loop): New.
18084 (nds32_expand_setmem_loop_v3m): New.
18085 (nds32_expand_setmem_unroll): New.
18086 (nds32_expand_setmem): New.
18088 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
18089 Chung-Ju Wu <jasonwucj@gmail.com>
18091 * config/nds32/nds32-memory-manipulation.c
18092 (nds32_emit_load_store): New.
18093 (nds32_emit_post_inc_load_store): New.
18094 (nds32_emit_mem_move): New.
18095 (nds32_emit_mem_move_block): New.
18096 (nds32_expand_movmemsi_loop_unknown_size): New.
18097 (nds32_expand_movmemsi_loop_known_size): New.
18098 (nds32_expand_movmemsi_loop): New.
18099 (nds32_expand_movmemsi_unroll): New.
18100 (nds32_expand_movmemqi): Rename ...
18101 (nds32_expand_movmemsi): ... to this.
18102 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
18103 (movmemsi): ... to this.
18104 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
18105 (nds32_expand_movmemsi): ... to this.
18107 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
18108 Monk Chiang <sh.chiang04@gmail.com>
18109 Chung-Ju Wu <jasonwucj@gmail.com>
18111 * config/nds32/nds32-protos.h
18112 (nds32_expand_load_multiple): New arguments.
18113 (nds32_expand_store_multiple): Ditto.
18114 (nds32_valid_multiple_load_store): Rename ...
18115 (nds32_valid_multiple_load_store_p): ... to this.
18116 * config/nds32/nds32-memory-manipulation.c
18117 (nds32_expand_load_multiple): Refine implementation.
18118 (nds32_expand_store_multiple): Ditto.
18119 * config/nds32/nds32-multiple.md
18120 (load_multiple): Update nds32_expand_load_multiple interface.
18121 (store_multiple): Update nds32_expand_store_multiple interface.
18122 * config/nds32/nds32-predicates.c
18123 (nds32_valid_multiple_load_store): Rename ...
18124 (nds32_valid_multiple_load_store_p): ... to this and refine
18126 * config/nds32/predicates.md
18127 (nds32_load_multiple_and_update_address_operation): New predicate.
18128 (nds32_store_multiple_and_update_address_operation): New predicate.
18130 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
18131 Chung-Ju Wu <jasonwucj@gmail.com>
18133 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
18134 (combo): New attribute.
18135 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
18137 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
18139 * config/nds32/nds32.opt: Change -mcmodel= default value.
18141 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
18142 Monk Chiang <sh.chiang04@gmail.com>
18143 Chung-Ju Wu <jasonwucj@gmail.com>
18145 * config/nds32/constants.md (unspec_element): New enum.
18146 * config/nds32/constraints.md (Umw): New constraint.
18147 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
18148 * config/nds32/nds32-intrinsic.md: Likewise.
18149 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
18150 (nds32_valid_smw_lwm_base_p): New.
18151 (nds32_output_smw_single_word): New.
18152 (nds32_output_lmw_single_word): New.
18153 (nds32_expand_unaligned_load): New.
18154 (nds32_expand_unaligned_store): New.
18155 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
18156 (nds32_output_smw_single_word): Declare.
18157 (nds32_output_lmw_single_word): Declare.
18158 (nds32_expand_unaligned_load): Declare.
18159 (nds32_expand_unaligned_store): Declare.
18160 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
18161 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
18162 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
18163 NDS32_BUILTIN_UASTORE_DW.
18164 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
18167 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
18168 Kito Cheng <kito.cheng@gmail.com>
18169 Chung-Ju Wu <jasonwucj@gmail.com>
18171 * config/nds32/nds32-intrinsic.c
18172 (nds32_expand_builtin_null_ftype_reg): Delete.
18173 (nds32_expand_builtin_reg_ftype_imm): Ditto.
18174 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
18175 (nds32_read_argument): New.
18176 (nds32_legitimize_target): Ditto.
18177 (nds32_legitimize_argument): Ditto.
18178 (nds32_check_constant_argument): Ditto.
18179 (nds32_expand_unop_builtin): Ditto.
18180 (nds32_expand_unopimm_builtin): Ditto.
18181 (nds32_expand_binop_builtin): Ditto.
18182 (nds32_builtin_decl_impl): Ditto.
18183 (builtin_description): Ditto.
18184 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
18185 (nds32_init_builtins_impl): Ditto.
18186 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
18187 (nds32_builtin_decl): New.
18188 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
18189 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
18191 2018-03-02 Jeff Law <law@redhat.com>
18193 * reorg.c (stop_search_p): Handle DEBUG_INSN.
18194 (redundant_insn, fill_simple_delay_slots): Likewise.
18195 (fill_slots_from_thread): Likewise.
18196 * resource.c (mark_referenced_resources): Likewise.
18197 (mark_set_resources, find_dead_or_set_registers): Likewise.
18199 2018-03-02 Jakub Jelinek <jakub@redhat.com>
18201 * substring-locations.h (format_warning_va): Formatting fix for
18202 ATTRIBUTE_GCC_DIAG.
18203 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
18205 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
18206 * substring-locations.c: Include intl.h.
18207 (format_warning_va): Turned into small wrapper around
18208 format_warning_n_va, renamed to ...
18209 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
18210 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
18212 (format_warning_at_substring_n): New function.
18213 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
18214 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
18215 format_warning_at_substring with just a shorter name instead of
18216 const function pointer.
18217 (fmtwarn_n): New function.
18218 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
18219 appropriate, get rid of all the fmtstr temporaries, move conditionals
18220 with G_() wrapped string literals directly into fmtwarn arguments,
18221 cast dir.len to (int), formatting fixes.
18223 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
18225 * doc/invoke.texi: Remove "Cilk Plus" references.
18227 2018-03-02 Jakub Jelinek <jakub@redhat.com>
18228 Richard Biener <rguenther@suse.de>
18231 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
18232 for error or warning attributes if CALL_FROM_THUNK_P is set.
18235 2018-03-02 Jakub Jelinek <jakub@redhat.com>
18238 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
18239 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
18242 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
18243 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
18245 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
18246 instead of -1U in last predictors element's probability member.
18248 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
18251 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
18252 arguments if they are comparable.
18254 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
18256 PR tree-optimization/84634
18257 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
18258 masks and masked_loop_p with a single loop_masks, making sure it's
18259 null for bb vectorization.
18261 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
18263 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
18264 (vect_analyze_data_ref_access): Use loop->safe_len rather than
18265 loop->force_vectorize to check whether there is no alias.
18267 2018-03-02 Jakub Jelinek <jakub@redhat.com>
18270 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
18272 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
18274 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
18275 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
18276 instead of a loop around prev_real_insn.
18277 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
18280 PR inline-asm/84625
18281 * config/i386/i386.c (ix86_print_operand): Use conditional
18282 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
18285 2018-03-02 Richard Biener <rguenther@suse.de>
18287 PR tree-optimization/84427
18288 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
18289 (bitmap_set_subtract_values): Rewrite to handle multiple
18292 (prune_clobbered_mems): Likewise.
18293 (phi_translate): Take edge instead of pred/phiblock.
18294 (phi_translate_1): Likewise.
18295 (phi_translate_set): Likewise. Insert all translated
18296 exprs for a value into the set, keeping possibly multiple
18297 expressions per value.
18298 (compute_antic_aux): Adjust for phi_translate changes.
18299 When intersecting union the expressions and prune those
18300 not in the final value set, keeping possibly multiple
18301 expressions per value. Do not use value-insertion
18302 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
18303 all expressions. Add verification that the value-sets
18304 only shrink during iteration.
18305 (compute_partial_antic_aux): Adjust for the phi_translate changes.
18306 (do_pre_regular_insertion): Likewise.
18307 (do_pre_partial_partial_insertion): Likewise.
18309 2018-03-02 Richard Biener <rguenther@suse.de>
18312 * config/darwin.c (saved_debug_info_level): New static global.
18313 (darwin_asm_lto_start): Disable debug info generation for LTO out.
18314 (darwin_asm_lto_end): Restore debug info generation settings.
18316 2018-03-01 Martin Liska <mliska@suse.cz>
18319 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
18320 volatile arguments.
18322 2018-03-01 Richard Biener <rguenther@suse.de>
18325 * dwarf2out.c (gen_variable_die): Properly handle late VLA
18326 type annotation with LTO when debug was disabled at compile-time.
18328 2018-03-01 Matthew Fortune <mfortune@gmail.com>
18330 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
18332 (mips_final_postscan_insn): Likewise.
18334 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
18336 PR rtl-optimization/84528
18337 * alias.c (init_alias_target): Add commentary.
18338 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
18339 a unique base value if the frame pointer is not eliminated
18340 to the stack pointer.
18342 2018-03-01 Tom de Vries <tom@codesourcery.com>
18344 PR rtl-optimization/83327
18345 * lra-int.h (hard_regs_spilled_into): Declare.
18346 * lra.c (hard_regs_spilled_into): Define.
18347 (init_reg_info): Init hard_regs_spilled_into.
18348 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
18349 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
18350 (process_bb_lives): Handle hard_regs_spilled_into.
18351 (lra_create_live_ranges_1): Before doing liveness propagation, clear
18352 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
18354 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
18356 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
18357 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
18358 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
18359 * config/rs6000/aix72.h: New file.
18361 2018-02-28 Jakub Jelinek <jakub@redhat.com>
18363 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
18364 instead of warning_at with conditional singular and plural messages
18368 * stor-layout.c (update_alignment_for_field): For
18369 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
18370 && !DECL_PACKED (field), do the alignment update, just use
18371 only desired_align instead of MAX (type_align, desired_align)
18373 (place_field): Don't do known_align < desired_align handling
18374 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
18375 is non-NULL, instead do it after rli->prev_field handling and
18376 only if not within a bitfield word. For DECL_PACKED (field)
18377 use type_align of BITS_PER_UNIT.
18379 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
18381 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
18382 superfluous parentheses and trailing spaces.
18384 2018-02-28 Richard Biener <rguenther@suse.de>
18386 PR tree-optimization/84584
18387 * graphite-scop-detection.c (scop_detection::add_scop): Discard
18388 SCoPs with fake exit edge.
18390 2018-02-28 Martin Liska <mliska@suse.cz>
18393 * timevar.c (timer::print): Fix format to properly print 100%
18396 2018-02-28 Richard Biener <rguenther@suse.de>
18398 PR middle-end/84607
18399 * genmatch.c (capture_info::walk_match): Do not mark
18400 captured expressions without operands as expr_p given
18401 they act more like predicates and should be subject to
18402 "lost tail" side-effect preserving.
18404 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
18406 PR rtl-optimization/81611
18407 * auto-inc-dec.c (attempt_change): Move dead note from
18408 mem_insn if it's the next use of regno
18409 (find_address): Take address use of reg holding
18410 non-incremented value. Add parm to limit search to the named
18412 (merge_in_block): Attempt to use a mem insn that is the next
18413 use of the original regno.
18415 2018-02-27 Martin Sebor <msebor@redhat.com>
18418 * doc/invoke.texi (-Wmissing-attributes): New option.
18419 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
18421 2018-02-27 Martin Sebor <msebor@redhat.com>
18423 PR translation/84207
18424 * diagnostic-core.h (warning_n, error_n, inform_n): Change
18425 n argument to unsigned HOST_WIDE_INT.
18426 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
18427 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
18428 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
18429 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
18431 2018-02-27 Richard Biener <rguenther@suse.de>
18433 PR tree-optimization/84512
18434 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
18435 Do not use the estimate returned from record_stmt_cost for
18436 the scalar iteration cost but sum properly using add_stmt_cost.
18438 2018-02-27 Richard Biener <rguenther@suse.de>
18440 PR tree-optimization/84466
18441 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
18442 Adjust last change to less strictly validate use operands.
18444 2018-02-27 Martin Liska <mliska@suse.cz>
18446 PR gcov-profile/84548
18447 * gcov.c (process_file): Allow partial overlap and consider it
18448 also as group functions.
18449 (output_lines): Properly calculate range of lines for a group.
18451 2018-02-27 Martin Liska <mliska@suse.cz>
18453 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
18454 'ggc' suffixes. Change first column width.
18455 (timer::print): Fix formatting of the column.
18457 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
18459 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
18460 preserve inline entry blocks for the sake of debug inline
18461 entry point markers alone.
18462 (remove_unused_locals): Suggest in comments a better place to
18463 force the preservation of inline entry blocks that are
18464 otherwise unused, but do not preserve them.
18466 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
18468 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
18470 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
18473 * config/i386/constraints.md (Bs): Replace
18474 ix86_indirect_branch_register with
18475 TARGET_INDIRECT_BRANCH_REGISTER.
18477 * config/i386/i386.md (indirect_jump): Likewise.
18478 (tablejump): Likewise.
18479 (*sibcall_memory): Likewise.
18480 (*sibcall_value_memory): Likewise.
18481 Peepholes of indirect call and jump via memory: Likewise.
18482 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
18483 (*sibcall_value_GOT_32): Likewise.
18484 * config/i386/predicates.md (indirect_branch_operand): Likewise.
18485 (GOT_memory_operand): Likewise.
18486 (call_insn_operand): Likewise.
18487 (sibcall_insn_operand): Likewise.
18488 (GOT32_symbol_operand): Likewise.
18489 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
18491 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
18493 PR rtl-optimization/83496
18494 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
18495 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
18496 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
18497 redundant insn, if any.
18498 (relax_delay_slots): Likewise.
18499 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
18501 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
18503 PR tree-optimization/83965
18504 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
18505 that grouped statements are part of a reduction chain. Return
18506 true if the statement is not marked as a reduction itself but
18507 is part of a group.
18508 (vect_recog_dot_prod_pattern): Don't check whether the statement
18509 is part of a group here.
18510 (vect_recog_sad_pattern): Likewise.
18511 (vect_recog_widen_sum_pattern): Likewise.
18513 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
18516 * final.c (rest_of_clean_state): Also look for calls inside sequences.
18518 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
18521 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
18523 (ix86_output_indirect_function_return): New prototype.
18524 (ix86_split_simple_return_pop_internal): Likewise.
18525 * config/i386/i386.c (indirect_return_via_cx): New.
18526 (indirect_return_via_cx_bnd): Likewise.
18527 (indirect_thunk_name): Handle return va CX_REG.
18528 (output_indirect_thunk_function): Create alias for
18529 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
18530 (ix86_output_indirect_jmp): Remove the bool argument.
18531 (ix86_output_indirect_function_return): New function.
18532 (ix86_split_simple_return_pop_internal): Likewise.
18533 * config/i386/i386.md (*indirect_jump): Don't pass false
18534 to ix86_output_indirect_jmp.
18535 (*tablejump_1): Likewise.
18536 (simple_return_pop_internal): Change it to define_insn_and_split.
18537 Call ix86_split_simple_return_pop_internal to split it for
18538 -mfunction-return=.
18539 (simple_return_indirect_internal): Call
18540 ix86_output_indirect_function_return instead of
18541 ix86_output_indirect_jmp.
18543 2018-02-26 Jakub Jelinek <jakub@redhat.com>
18546 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
18547 memset and value initialization afterwards.
18549 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
18551 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
18553 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
18556 * common/config/aarch64/aarch64-common.c
18557 (aarch_option_optimization_table[]): Switch
18558 off fomit-frame-pointer
18560 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
18561 Chung-Ju Wu <jasonwucj@gmail.com>
18563 * config/nds32/nds32-multiple.md (load_multiple): Disallow
18565 (store_multiple): Ditto.
18567 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
18569 * config.gcc: Add --with-cpu support for nds32 target.
18570 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
18571 * config/nds32/nds32.opt: Add -mcpu= option.
18573 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
18575 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
18576 isel=yes): Warn for these deprecated options.
18578 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
18580 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
18581 ISA_2_5_MASKS_EMBEDDED.
18583 2018-02-23 Jakub Jelinek <jakub@redhat.com>
18585 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
18586 p->max as pointers rather than using iterative_hash_expr.
18588 2018-02-23 Carl Love <cel@us.ibm.com>
18590 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
18591 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
18593 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
18594 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
18595 P8V_BUILTIN_VEC_VUNSIGNED2.
18597 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
18600 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
18601 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
18603 * lra-constraints.c (curr_insn_transform): Set up
18604 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
18606 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
18607 * lra-eliminations.c (spill_pseudos): Ditto.
18608 (process_insn_for_elimination): Ditto.
18609 * lra-lives.c (reg_early_clobber_p): Use the new macros.
18610 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
18611 LRA_NON_CLOBBERED_ALT.
18613 2018-02-22 Martin Sebor <msebor@redhat.com>
18615 PR tree-optimization/84480
18616 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
18617 to maybe_diag_stxncpy_trunc. Call it.
18618 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
18619 from gimple_fold_builtin_strcpy. Print inlining stack.
18620 (handle_builtin_stxncpy): Print inlining stack.
18621 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
18623 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
18626 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
18627 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
18628 and -fcheck-pointer-bounds are used together.
18629 (indirect_thunk_prefix): New enum.
18630 (indirect_thunk_need_prefix): New function.
18631 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
18632 "_nt" instead of "_bnd" for NOTRACK prefix.
18633 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
18634 (output_indirect_thunk_function): Likewise.
18636 (ix86_code_end): Update output_indirect_thunk_function calls.
18637 (ix86_output_indirect_branch_via_reg): Replace
18638 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
18639 (ix86_output_indirect_branch_via_push): Likewise.
18640 (ix86_output_function_return): Likewise.
18641 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
18642 incompatible with -fcf-protection=branch and
18643 -fcheck-pointer-bounds.
18645 2018-02-22 Steve Ellcey <sellcey@cavium.com>
18648 * config/aarch64/aarch64.c (aarch64_print_address_internal):
18649 Change gcc_assert call to output_operand_lossage.
18651 2018-02-22 Steve Ellcey <sellcey@cavium.com>
18653 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
18655 2018-02-22 DJ Delorie <dj@redhat.com>
18656 Sebastian Perta <sebastian.perta@renesas.com>
18657 Oleg Endo <olegendo@gcc.gnu.org>
18659 * config/rx/rx.c (rx_rtx_costs): New function.
18660 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
18662 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
18664 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
18666 2018-02-22 Martin Liska <mliska@suse.cz>
18669 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
18670 Add "native" as a possible value.
18672 2018-02-22 Martin Liska <mliska@suse.cz>
18675 * config/i386/i386.c (ix86_option_override_internal):
18676 Add "native" as a possible value for -march and -mtune.
18678 2018-02-22 Jakub Jelinek <jakub@redhat.com>
18681 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
18682 to all type variants.
18684 PR tree-optimization/84503
18685 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
18686 width as info->bitpos + info->bitsize - start.
18687 (merged_store_group::merge_overlapping): Simplify width computation.
18688 (check_no_overlap): New function.
18689 (imm_store_chain_info::try_coalesce_bswap): Compute expected
18690 start + width and last_order of the group, fail if check_no_overlap
18692 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
18693 to group if check_no_overlap fails.
18695 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
18697 * config/rs6000/altivec.md: Delete contraint arguments to
18698 define_expand, define_split, and define_peephole2, and in
18699 define_insn_and_split if always unused.
18700 * config/rs6000/darwin.md: Ditto.
18701 * config/rs6000/dfp.md: Ditto.
18702 * config/rs6000/rs6000.md: Ditto.
18703 * config/rs6000/sync.md: Ditto.
18704 * config/rs6000/vector.md: Ditto.
18705 * config/rs6000/vsx.md: Ditto.
18707 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
18709 * config/rs6000/altivec.md: Write output control strings as braced
18710 blocks instead of double-quoted strings.
18711 * config/rs6000/darwin.md: Ditto.
18712 * config/rs6000/rs6000.md: Ditto.
18713 * config/rs6000/vector.md: Ditto.
18714 * config/rs6000/vsx.md: Ditto.
18716 2018-02-21 Jason Merrill <jason@redhat.com>
18718 PR c++/84314 - ICE with templates and fastcall attribute.
18719 * attribs.c (build_type_attribute_qual_variant): Remove assert.
18721 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
18723 * ipa-cp.c (determine_versionability): Fix comment typos.
18725 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
18728 * ipa-cp.c (determine_versionability): Do not version functions caling
18731 2018-02-21 Martin Liska <mliska@suse.cz>
18734 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
18735 Add "native" as a possible value.
18736 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
18737 the macro when native cpu detection is available.
18739 2018-02-21 Martin Liska <mliska@suse.cz>
18742 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
18743 Add "native" as a possible value.
18744 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
18745 when native cpu detection is available.
18747 2018-02-21 Jakub Jelinek <jakub@redhat.com>
18748 Martin Sebor <msebor@redhat.com>
18750 PR tree-optimization/84478
18751 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
18753 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
18754 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
18755 support which is conservatively correct, for 2 only stay conservative
18756 for maxlen. Formatting and comment capitalization fixes. Add STRICT
18757 argument to the 2 argument get_range_strlen, adjust 6 arg
18758 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
18760 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
18761 (gimple_fold_builtin_strlen): Pass true as last argument to
18764 2018-02-20 Martin Sebor <msebor@redhat.com>
18766 PR middle-end/84095
18767 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
18768 (builtin_memref::set_base_and_offset): Same. Handle inner references.
18769 (builtin_memref::builtin_memref): Factor out parts into
18770 set_base_and_offset and call it.
18772 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
18774 PR middle-end/84406
18775 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
18776 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
18777 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
18778 search at the associated MODE_INT.
18780 2018-02-20 Jeff Law <law@redhat.com>
18782 PR middle-end/82123
18783 PR tree-optimization/81592
18784 PR middle-end/79257
18785 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
18786 for range data rather than using global data.
18787 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
18788 range data rather than using global data.
18789 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
18790 pass it to children as needed.
18791 (struct directive::fmtresult): Similarly.
18792 (struct directive::set_width): Similarly.
18793 (struct directive::set_precision): Similarly.
18794 (format_integer, format_directive, parse_directive): Similarly.
18795 (format_none): Accept unnamed vr_values parameter.
18796 (format_percent, format_floating, format_character): Similarly.
18797 (format_string, format_plain): Similarly.
18798 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
18799 the EVRP range analyzer for range data rather than using global data.
18800 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
18801 gimple-ssa-evrp-analyze.h
18802 (class sprintf_dom_walker): Add after_dom_children member function.
18803 Add evrp_range_analyzer member.
18804 (sprintf_dom_walker::before_dom_children): Call into the EVRP
18805 range analyzer as needed.
18806 (sprintf_dom_walker::after_dom_children): New member function.
18807 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
18809 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
18810 (evrp_range_analyzer::pop_to_marker): Likewise.
18812 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
18814 PR tree-optimization/84419
18815 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
18816 with the required type if its current type is compatible but
18819 2018-02-20 Jakub Jelinek <jakub@redhat.com>
18821 PR middle-end/82004
18822 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
18823 after vectorization.
18825 2018-02-20 Martin Liska <mliska@suse.cz>
18828 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
18829 possible values if we don't have a hint.
18831 2018-02-20 Martin Liska <mliska@suse.cz>
18835 * final.c (shorten_branches): Build align_tab array with one
18837 * opts.c (finish_options): Add alignment option limit check.
18838 (MAX_CODE_ALIGN): Likewise.
18839 (MAX_CODE_ALIGN_VALUE): Likewise.
18840 * doc/invoke.texi: Document maximum allowed option value for
18841 all -falign-* options.
18843 2018-02-19 Jakub Jelinek <jakub@redhat.com>
18846 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
18847 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
18848 * var-tracking.c (emit_note_insn_var_location): Remove all references
18849 to NOTE_INSN_CALL_ARG_LOCATION.
18850 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
18851 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
18852 Use copy_rtx_if_shared.
18853 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
18854 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
18855 (dwarf2out_var_location): Remove handling of
18856 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
18858 * final.c (final_scan_insn): Remove all references to
18859 NOTE_INSN_CALL_ARG_LOCATION.
18860 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
18861 before dumping final insns.
18862 * except.c (emit_note_eh_region_end): Remove all references to
18863 NOTE_INSN_CALL_ARG_LOCATION.
18864 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
18865 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
18866 * config/arc/arc.c (hwloop_optimize): Likewise.
18867 * config/arm/arm.c (create_fix_barrier): Likewise.
18868 * config/s390/s390.c (s390_chunkify_start): Likewise.
18869 * config/sh/sh.c (find_barrier): Likewise.
18870 * config/i386/i386.c (rest_of_insert_endbranch,
18871 ix86_seh_fixup_eh_fallthru): Likewise.
18872 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
18873 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
18874 * config/frv/frv.c (frv_function_prologue): Likewise.
18875 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
18877 (note_outside_basic_block_p): Remove all references to
18878 NOTE_INSN_CALL_ARG_LOCATION.
18879 * gengtype.c (adjust_field_rtx_def): Likewise.
18880 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
18882 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
18883 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
18886 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
18889 PR tree-optimization/84452
18890 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
18891 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
18894 2018-02-19 Martin Liska <mliska@suse.cz>
18897 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
18899 2018-02-19 Martin Liska <mliska@suse.cz>
18900 Richard Sandiford <richard.sandiford@linaro.org>
18902 PR tree-optimization/82491
18903 * gimple-fold.c (get_base_constructor): Make earlier bail out
18906 2018-02-19 Carl Love <cel@us.ibm.com>
18908 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
18909 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
18911 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
18912 P8V_BUILTIN_VEC_NEG.
18914 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
18916 * config/rl78/rl78.md (movdf): New define expand.
18918 2018-02-19 Martin Liska <mliska@suse.cz>
18921 * doc/invoke.texi: Fix typo.
18922 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
18924 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
18926 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
18927 handle rs6000_single_float and rs6000_double_float specially for
18930 2018-02-16 Jeff Law <law@redhat.com>
18932 * config/rx/rx.c (add_pop_cfi_notes): New function.;
18933 (pop_regs): Use it.
18935 2018-02-16 Jakub Jelinek <jakub@redhat.com>
18938 * ipa-inline.c (inline_small_functions): Fix a typo.
18940 2018-02-16 Nathan Sidwell <nathan@acm.org>
18942 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
18944 2018-02-16 Carl Love <cel@us.ibm.com>
18946 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
18947 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
18948 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
18949 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
18950 expansion to P8V_BUILTIN_VEC_FLOAT2.
18952 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
18954 PR rtl-optimization/70023
18955 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
18956 src_regno into account.
18958 2018-02-16 Carl Love <cel@us.ibm.com>
18960 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
18961 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
18962 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
18963 * config/rs6000/rs6000.c: Remove case statements for
18964 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
18965 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
18966 and P9V_BUILTIN_VEC_VINSERT4B.
18967 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
18968 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
18969 * config/rs6000/vsx.md:
18970 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
18973 2018-02-16 Carl Love <cel@us.ibm.com>
18975 * config/rs6000/altivec.h: Add builtin names vec_extract4b
18977 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
18979 * config/rs6000/rs6000-c.c: Add the definitions for
18980 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
18981 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
18982 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
18983 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
18984 definition for insert4b and define insn *insert3b_internal.
18985 * doc/extend.texi: Add documentation for vec_extract4b.
18987 2018-02-16 Nathan Sidwell <nathan@acm.org>
18989 * doc/extend.texi (Backwards Compatibility): Mention friend
18990 injection. Note for-scope is deprecated.
18991 * doc/invoke.texi (-ffriend-injection): Deprecate.
18993 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
18995 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
18996 that moved to I2, also allow destinations that are a paradoxical
18997 subreg (instead of a normal reg).
18999 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
19002 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
19005 2018-02-16 Richard Biener <rguenther@suse.de>
19007 PR tree-optimization/84037
19008 PR tree-optimization/84016
19010 * config/i386/i386.c (ix86_builtin_vectorization_cost):
19011 Adjust vec_construct for the fact we need additional higher latency
19012 128bit inserts for AVX256 and AVX512 vector builds.
19013 (ix86_add_stmt_cost): Scale vector construction cost for
19016 2018-02-16 Richard Biener <rguenther@suse.de>
19018 PR tree-optimization/84417
19019 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
19020 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
19021 (non_rewritable_lvalue_p): Likewise, use poly-ints.
19023 2018-02-16 Martin Liska <mliska@suse.cz>
19026 * internal-fn.def (ASAN_CHECK): Set proper flags.
19027 (ASAN_MARK): Likewise.
19029 2018-02-16 Julia Koval <julia.koval@intel.com>
19031 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
19032 from PTA_CANNONLAKE.
19034 2018-02-16 Jakub Jelinek <jakub@redhat.com>
19037 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
19038 Use ++iter rather than iter++ for std::list iterators.
19039 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
19040 defer deleting them until all nodes in the forest are processed. Do
19041 free even leaf nodes. Change to_process into auto_vec.
19044 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
19045 * vec.h (vec_default_construct): Use memset instead of placement new
19046 if BROKEN_VALUE_INITIALIZATION is defined.
19047 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
19048 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
19051 PR rtl-optimization/83723
19052 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
19053 * lra.c (lra_substitute_pseudo): Likewise. If true, use
19054 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
19056 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
19058 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
19060 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
19062 PR rtl-optimization/81443
19063 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
19064 from inner REGs to paradoxical SUBREGs.
19066 2018-02-16 Richard Biener <rguenther@suse.de>
19068 PR tree-optimization/84399
19069 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
19070 For operands we can analyze at their definition make sure we can
19071 analyze them at each use as well.
19073 2018-02-16 Richard Biener <rguenther@suse.de>
19075 PR tree-optimization/84190
19076 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
19077 volatile accesses if the decl isn't volatile.
19079 2018-02-15 Jason Merrill <jason@redhat.com>
19081 PR c++/84314 - ICE with templates and fastcall attribute.
19082 * attribs.c (build_type_attribute_qual_variant): Don't clobber
19083 TYPE_CANONICAL on an existing type.
19085 2018-02-15 Jakub Jelinek <jakub@redhat.com>
19087 PR tree-optimization/84383
19088 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
19089 dstoff nor call operand_equal_p if dstbase is NULL.
19091 PR tree-optimization/84334
19092 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
19093 also a CONSTANT_CLASS_P, punt.
19095 2018-02-14 Jim Wilson <jimw@sifive.com>
19097 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
19098 first SMALL_OPERAND check. New local min_second_step. Move assert
19099 to where locals are set. Add TARGET_RVC support.
19100 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
19102 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
19104 * doc/invoke.texi: Correct -Wformat-overflow code sample.
19106 2018-02-14 Martin Sebor <msebor@redhat.com>
19108 PR tree-optimization/83698
19109 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
19110 arrays constrain the offset range to their bounds.
19111 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
19112 (builtin_access::overlap): Avoid setting the size of overlap if it's
19114 (maybe_diag_overlap): Also consider arrays when deciding what values
19115 of offsets to include in diagnostics.
19117 2018-02-14 Martin Sebor <msebor@redhat.com>
19120 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
19121 that correspond to the kind of a declaration.
19123 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
19126 * config/pa/pa.md: Load address of PIC label using the linkage table
19127 if the label is nonlocal.
19129 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
19131 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
19132 warning message if user requests -maltivec=be.
19133 * doc/invoke.texi: Document deprecation of -maltivec=be.
19135 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
19138 * config/rs6000/rs6000-c.c: Update definitions for
19139 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
19140 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
19142 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
19145 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
19146 add _get_ssp intrinsics. Remove argument from
19147 __builtin_ia32_rdssp[d|q].
19148 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
19149 * config/i386/i386-builtin.def: Remove argument from
19150 __builtin_ia32_rdssp[d|q].
19151 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
19152 ix86_expand_special_args_builtin for _rdssp[d|q].
19153 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
19154 Clear register before usage.
19155 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
19156 Add documentation for new _get_ssp and _inc_ssp intrinsics.
19158 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
19160 PR tree-optimization/84357
19161 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
19162 operand 1 of an ARRAY_REF too.
19164 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
19167 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
19168 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
19170 (set_of_reg): New struct.
19171 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
19172 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
19173 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
19175 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
19176 Split into bitclr, bitset, bitinvert patterns if appropriate.
19177 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
19178 use rx_fuse_in_memory_bitop.
19179 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
19180 to named insn, correct maximum insn length.
19182 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
19185 * machmode.def: Define a complex mode for PARTIAL_INT.
19186 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
19188 * doc/rtl.texi: Document CSPImode.
19189 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
19191 (msp430_hard_regno_nregs_with_padding): Likewise.
19193 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
19196 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
19198 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
19200 PR rtl-optimization/84169
19201 * combine.c (try_combine): New variable split_i2i3. Set it to true if
19202 we generated a parallel as new i3 and we split that to new i2 and i3
19203 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
19204 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
19205 those to i2, not i1. Partially rewrite this scan code.
19207 2018-02-13 Jakub Jelinek <jakub@redhat.com>
19210 * stor-layout.c (place_field): For variable length fields, adjust
19211 offset_align afterwards not just based on the field's alignment,
19212 but also on the size.
19214 PR middle-end/84309
19215 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
19216 of exps and logs in the use_exp2 case.
19218 2018-02-13 Jeff Law <law@redhat.com>
19220 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
19221 entry for "vector".
19223 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
19226 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
19230 * common.opt (gas-loc-support, gas-locview-support): New.
19231 (ginline-points, ginternal-reset-location-views): New.
19232 * doc/invoke.texi: Document them. Use @itemx where intended.
19233 (gvariable-location-views): Adjust.
19234 * target.def (reset_location_view): New.
19235 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
19236 (TARGET_RESET_LOCATION_VIEW): New.
19237 * doc/tm.texi: Rebuilt.
19238 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
19239 (dwarf2out_default_as_locview_support): New.
19240 (output_asm_line_debug_info): Use option variables.
19241 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
19242 (output_loc_list): Likewise.
19243 (add_high_low_attributes): Check option variables.
19244 Don't output entry view attribute in strict mode.
19245 (gen_inlined_subroutine_die): Check option variables.
19246 (dwarf2out_inline_entry): Likewise.
19247 (init_sections_and_labels): Likewise.
19248 (dwarf2out_early_finish): Likewise.
19249 (maybe_reset_location_view): New, from...
19250 (dwarf2out_var_location): ... here. Call it.
19251 * debug.h (dwarf2out_default_as_loc_support): Declare.
19252 (dwarf2out_default_as_locview_support): Declare.
19253 * hooks.c (hook_int_rtx_insn_0): New.
19254 * hooks.h (hook_int_rtx_insn_0): Declare.
19255 * toplev.c (process_options): Take -gas-loc-support and
19256 -gas-locview-support from dwarf2out. Enable
19257 -gvariable-location-views by default only with locview
19258 assembler support. Enable -ginternal-reset-location-views by
19259 default only if the target defines the corresponding hook.
19260 Enable -ginline-points by default if location views are
19261 enabled; force it disabled if statement frontiers are
19263 * tree-inline.c (expand_call_inline): Check option variables.
19264 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
19266 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
19268 PR tree-optimization/84321
19269 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
19270 handling. Also check whether the anti-range contains any values
19271 that satisfy the mask; switch to a VR_RANGE if not.
19273 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
19276 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
19278 2018-02-13 Martin Jambor <mjambor@suse.cz>
19281 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
19282 of call statements, also set location of a load to a temporary.
19284 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
19286 * config/rl78/rl78.c (add_vector_labels): New function.
19287 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
19288 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
19289 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
19290 which checks that no arguments are passed.
19291 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
19292 * doc/extend.texi: Documentation for the new attribute.
19294 2018-02-13 Andreas Schwab <schwab@suse.de>
19296 * config/riscv/linux.h (CPP_SPEC): Define.
19298 2018-02-13 Jakub Jelinek <jakub@redhat.com>
19301 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
19302 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
19303 OPTION_MASK_ISA_AES as first argument to def_builtin_const
19304 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
19305 instead of OPTION_MASK_ISA_PCLMUL as first argument to
19306 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
19307 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
19308 temporarily for AES and PCLMUL builtins.
19310 PR tree-optimization/84339
19311 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
19312 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
19315 PR middle-end/84309
19316 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
19317 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
19318 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
19320 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
19322 * omp-simd-clone.h: New file.
19323 * omp-simd-clone.c: Include omp-simd-clone.h.
19324 (expand_simd_clones): No longer static.
19325 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
19326 cgraph.h and omp-simd-clone.h.
19327 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
19328 (vect_recog_widen_shift_pattern): Formatting fix.
19329 (vect_pattern_recog_1): Don't check optab for calls.
19332 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
19333 operands[2] into a REG before using gen_lowpart on it.
19335 2018-02-12 Jeff Law <law@redhat.com>
19338 * config/sh/sh.c (find_barrier): Consider a sibling call
19341 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
19342 successfully back substituting a reg.
19344 2018-02-12 Richard Biener <rguenther@suse.de>
19346 PR tree-optimization/84037
19347 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
19348 parameter, move visited init to caller.
19349 (vect_slp_analyze_operations): Separate cost from validity
19350 check, initialize visited once for all instances.
19351 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
19353 * tree-vect-stmts.c (vect_model_simple_cost): Make early
19355 (vect_model_promotion_demotion_cost): Likewise.
19356 (vectorizable_bswap): Guard cost modeling with !slp_node
19357 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
19359 (vectorizable_call): Likewise.
19360 (vectorizable_conversion): Likewise.
19361 (vectorizable_assignment): Likewise.
19362 (vectorizable_shift): Likewise.
19363 (vectorizable_operation): Likewise.
19364 (vectorizable_store): Likewise.
19365 (vectorizable_load): Likewise.
19366 (vectorizable_condition): Likewise.
19367 (vectorizable_comparison): Likewise.
19369 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
19372 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
19373 (ASAN_MARK): Fix fnspec to account for return value, change pointer
19374 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
19376 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
19378 PR middle-end/83665
19379 * params.def (inline-min-speedup): Increase from 8 to 15.
19380 (max-inline-insns-auto): Decrease from 40 to 30.
19381 * ipa-split.c (consider_split): Add some buffer for function to
19382 be considered inlining candidate.
19383 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
19386 2018-02-12 Richard Biener <rguenther@suse.de>
19388 PR tree-optimization/84037
19389 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
19390 matched stmts if we cannot swap the non-matched ones.
19392 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
19394 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
19395 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
19396 _mm_maskz_scalef_round_ss): New intrinsics.
19397 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
19398 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
19399 __builtin_ia32_scalefss_round): Remove.
19400 (__builtin_ia32_scalefsd_mask_round,
19401 __builtin_ia32_scalefss_mask_round): New intrinsics.
19402 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
19403 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
19404 ((match_operand:VF_128 2 "<round_nimm_predicate>"
19405 "<round_constraint>")): Changed to ...
19406 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
19407 "<round_scalar_constraint>")): ... this.
19408 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
19409 %0, %1, %2<round_op3>}"): Changed to ...
19410 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
19411 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
19412 %2<round_scalar_mask_op3>}"): ... this.
19413 * config/i386/subst.md (round_scalar_nimm_predicate): New.
19415 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
19417 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
19418 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
19419 (_mm_maskz_sqrt_round_ss): New intrinsics.
19420 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
19421 (__builtin_ia32_sqrtsd_mask_round)
19422 (__builtin_ia32_sqrtss_mask_round): New builtins.
19423 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
19424 (__builtin_ia32_sqrtss_round): Remove.
19425 (__builtin_ia32_sqrtsd_mask_round)
19426 (__builtin_ia32_sqrtss_mask_round): New builtins.
19427 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
19428 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
19429 ((match_operand:VF_128 1 "vector_operand"
19430 "xBm,<round_constraint>")): Changed to ...
19431 ((match_operand:VF_128 1 "vector_operand"
19432 "xBm,<round_scalar_constraint>")): ... this.
19433 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
19434 %0, %2, %<iptr>1<round_op3>}): Changed to ...
19435 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
19436 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
19437 %<iptr>1<round_scalar_mask_op3>}): ... this.
19438 ((set_attr "prefix" "<round_prefix>")): Changed to ...
19439 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
19441 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
19444 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
19445 Cast vec_cmpeq result to correct type.
19446 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
19447 Cast vec_cmpgt result to correct type.
19449 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
19451 * final.c (final_scan_insn_1): Renamed from...
19452 (final_scan_insn): ... this. New wrapper, to recover
19453 seen from the outermost call in recursive ones.
19454 * config/sparc/sparc.c (output_return): Drop seen from call.
19455 (output_sibcall): Likewise.
19456 * config/visium/visium.c (output_branch): Likewise.
19458 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
19460 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
19463 2018-02-10 Alan Modra <amodra@gmail.com>
19466 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
19467 Specify LR as an input.
19469 2018-02-10 Jakub Jelinek <jakub@redhat.com>
19472 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
19473 remove_member_access_dummy_vars): New functions.
19474 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
19475 lower_omp_1, execute_lower_omp): Use them.
19477 PR rtl-optimization/84308
19478 * shrink-wrap.c (spread_components): Release todo vector.
19480 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
19482 PR rtl-optimization/57193
19483 * ira-color.c (struct allocno_color_data): Add member
19484 conflict_allocno_hard_prefs.
19485 (update_conflict_allocno_hard_prefs): New.
19486 (bucket_allocno_compare_func): Add a preference based on
19487 conflict_allocno_hard_prefs.
19488 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
19489 (color_allocnos): Remove a dead code. Initiate
19490 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
19492 2018-02-09 Jakub Jelinek <jakub@redhat.com>
19495 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
19496 constraint from =wa to wa. Avoid a subreg on the output operand,
19497 instead use a pseudo and subreg it in a move.
19498 (p9_xxbrd_<mode>): Changed to ...
19499 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
19500 (p9_xxbrd_v2df): New expander.
19501 (p9_xxbrw_<mode>): Changed to ...
19502 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
19503 (p9_xxbrw_v4sf): New expander.
19505 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
19507 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
19509 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
19512 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
19513 multiply in 32-bit mode.
19514 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
19515 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
19518 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
19520 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
19521 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
19522 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
19523 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
19525 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
19528 * dwarf2out.c (is_trivial_indirect_ref): New function.
19529 (dwarf2out_late_global_decl): Do not generate a location
19530 attribute for variables that have a non-trivial DECL_VALUE_EXPR
19531 and that are not defined in the current unit.
19533 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
19535 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
19536 instead of a libcall for UNORDERED.
19538 2018-02-09 Tamar Christina <tamar.christina@arm.com>
19541 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
19542 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
19544 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
19547 * config/s390/s390.c (s390_set_current_function): Invoke
19548 s390_indirect_branch_settings also if fndecl didn't change.
19550 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
19552 * config/rs6000/rs6000.md (blockage): Set length to zero.
19554 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
19556 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
19558 2018-02-09 Jakub Jelinek <jakub@redhat.com>
19561 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
19562 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
19566 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
19567 PARALLEL incoming that failed vt_get_decl_and_offset check.
19569 PR middle-end/84237
19570 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
19571 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
19573 (get_variable_section): For decls in named .bss* sections pass true as
19574 second argument to bss_initializer_p.
19576 2018-02-09 Marek Polacek <polacek@redhat.com>
19577 Jakub Jelinek <jakub@redhat.com>
19580 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
19581 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
19582 Sync some changes from cxx_fold_indirect_ref.
19584 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
19586 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
19588 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
19589 (BLOCK_INLINE_ENTRY_LABEL): New.
19590 (dwarf2out_var_location): Disregard inline entry markers.
19591 (inline_entry_data): New struct.
19592 (inline_entry_data_hasher): New hashtable type.
19593 (inline_entry_data_hasher::hash): New.
19594 (inline_entry_data_hasher::equal): New.
19595 (inline_entry_data_table): New variable.
19596 (add_high_low_attributes): Add DW_AT_entry_pc and
19597 DW_AT_GNU_entry_view attributes if a pending entry is found
19598 in inline_entry_data_table. Add old entry_pc attribute only
19599 if debug nonbinding markers are disabled.
19600 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
19601 markers are enabled.
19602 (block_within_block_p, dwarf2out_inline_entry): New.
19603 (dwarf2out_finish): Check that no entries remained in
19604 inline_entry_data_table.
19605 * final.c (reemit_insn_block_notes): Handle inline entry notes.
19606 (final_scan_insn, notice_source_line): Likewise.
19607 (rest_of_clean_state): Skip inline entry markers.
19608 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
19610 * gimple.c (gimple_build_debug_inline_entry): New.
19611 * gimple.h (enum gimple_debug_subcode): Add
19612 GIMPLE_DEBUG_INLINE_ENTRY.
19613 (gimple_build_debug_inline_entry): Declare.
19614 (gimple_debug_inline_entry_p): New.
19615 (gimple_debug_nonbind_marker_p): Adjust.
19616 * insn-notes.def (INLINE_ENTRY): New.
19617 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
19618 inline entry marker notes.
19619 (print_insn): Likewise.
19620 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
19621 (INSN_DEBUG_MARKER_KIND): Likewise.
19622 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
19623 * tree-inline.c (expand_call_inline): Build and insert
19624 debug_inline_entry stmt.
19625 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
19626 inline entry blocks early, if nonbind markers are enabled.
19627 (dump_scope_block): Dump fragment info.
19628 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
19629 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
19630 (gimple_build_debug_inline_entry): New.
19631 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
19632 Enable/disable inline entry points too.
19633 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
19634 (DEBUG_INSN): Describe inline entry markers.
19636 * common.opt (gvariable-location-views): New.
19637 (gvariable-location-views=incompat5): New.
19638 * config.in: Rebuilt.
19639 * configure: Rebuilt.
19640 * configure.ac: Test assembler for view support.
19641 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
19642 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
19643 * dwarf2out.c (var_loc_view): New typedef.
19644 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
19645 (dwarf2out_locviews_in_attribute): New.
19646 (dwarf2out_locviews_in_loclist): New.
19647 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
19648 (enum dw_line_info_opcode): Add LI_adv_address.
19649 (struct dw_line_info_table): Add view.
19650 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
19651 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
19652 (zero_view_p): New variable.
19653 (ZERO_VIEW_P): New macro.
19654 (output_asm_line_debug_info): New.
19655 (struct var_loc_node): Add view.
19656 (add_AT_view_list, AT_loc_list): New.
19657 (add_var_loc_to_decl): Add view param. Test it against last.
19658 (new_loc_list): Add view params. Record them.
19659 (AT_loc_list_ptr): Handle loc and view lists.
19660 (view_list_to_loc_list_val_node): New.
19661 (print_dw_val): Handle dw_val_class_view_list.
19662 (size_of_die): Likewise.
19663 (value_format): Likewise.
19664 (loc_list_has_views): New.
19665 (gen_llsym): Set vl_symbol too.
19666 (maybe_gen_llsym, skip_loc_list_entry): New.
19667 (dwarf2out_maybe_output_loclist_view_pair): New.
19668 (output_loc_list): Output view list or entries too.
19669 (output_view_list_offset): New.
19670 (output_die): Handle dw_val_class_view_list.
19671 (output_dwarf_version): New.
19672 (output_compilation_unit_header): Use it.
19673 (output_skeleton_debug_sections): Likewise.
19674 (output_rnglists, output_line_info): Likewise.
19675 (output_pubnames, output_aranges): Update version comments.
19676 (output_one_line_info_table): Output view numbers in asm comments.
19677 (dw_loc_list): Determine current endview, pass it to new_loc_list.
19678 Call maybe_gen_llsym.
19679 (loc_list_from_tree_1): Adjust.
19680 (add_AT_location_description): Create view list attribute if
19681 needed, check it's absent otherwise.
19682 (convert_cfa_to_fb_loc_list): Adjust.
19683 (maybe_emit_file): Call output_asm_line_debug_info for test.
19684 (dwarf2out_var_location): Reset views as needed. Precompute
19685 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
19686 attribute. Set view.
19687 (new_line_info_table): Reset next view.
19688 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
19689 (dwarf2out_source_line): Likewise. Output view resets and labels to
19690 the assembler, or select appropriate line info opcodes.
19691 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
19692 (optimize_string_length): Catch it. Adjust.
19693 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
19694 dw_val_class_view_list, and remove it if no longer needed.
19695 (hash_loc_list): Hash view numbers.
19696 (loc_list_hasher::equal): Compare them.
19697 (optimize_location_lists): Check whether a view list symbol is
19698 needed, and whether the locview attribute is present, and
19699 whether they match. Remove the locview attribute if no longer
19701 (index_location_lists): Call skip_loc_list_entry for test.
19702 (dwarf2out_finish): Call output_asm_line_debug_info for test.
19703 Use output_dwarf_version.
19704 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
19705 (struct dw_val_node): Add val_view_list.
19706 * final.c (SEEN_NEXT_VIEW): New.
19707 (set_next_view_needed): New.
19708 (clear_next_view_needed): New.
19709 (maybe_output_next_view): New.
19710 (final_start_function): Rename to...
19711 (final_start_function_1): ... this. Take pointer to FIRST,
19712 add SEEN parameter. Emit param bindings in the initial view.
19713 (final_start_function): Reintroduce SEEN-less interface.
19714 (final): Rename to...
19715 (final_1): ... this. Take SEEN parameter. Output final pending
19716 next view at the end.
19717 (final): Reintroduce seen-less interface.
19718 (final_scan_insn): Output pending next view before switching
19719 sections or ending a block. Mark the next view as needed when
19720 outputting variable locations. Notify debug backend of section
19721 changes, and of location view changes.
19722 (rest_of_handle_final): Adjust.
19723 * toplev.c (process_options): Autodetect value for debug variable
19724 location views option. Warn on incompat5 without -gdwarf-5.
19725 * doc/invoke.texi (gvariable-location-views): New.
19726 (gvariable-location-views=incompat5): New.
19727 (gno-variable-location-views): New.
19729 2018-02-08 David Malcolm <dmalcolm@redhat.com>
19731 PR tree-optimization/84136
19732 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
19733 that the result of find_edge is non-NULL.
19735 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
19738 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
19739 storing integer register in SImode. Fix cost of 256 and 512
19740 byte aligned SSE register store.
19742 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
19744 * config/i386/i386.c (ix86_multiplication_cost): Fix
19745 multiplication cost for TARGET_AVX512DQ.
19747 2018-02-08 Marek Polacek <polacek@redhat.com>
19749 PR tree-optimization/84238
19750 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
19753 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
19755 PR tree-optimization/84265
19756 * tree-vect-stmts.c (vectorizable_store): Don't treat
19757 VMAT_CONTIGUOUS accesses as grouped.
19758 (vectorizable_load): Likewise.
19760 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
19762 PR tree-optimization/81635
19763 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
19764 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
19765 (test_round_for_mask): New functions.
19766 (wide_int_cc_tests): Call test_round_for_mask.
19767 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
19768 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
19769 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
19770 range returned by get_range_info.
19772 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
19775 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
19776 * symtab.c: Include builtins.h
19777 (symtab_node::output_to_lto_symbol_table_p): Move here
19778 from lto-streamer-out.c:output_symbol_p.
19779 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
19780 (output_symbol_p): Move all logic to symtab.c
19781 (produce_symtab): Update.
19783 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
19785 * config/s390/s390-opts.h (enum indirect_branch): Define.
19786 * config/s390/s390-protos.h (s390_return_addr_from_memory)
19787 (s390_indirect_branch_via_thunk)
19788 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
19789 (enum s390_indirect_branch_type): Define.
19790 * config/s390/s390.c (struct s390_frame_layout, struct
19791 machine_function): Remove.
19792 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
19793 (indirect_branch_table_label_no, indirect_branch_table_name):
19795 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
19796 (enum s390_indirect_branch_option): Define.
19797 (s390_return_addr_from_memory): New function.
19798 (s390_handle_string_attribute): New function.
19799 (s390_attribute_table): Add new attribute handler.
19800 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
19801 (s390_indirect_branch_via_thunk): New function.
19802 (s390_indirect_branch_via_inline_thunk): New function.
19803 (s390_function_ok_for_sibcall): When jumping via thunk disallow
19804 sibling call optimization for non z10 compiles.
19805 (s390_emit_call): Force indirect branch target to be a single
19806 register. Add r1 clobber for non-z10 compiles.
19807 (s390_emit_epilogue): Emit return jump via return_use expander.
19808 (s390_reorg): Handle JUMP_INSNs as execute targets.
19809 (s390_option_override_internal): Perform validity checks for the
19810 new command line options.
19811 (s390_indirect_branch_attrvalue): New function.
19812 (s390_indirect_branch_settings): New function.
19813 (s390_set_current_function): Invoke s390_indirect_branch_settings.
19814 (s390_output_indirect_thunk_function): New function.
19815 (s390_code_end): Implement target hook.
19816 (s390_case_values_threshold): Implement target hook.
19817 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
19819 * config/s390/s390.h (struct s390_frame_layout)
19820 (struct machine_function): Move here from s390.c.
19821 (TARGET_INDIRECT_BRANCH_NOBP_RET)
19822 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
19823 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
19824 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
19825 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
19826 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
19827 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
19828 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
19829 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
19830 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
19831 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
19832 (mnemonic attribute): Add values which aren't recognized
19834 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
19835 pattern for branch conversion. Fix mnemonic attribute.
19836 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
19837 indirect branch via thunk if requested.
19838 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
19839 ("*indirect_jump"): Disable for branch conversion using out of
19841 ("indirect_jump_via_thunk<mode>_z10")
19842 ("indirect_jump_via_thunk<mode>")
19843 ("indirect_jump_via_inlinethunk<mode>_z10")
19844 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
19845 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
19846 ("casesi_jump_via_inlinethunk<mode>_z10")
19847 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
19848 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
19849 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
19850 ("*indirect2_jump"): Disable for branch conversion.
19851 ("casesi_jump"): Turn into expander and expand patterns for branch
19853 ("return_use"): New expander.
19854 ("*return"): Emit return via thunk and rename it to ...
19855 ("*return<mode>"): ... this one.
19856 * config/s390/s390.opt: Add new options and and enum for the
19859 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
19861 * lra-constraints.c (match_reload): Unconditionally use
19862 gen_lowpart_SUBREG, rather than selecting between that
19863 and equivalent gen_rtx_SUBREG code.
19865 2018-02-08 Richard Biener <rguenther@suse.de>
19867 PR tree-optimization/84233
19868 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
19869 changed flag instead of boguously re-using phi_inserted.
19871 2018-02-08 Martin Jambor <mjambor@suse.cz>
19873 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
19874 static local variables.
19876 2018-02-08 Richard Biener <rguenther@suse.de>
19878 PR tree-optimization/84278
19879 * tree-vect-stmts.c (vectorizable_store): When looking for
19880 smaller vector types to perform grouped strided loads/stores
19881 make sure the mode is supported by the target.
19882 (vectorizable_load): Likewise.
19884 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
19886 * config/aarch64/aarch64.c (aarch64_components_for_bb):
19887 Increase LDP/STP opportunities by adding adjacent callee-saves.
19889 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
19891 PR rtl-optimization/84068
19892 PR rtl-optimization/83459
19893 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
19895 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
19897 PR tree-optimization/84224
19898 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
19899 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
19900 non-zero arguments.
19902 2018-02-07 Iain Sandoe <iain@codesourcery.com>
19905 * config/rs6000/altivec.md (*restore_world): Remove LR use.
19906 * config/rs6000/predicates.md (restore_world_operation): Adjust op
19907 count, remove one USE.
19909 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
19911 * doc/install.texi (Configuration): Document the
19912 --with-long-double-format={ibm,ieee} PowerPC configuration
19916 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
19917 Convert from define_expand to be define_insn_and_split. Rework
19918 float/double/_Float128 conversions to QI/HI/SImode to work with
19919 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
19920 conversions to QI/HImode types did a store and then a load to
19921 truncate the value. For conversions to VSX registers, don't split
19922 the insn, instead emit the code directly. Use the code iterator
19923 any_fix to combine signed and unsigned conversions.
19924 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
19925 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
19926 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
19927 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
19928 (fix_<mode>di2_hw): Likewise.
19929 (fixuns_<mode>di2_hw): Likewise.
19930 (fix_<mode>si2_hw): Likewise.
19931 (fixuns_<mode>si2_hw): Likewise.
19932 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
19933 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
19934 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
19935 fix<uns>_trunc<SFDF:mode>si2_p8.
19936 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
19937 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
19938 (fix<uns>_<mode>_mem): Likewise.
19939 (fctiw<u>z_<mode>_mem): Likewise.
19940 (fix<uns>_<mode>_mem): Likewise.
19941 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
19942 the register allocator from doing a direct move to the GPRs to do
19943 a store, and instead use the ISA 3.0 store byte/half-word from
19944 vector register instruction. For IEEE 128-bit floating point,
19945 also optimize stores of 32-bit ints.
19946 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
19948 2018-02-07 Alan Hayward <alan.hayward@arm.com>
19950 * genextract.c (push_pathstr_operand): New function to support
19952 (walk_rtx): Call push_pathstr_operand.
19953 (print_path): Support [a-zA-Z].
19955 2018-02-07 Richard Biener <rguenther@suse.de>
19957 PR tree-optimization/84037
19958 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
19959 (cse_and_gimplify_to_preheader): Declare.
19960 (vect_get_place_in_interleaving_chain): Likewise.
19961 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19963 (_loop_vec_info::~_loop_vec_info): Delete it.
19964 (cse_and_gimplify_to_preheader): New function.
19965 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
19966 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
19967 (vectorizable_load): Likewise. For grouped stores always base
19968 the IV on the first element.
19969 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
19970 condition before gimplifying.
19972 2018-02-07 Jakub Jelinek <jakub@redhat.com>
19974 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
19975 *DIV_EXPR and *MOD_EXPR.
19977 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
19980 * config/i386/i386.c (ix86_option_override_internal): Mask out
19981 the CF_SET bit when checking -fcf-protection.
19983 2018-02-07 Tom de Vries <tom@codesourcery.com>
19986 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
19989 2018-02-07 Richard Biener <rguenther@suse.de>
19991 PR tree-optimization/84204
19992 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
19995 PR tree-optimization/84205
19996 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
19997 special-case isl_ast_op_zdiv_r.
19999 PR tree-optimization/84223
20000 * graphite-scop-detection.c (gather_bbs::before_dom_children):
20001 Only add conditions from within the region.
20002 (gather_bbs::after_dom_children): Adjust.
20004 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
20007 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
20008 * config/avr/avr.md: Only post-reload split REG-REG moves if
20009 either register is GENERAL_REG_P.
20011 2018-02-07 Jakub Jelinek <jakub@redhat.com>
20013 PR tree-optimization/84235
20014 * tree-ssa-scopedtables.c
20015 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
20016 if the subtraction is performed in floating point type where NaNs are
20017 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
20018 build 1. Formatting fix.
20020 2018-02-06 Jakub Jelinek <jakub@redhat.com>
20023 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
20024 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
20025 and skip it regardless of bb boundaries. Use CALL_P macro,
20026 don't test INSN_P (insn) together with CALL_P or JUMP_P check
20027 unnecessarily, formatting fix.
20029 2018-02-06 Michael Collison <michael.collison@arm.com>
20031 * config/arm/thumb2.md:
20032 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
20033 (*thumb_mov_notscc): Ditto.
20035 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
20038 * config/rs6000/rs6000.md (su code attribute): Use "u" for
20039 unsigned_fix, not "s".
20041 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20043 * configure.ac (gcc_fn_eh_frame_ro): New function.
20044 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
20045 correct .eh_frame permissions.
20046 * configure: Regenerate.
20048 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
20050 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
20051 irrelevant options.
20053 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20055 * config/rs6000/rs6000.c (rs6000_option_override_internal):
20056 Display warning message for -mno-speculate-indirect-jumps.
20058 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
20060 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
20062 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
20064 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
20066 PR tree-optimization/84225
20067 * tree-eh.c (find_trapping_overflow): Only call
20068 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
20070 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
20073 * config/i386/i386.c: Reimplement the check of possible options
20074 -mibt/-mshstk conbination. Change error messages.
20075 * doc/invoke.texi: Fix a typo: remove extra '='.
20077 2018-02-06 Marek Polacek <polacek@redhat.com>
20079 PR tree-optimization/84228
20080 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
20082 2018-02-06 Tamar Christina <tamar.christina@arm.com>
20085 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
20086 emitted arch directives.
20087 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
20088 __ARM_FEATURE_COPROC before changing architectures.
20090 2018-02-06 Richard Biener <rguenther@suse.de>
20092 * config/i386/i386.c (print_reg): Fix typo.
20093 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
20095 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
20097 * configure: Regenerate.
20099 2018-02-05 Martin Sebor <msebor@redhat.com>
20101 PR tree-optimization/83369
20102 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
20105 2018-02-05 Martin Liska <mliska@suse.cz>
20107 * doc/invoke.texi: Cherry-pick upstream r323995.
20109 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
20111 * ira.c (ira_init_register_move_cost): Adjust comment.
20113 2018-02-05 Martin Liska <mliska@suse.cz>
20115 PR gcov-profile/84137
20116 * doc/gcov.texi: Fix typo in documentation.
20118 2018-02-05 Martin Liska <mliska@suse.cz>
20120 PR gcov-profile/83879
20121 * doc/gcov.texi: Document necessity of --dynamic-list-data when
20122 using dlopen functionality.
20124 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
20126 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
20127 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
20128 _mm_maskz_range_ss, _mm_mask_range_round_ss,
20129 _mm_maskz_range_round_ss): New intrinsics.
20130 (__builtin_ia32_rangesd128_round)
20131 (__builtin_ia32_rangess128_round): Remove.
20132 (__builtin_ia32_rangesd128_mask_round,
20133 __builtin_ia32_rangess128_mask_round): New builtins.
20134 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
20135 __builtin_ia32_rangess128_round): Remove.
20136 (__builtin_ia32_rangesd128_mask_round,
20137 __builtin_ia32_rangess128_mask_round): New builtins.
20138 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
20139 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
20140 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
20141 "<round_saeonly_constraint>")): Changed to ...
20142 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
20143 "<round_saeonly_scalar_constraint>")): ... this.
20144 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
20145 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
20146 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
20147 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
20148 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
20150 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
20152 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
20154 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
20155 Remove all values except native, 8540 and 8548.
20157 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
20159 * config/i386/i386.c (ix86_output_function_return): Pass
20160 INVALID_REGNUM, instead of -1, as invalid register number to
20161 indirect_thunk_name and output_indirect_thunk.
20163 2018-02-02 Julia Koval <julia.koval@intel.com>
20165 * config.gcc: Add -march=icelake.
20166 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
20167 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
20168 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
20169 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
20170 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
20171 (processor_target_table): Add icelake.
20172 (ix86_option_override_internal): Handle new PTAs.
20173 (get_builtin_code_for_version): Handle icelake.
20174 (M_INTEL_COREI7_ICELAKE): New.
20175 (fold_builtin_cpu): Handle icelake.
20176 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
20177 * doc/invoke.texi: Add -march=icelake.
20179 2018-02-02 Julia Koval <julia.koval@intel.com>
20181 * config/i386/i386.c (ix86_option_override_internal): Change flags type
20182 to wide_int_bitmask.
20183 * wide-int-bitmask.h: New.
20185 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
20188 * config/i386/i386.md: Replace Pmode with word_mode in
20189 builtin_setjmp_setup and builtin_longjmp to support x32.
20191 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
20195 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
20197 (rs6000_supported_cpu_names): New static variable.
20198 (linux_cpu_translation_table): Likewise.
20199 (elf_platform) <cpu>: Define new static variable and use it.
20200 Translate kernel AT_PLATFORM name to canonical name if needed.
20201 Error if platform name is unknown.
20203 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
20206 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
20208 2018-02-01 Jeff Law <law@redhat.com>
20211 * config/i386/i386.c (release_scratch_register_on_entry): Add new
20212 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
20213 the scratch if RELEASE_VIA_POP is false.
20214 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
20215 If we have to save a temporary register, decrement SIZE appropriately.
20216 Pass new arguments to release_scratch_register_on_entry.
20217 (ix86_adjust_stack_and_probe): Likewise.
20218 (ix86_emit_probe_stack_range): Pass new arguments to
20219 release_scratch_register_on_entry.
20221 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
20223 PR rtl-optimization/84157
20224 * combine.c (change_zero_ext): Use REG_P predicate in
20225 front of HARD_REGISTER_P predicate.
20227 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
20229 * config/avr/avr.c (avr_option_override): Move disabling of
20230 -fdelete-null-pointer-checks to...
20231 * common/config/avr/avr-common.c (avr_option_optimization_table):
20234 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20236 PR tree-optimization/81635
20237 * tree-data-ref.c (split_constant_offset_1): For types that
20238 wrap on overflow, try to use range info to prove that wrapping
20241 2018-02-01 Renlin Li <renlin.li@arm.com>
20244 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
20245 TAILCALL_ADDR_REGS.
20246 (aarch64_register_move_cost): Likewise.
20247 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
20248 TAILCALL_ADDR_REGS.
20249 (REG_CLASS_NAMES): Likewise.
20250 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
20251 TAILCALL_ADDR_REGS. Remove IP registers.
20252 * config/aarch64/aarch64.md (Ucs): Update register constraint.
20254 2018-02-01 Richard Biener <rguenther@suse.de>
20256 * domwalk.h (dom_walker::dom_walker): Add additional constructor
20257 for specifying RPO order and allow NULL for that.
20258 * domwalk.c (dom_walker::dom_walker): Likewise.
20259 (dom_walker::walk): Handle NULL RPO order.
20260 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
20262 (rewrite_update_dom_walker): Likewise.
20263 (mark_def_dom_walker): Likewise.
20265 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20267 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
20268 (aarch64_maybe_expand_sve_subreg_move): Declare.
20269 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
20270 * config/aarch64/predicates.md (aarch64_any_register_operand): New
20272 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
20273 that are semantically a reverse operation.
20274 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
20275 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
20276 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
20278 (aarch64_can_change_mode_class): For big-endian, forbid changes
20279 between two SVE modes if they have different element sizes.
20281 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20283 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
20284 the TImode handling for big-endian targets.
20286 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20288 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
20289 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
20291 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
20292 Remove BSWAP handing for big-endian targets and use the form of
20293 LD1RQ appropariate for the mode.
20295 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20297 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
20298 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
20299 duplicated element.
20301 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20304 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
20305 check for operands that need to go through aarch64_sve_reload_be.
20307 2018-02-01 Jakub Jelinek <jakub@redhat.com>
20309 PR tree-optimization/81661
20310 PR tree-optimization/84117
20311 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
20312 * tree-eh.c: Include gimplify.h.
20313 (find_trapping_overflow, replace_trapping_overflow,
20314 rewrite_to_non_trapping_overflow): New functions.
20315 * tree-vect-loop.c: Include tree-eh.h.
20316 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
20317 * tree-data-ref.c: Include tree-eh.h.
20318 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
20320 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
20322 PR rtl-optimization/84123
20323 * combine.c (change_zero_ext): Check if hard register satisfies
20324 can_change_dest_mode before calling gen_lowpart_SUBREG.
20326 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
20329 * ira.c (ira_init_register_move_cost): Remove assert.
20331 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
20333 PR rtl-optimization/84071
20334 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
20335 * doc/tm.texi: Regenerate.
20337 2018-01-31 Richard Biener <rguenther@suse.de>
20339 PR tree-optimization/84132
20340 * tree-data-ref.c (analyze_miv_subscript): Properly
20341 check whether evolution_function_is_affine_multivariate_p
20342 before calling gcd_of_steps_may_divide_p.
20344 2018-01-31 Julia Koval <julia.koval@intel.com>
20347 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
20348 * config/i386/i386.md (rdpid_rex64) New.
20349 (rdpid): Make 32bit only.
20351 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
20354 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
20355 an IDENTIFIER_NODE for FUNCTION_TYPE's.
20357 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
20360 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
20362 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
20364 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
20366 PR rtl-optimization/84071
20367 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
20368 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
20370 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
20372 * config/arc/arc.c (arc_handle_aux_attribute): New function.
20373 (arc_attribute_table): Add 'aux' attribute.
20374 (arc_in_small_data_p): Consider aux like variables.
20375 (arc_is_aux_reg_p): New function.
20376 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
20377 (arc_get_aux_arg): New function.
20378 (prepare_move_operands): Handle aux-register access.
20379 (arc_handle_aux_attribute): New function.
20380 * doc/extend.texi (ARC Variable attributes): Add subsection.
20382 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
20384 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
20385 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
20386 (arc_attribute_table): Add 'uncached' attribute.
20387 (arc_print_operand): Print '.di' flag for uncached memory
20389 (arc_in_small_data_p): Do not consider for small data the uncached
20391 (arc_is_uncached_mem_p): New function.
20392 * config/arc/predicates.md (compact_store_memory_operand): Check
20393 for uncached memory accesses.
20394 (nonvol_nonimm_operand): Likewise.
20395 * doc/extend.texi (ARC Type Attribute): New subsection.
20397 2018-01-31 Jakub Jelinek <jakub@redhat.com>
20400 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
20401 falign-loops=): Add Optimization flag.
20403 2018-01-30 Jeff Law <law@redhat.com>
20406 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
20407 INT_REGISTERS_SAVED. Check it prior to calling
20408 get_scratch_register_on_entry.
20409 (ix86_adjust_stack_and_probe): Similarly.
20410 (ix86_emit_probe_stack_range): Similarly.
20411 (ix86_expand_prologue): Corresponding changes.
20413 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20416 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
20417 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
20419 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
20422 * lra-constraints.c (curr_insn_transform): Process AND in the
20425 2018-01-30 Jakub Jelinek <jakub@redhat.com>
20427 PR rtl-optimization/83986
20428 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
20429 dependence against last_pending_memory_flush in addition to
20430 pending_jump_insns.
20432 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
20434 PR tree-optimization/81611
20435 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
20438 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20441 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
20444 2018-01-30 Richard Biener <rguenther@suse.de>
20445 Jakub Jelinek <jakub@redhat.com>
20447 PR tree-optimization/84111
20448 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
20449 inner loops added during recursion, as they don't have up-to-date
20452 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
20455 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
20456 (can_inline_edge_by_limits_p): ... here.
20457 (can_early_inline_edge_p, check_callers,
20458 update_caller_keys, update_callee_keys, recursive_inlining,
20459 add_new_edges_to_heap, speculation_useful_p,
20460 inline_small_functions,
20461 inline_small_functions, flatten_function,
20462 inline_to_all_callers_1): Update.
20464 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
20466 * profile-count.c (profile_count::combine_with_ipa_count): Handle
20469 2018-01-30 Richard Biener <rguenther@suse.de>
20471 PR tree-optimization/83008
20472 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
20473 invariant and constant vector uses in stmts when they need
20474 more than one stmt.
20476 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20479 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
20480 * configure: Regenerate.
20482 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
20484 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
20486 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
20487 Use gen_rtx_REG rather than gen_lowpart.
20489 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
20491 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
20492 rather than 0 when creating partial subregs.
20494 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
20496 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
20499 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
20502 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
20503 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
20504 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
20505 flags. This restores the settings used before the 2017-07-24.
20506 Turning off pre increment/decrement/modify allows IVOPTS to
20507 optimize DF/SF loops where the index is an int.
20509 2018-01-29 Richard Biener <rguenther@suse.de>
20510 Kelvin Nilsen <kelvin@gcc.gnu.org>
20513 * tree-vect-stmts.c (vectorizable_call): Don't call
20514 targetm.vectorize_builtin_md_vectorized_function if callee is
20517 2018-01-22 Carl Love <cel@us.ibm.com>
20519 * doc/extend.tex: Fix typo in second arg in
20520 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
20522 2018-01-29 Richard Biener <rguenther@suse.de>
20524 PR tree-optimization/84086
20525 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
20526 (flush_ssaname_freelist): When SSA names were released reset
20527 the SCEV hash table.
20529 2018-01-29 Richard Biener <rguenther@suse.de>
20531 PR tree-optimization/84057
20532 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
20533 removed paths when removing edges.
20535 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
20537 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
20538 -mfunction-return=@var{choice}.
20540 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
20542 PR diagnostic/84034
20543 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
20544 Handle CR like TAB.
20545 (layout::print_source_line): Likewise.
20546 (test_get_line_width_without_trailing_whitespace): Add test cases.
20548 2018-01-27 Jakub Jelinek <jakub@redhat.com>
20550 PR middle-end/84040
20551 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
20554 2018-01-26 Jim Wilson <jimw@sifive.com>
20556 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
20558 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
20561 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20563 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
20564 and CMP + SUB-immediate -> SUBS.
20566 2018-01-26 Martin Sebor <msebor@redhat.com>
20568 PR tree-optimization/83896
20569 * tree-ssa-strlen.c (get_string_len): Rename...
20570 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
20571 Avoid assuming length is constant.
20572 (handle_char_store): Use HOST_WIDE_INT for string length.
20574 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
20577 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
20578 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
20580 2018-01-26 Richard Biener <rguenther@suse.de>
20582 PR rtl-optimization/84003
20583 * dse.c (record_store): Only record redundant stores when
20584 the earlier store aliases at least all accesses the later one does.
20586 2018-01-26 Jakub Jelinek <jakub@redhat.com>
20588 PR rtl-optimization/83985
20589 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
20590 REG_CFA_RESTORE insns.
20591 (delete_unmarked_insns): Don't ignore separate shrink wrapping
20592 REG_CFA_RESTORE insns here.
20595 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
20596 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
20598 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20600 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
20601 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
20602 (arc_init): Likewise.
20603 (arc_override_options): Likewise.
20604 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
20606 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
20608 * config/arc/arc.h (TARGET_DBNZ): Define.
20609 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
20610 properly set the tune attribute.
20611 (dbnz): Use TARGET_DBNZ guard.
20612 * config/arc/arc.opt (mtune): Add core3 option.
20614 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20616 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
20617 recognize new pic like addresses.
20618 (arc_delegitimize_address): Clean up.
20620 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20622 * config/arc/arc-arches.def: Option mrf16 valid for all
20624 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
20625 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
20626 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
20627 * config/arc/arc-tables.opt: Regenerate.
20628 * config/arc/arc.c (arc_conditional_register_usage): Handle
20629 reduced register file case.
20630 (arc_file_start): Set must have build attributes.
20631 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
20632 mrf16 option value.
20633 * config/arc/arc.opt (mrf16): Add new option.
20634 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
20635 * config/arc/genmultilib.awk: Handle new mrf16 option.
20636 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
20637 * config/arc/t-multilib: Regenerate.
20638 * doc/invoke.texi (ARC Options): Document mrf16 option.
20640 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20642 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
20643 * config/arc/arc.c (arc_handle_secure_attribute): New function.
20644 (arc_attribute_table): Add 'secure_call' attribute.
20645 (arc_print_operand): Print secure call operand.
20646 (arc_function_ok_for_sibcall): Don't optimize tail calls when
20648 (arc_is_secure_call_p): New function. * config/arc/arc.md
20649 (call_i): Add support for sjli instruction.
20650 (call_value_i): Likewise.
20651 * config/arc/constraints.md (Csc): New constraint.
20653 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20654 John Eric Martin <John.Martin@emmicro-us.com>
20656 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
20657 * config/arc/arc.c (_arc_jli_section): New struct.
20658 (arc_jli_section): New type.
20659 (rc_jli_sections): New static variable.
20660 (arc_handle_jli_attribute): New function.
20661 (arc_attribute_table): Add jli_always and jli_fixed attribute.
20662 (arc_file_end): New function.
20663 (TARGET_ASM_FILE_END): Define.
20664 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
20665 (arc_add_jli_section): New function.
20666 (jli_call_scan): Likewise.
20667 (arc_reorg): Call jli_call_scan.
20668 (arc_output_addsi): Remove 'S' from printing asm operand.
20669 (arc_is_jli_call_p): New function.
20670 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
20672 (movhi_insn): Likewise.
20673 (movsi_insn): Likewise.
20674 (movsi_set_cc_insn): Likewise.
20675 (loadqi_update): Likewise.
20676 (load_zeroextendqisi_update): Likewise.
20677 (load_signextendqisi_update): Likewise.
20678 (loadhi_update): Likewise.
20679 (load_zeroextendhisi_update): Likewise.
20680 (load_signextendhisi_update): Likewise.
20681 (loadsi_update): Likewise.
20682 (loadsf_update): Likewise.
20683 (movsicc_insn): Likewise.
20684 (bset_insn): Likewise.
20685 (bxor_insn): Likewise.
20686 (bclr_insn): Likewise.
20687 (bmsk_insn): Likewise.
20688 (bicsi3_insn): Likewise.
20689 (cmpsi_cc_c_insn): Likewise.
20690 (movsi_ne): Likewise.
20691 (movsi_cond_exec): Likewise.
20692 (clrsbsi2): Likewise.
20693 (norm_f): Likewise.
20703 (call_i): Remove 'S' asm letter, add jli instruction.
20704 (call_value_i): Likewise.
20705 * config/arc/arc.op (mjli-always): New option.
20706 * config/arc/constraints.md (Cji): New constraint.
20707 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
20709 (subsf3_fpx): Likewise.
20710 (mulsf3_fpx): Likewise.
20711 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
20713 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
20714 function attrbutes.
20715 * doc/invoke.texi (ARC): Document mjli-always option.
20717 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
20719 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
20720 avoid addition with 0 and use incw and decw where possible.
20722 2018-01-26 Richard Biener <rguenther@suse.de>
20724 PR tree-optimization/81082
20725 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
20726 association if it requires casting to unsigned.
20727 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
20728 from fold_plusminus_mult_expr to catch important cases late when
20729 range info is available.
20731 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20733 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
20734 * configure.ac (hidden_linkonce): New test.
20735 * configure: Regenerate.
20736 * config.in: Regenerate.
20738 2018-01-26 Julia Koval <julia.koval@intel.com>
20740 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
20741 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
20742 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
20743 _mm_mask_bitshuffle_epi64_mask): Fix type.
20744 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
20745 USI_FTYPE_V4DI_V4DI_USI): Remove.
20746 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
20747 __builtin_ia32_vpshufbitqmb256_mask,
20748 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
20749 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
20750 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
20752 2018-01-26 Alan Modra <amodra@gmail.com>
20755 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
20756 UNSPEC_VBPERMQ. Sort other unspecs.
20758 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
20760 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
20762 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
20764 PR middle-end/83055
20765 * predict.c (drop_profile): Do not push/pop cfun; update also
20767 (handle_missing_profiles): Fix logic looking for zero profiles.
20769 2018-01-25 Jakub Jelinek <jakub@redhat.com>
20771 PR middle-end/83977
20772 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
20773 on functions with #pragma omp declare simd or functions with simd
20775 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
20776 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
20777 Remove trailing \n from warning_at calls.
20779 2018-01-25 Tom de Vries <tom@codesourcery.com>
20782 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
20783 for neutered workers.
20785 2018-01-24 Joseph Myers <joseph@codesourcery.com>
20788 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
20789 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
20791 2018-01-24 Jeff Law <law@redhat.com>
20794 * i386.c (get_probe_interval): Move to earlier point.
20795 (ix86_compute_frame_layout): If -fstack-clash-protection and
20796 the frame is larger than the probe interval, then use pushes
20797 to save registers rather than reg->mem moves.
20798 (ix86_expand_prologue): Remove conditional for int_registers_saved
20801 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
20804 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
20805 min/max for never referenced object.
20807 2018-01-24 Jakub Jelinek <jakub@redhat.com>
20809 PR middle-end/83977
20810 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
20812 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
20813 attributes from DECL_ATTRIBUTES (decl) without affecting
20814 DECL_ATTRIBUTES (current_function_decl).
20815 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
20816 functions with non-NULL DECL_ABSTRACT_ORIGIN.
20818 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
20820 PR tree-optimization/83979
20821 * fold-const.c (fold_comparison): Use constant_boolean_node
20822 instead of boolean_{true,false}_node.
20824 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
20826 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
20829 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20831 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
20832 Simplify the clause that sets the length attribute.
20833 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
20834 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
20835 clause that sets the length attribute.
20836 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
20838 2018-01-24 Tom de Vries <tom@codesourcery.com>
20841 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
20842 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
20843 Add strict parameter.
20844 (prevent_branch_around_nothing): Insert dummy insn between branch to
20845 label and label with no ptx insn inbetween.
20846 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
20848 2018-01-24 Tom de Vries <tom@codesourcery.com>
20851 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
20852 for neutered threads in warp.
20853 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
20855 2018-01-24 Richard Biener <rguenther@suse.de>
20857 PR tree-optimization/83176
20858 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
20861 2018-01-24 Richard Biener <rguenther@suse.de>
20863 PR tree-optimization/82819
20864 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
20865 code generating pluses that are no-ops in the target precision.
20867 2018-01-24 Richard Biener <rguenther@suse.de>
20869 PR middle-end/84000
20870 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
20872 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
20874 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
20875 to merge probabilities.
20876 * predict.c (probably_never_executed): Also mark as cold functions
20877 with global 0 profile and guessed local profile.
20878 * profile-count.c (profile_probability::combine_with_count): New
20880 * profile-count.h (profile_probability::operator*,
20881 profile_probability::operator*=, profile_probability::operator/,
20882 profile_probability::operator/=): Reduce precision to adjusted
20883 and set value to guessed on contradictory divisions.
20884 (profile_probability::combine_with_freq): Remove.
20885 (profile_probability::combine_wiht_count): Declare.
20886 (profile_count::force_nonzero):: Set to adjusted.
20887 (profile_count::probability_in):: Set quality to adjusted.
20888 * tree-ssa-tail-merge.c (replace_block_by): Use
20889 combine_with_count.
20891 2018-01-23 Andrew Waterman <andrew@sifive.com>
20892 Jim Wilson <jimw@sifive.com>
20894 * config/riscv/riscv.c (riscv_stack_boundary): New.
20895 (riscv_option_override): Set riscv_stack_boundary. Handle
20896 riscv_preferred_stack_boundary_arg.
20897 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
20898 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
20899 (STACK_BOUNDARY): Set to riscv_stack_boundary.
20900 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
20901 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
20902 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
20904 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
20907 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
20908 of struct ix86_frame.
20909 (ix86_expand_epilogue): Likewise. Add a local variable for
20910 the reg_save_offset field in struct ix86_frame.
20912 2018-01-23 Bin Cheng <bin.cheng@arm.com>
20914 PR tree-optimization/82604
20915 * tree-loop-distribution.c (enum partition_kind): New enum item
20916 PKIND_PARTIAL_MEMSET.
20917 (partition_builtin_p): Support above new enum item.
20918 (generate_code_for_partition): Ditto.
20919 (compute_access_range): Differentiate cases that equality can be
20920 proven at all loops, the innermost loops or no loops.
20921 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
20922 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
20923 (finalize_partitions, distribute_loop): Don't fuse partition of
20924 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
20925 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
20926 parloop is enabled.
20928 2018-01-23 Martin Liska <mliska@suse.cz>
20930 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
20931 order to ignore the predictor.
20932 (PRED_POLYMORPHIC_CALL): Likewise.
20933 (PRED_RECURSIVE_CALL): Likewise.
20935 2018-01-23 Martin Liska <mliska@suse.cz>
20937 * tree-profile.c (tree_profiling): Print function header to
20938 aware reader which function we are working on.
20939 * value-prof.c (gimple_find_values_to_profile): Do not print
20940 not interesting value histograms.
20942 2018-01-23 Martin Liska <mliska@suse.cz>
20944 * profile-count.h (enum profile_quality): Add
20945 profile_uninitialized as the first value. Do not number values
20946 as they are zero based.
20947 (profile_count::verify): Update sanity check.
20948 (profile_probability::verify): Likewise.
20950 2018-01-23 Nathan Sidwell <nathan@acm.org>
20952 * doc/invoke.texi (ffor-scope): Deprecate.
20954 2018-01-23 David Malcolm <dmalcolm@redhat.com>
20956 PR tree-optimization/83510
20957 * domwalk.c (set_all_edges_as_executable): New function.
20958 (dom_walker::dom_walker): Convert bool param
20959 "skip_unreachable_blocks" to enum reachability. Move setup of
20960 edge flags to set_all_edges_as_executable and only do it when
20961 reachability is REACHABLE_BLOCKS.
20962 * domwalk.h (enum dom_walker::reachability): New enum.
20963 (dom_walker::dom_walker): Convert bool param
20964 "skip_unreachable_blocks" to enum reachability.
20965 (set_all_edges_as_executable): New decl.
20966 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
20967 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
20969 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
20970 but converting true to REACHABLE_BLOCKS.
20971 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
20973 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
20974 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
20975 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
20977 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
20978 if check_all_array_refs will be called.
20980 2018-01-23 David Malcolm <dmalcolm@redhat.com>
20982 * tree.c (selftest::test_location_wrappers): Add more test
20985 2018-01-23 David Malcolm <dmalcolm@redhat.com>
20987 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
20988 (selftest::test_bit_in_range): Likewise.
20990 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
20993 * doc/sourcebuild.texi (vect_float): Say that the selector
20994 only describes the situation when -funsafe-math-optimizations is on.
20995 (vect_float_strict): Document.
20997 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
20999 PR tree-optimization/83965
21000 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
21001 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
21002 instead of checking only for a reduction.
21003 (vect_recog_widen_sum_pattern): Likewise.
21005 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
21007 * predict.c (probably_never_executed): Only use precise profile info.
21008 (compute_function_frequency): Skip after inlining hack since we now
21009 have quality checking.
21011 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
21013 * profile-count.h (profile_probability::very_unlikely,
21014 profile_probability::unlikely, profile_probability::even): Set
21015 precision to guessed.
21017 2018-01-23 Richard Biener <rguenther@suse.de>
21019 PR tree-optimization/83963
21020 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
21021 Properly terminate dominator walk when crossing the exit edge not
21022 when visiting its source block.
21024 2018-01-23 Jakub Jelinek <jakub@redhat.com>
21027 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
21028 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
21030 2018-01-22 Jakub Jelinek <jakub@redhat.com>
21032 PR tree-optimization/83957
21033 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
21034 semicolon after for body surrounded by braces.
21036 PR tree-optimization/83081
21037 * profile-count.h (profile_probability::split): New method.
21038 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
21039 Use profile_probability::split.
21040 (do_compare_rtx_and_jump): Fix adjustment of probabilities
21041 when splitting a single conditional jump into 2.
21043 2018-01-22 David Malcolm <dmalcolm@redhat.com>
21045 PR tree-optimization/69452
21046 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
21049 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21051 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
21052 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
21053 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
21055 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21057 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
21059 * config/rl78/rl78.md (movdi): New define_expand.
21060 * config/rl78/rl78.c (rl78_split_movdi): New function.
21062 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
21065 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
21067 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
21068 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
21069 128-bit to produce an UNSPEC move to get the double word with the
21070 signbit and then a shift directly to do signbit.
21071 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
21072 implementation with a new version that just does either a direct
21073 move or a regular move. Move memory interface to separate insns.
21074 Move insns so they are next to the expander.
21075 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
21076 with signbit move. Split big and little endian case.
21077 (signbit<mode>2_dm_mem_le): Likewise.
21078 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
21079 (signbit<mode>2_dm2): Likewise.
21081 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21083 * config/rl78/rl78.md (anddi3): New define_expand.
21085 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21087 * config/rl78/rl78.md (umindi3): New define_expand.
21089 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21091 * config/rl78/rl78.md (smindi3): New define_expand.
21093 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21095 * config/rl78/rl78.md (smaxdi3): New define_expand.
21097 2018-01-22 Carl Love <cel@us.ibm.com>
21099 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
21100 LVX_V1TI): Add macro expansion.
21101 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
21102 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
21103 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
21104 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
21105 Change check to determine if the instruction is a byte reversing
21106 entry. Fix typo in comment.
21107 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
21108 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
21109 Add def_builtin calls for new builtins.
21110 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
21111 Add define_insn expansion.
21113 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21115 * config/rl78/rl78.md (umaxdi3): New define_expand.
21117 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21119 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
21120 for non-QImode registers.
21122 2018-01-22 Richard Biener <rguenther@suse.de>
21124 PR tree-optimization/83963
21125 * graphite-scop-detection.c (scop_detection::get_sese): Delay
21126 including the loop exit block.
21127 (scop_detection::merge_sese): Likewise.
21128 (scop_detection::add_scop): Do it here instead.
21130 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21132 * doc/sourcebuild.texi (arm_softfloat): Document.
21134 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
21137 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
21138 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
21139 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
21141 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21142 David Edelsohn <dje.gcc@gmail.com>
21145 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
21146 Change "crset eq" to "crset 2".
21147 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
21148 (*call_indirect_aix<mode>_nospec): Likewise.
21149 (*call_value_indirect_aix<mode>_nospec): Likewise.
21150 (*call_indirect_elfv2<mode>_nospec): Likewise.
21151 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
21152 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
21153 change assembly output from . to $.
21154 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
21155 (indirect_jump<mode>_nospec): Change assembly output from . to $.
21156 (*tablejump<mode>_internal1_nospec): Likewise.
21158 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
21161 * config/sh/sh_optimize_sett_clrt.cc:
21162 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
21164 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
21166 PR tree-optimization/83940
21167 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
21168 offset_dt to vect_constant_def rather than vect_unknown_def_type.
21169 (vect_check_load_store_mask): Add a mask_dt_out parameter and
21170 use it to pass back the definition type.
21171 (vect_check_store_rhs): Likewise rhs_dt_out.
21172 (vect_build_gather_load_calls): Add a mask_dt argument and use
21173 it instead of a call to vect_is_simple_use.
21174 (vectorizable_store): Update calls to vect_check_load_store_mask
21175 and vect_check_store_rhs. Use the dt returned by the latter instead
21176 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
21177 instead of calls to vect_is_simple_use. Pass the scalar rather
21178 than the vector operand to vect_is_simple_use when handling
21179 second and subsequent copies of an rhs value.
21180 (vectorizable_load): Update calls to vect_check_load_store_mask
21181 and vect_build_gather_load_calls. Use the cached mask_dt and
21182 gs_info.offset_dt instead of calls to vect_is_simple_use.
21184 2018-01-20 Jakub Jelinek <jakub@redhat.com>
21186 PR middle-end/83945
21187 * tree-emutls.c: Include gimplify.h.
21188 (lower_emutls_2): New function.
21189 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
21190 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
21191 it before further processing.
21194 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
21195 UINTVAL (trueop1) instead of INTVAL (op1).
21197 2018-01-19 Jakub Jelinek <jakub@redhat.com>
21201 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
21202 INCOMING_FRAME_SP_OFFSET if not defined.
21203 (scan_trace): Add ENTRY argument. If true and
21204 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
21205 emit a note to adjust the CFA offset.
21206 (create_cfi_notes): Adjust scan_trace callers.
21207 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
21208 INCOMING_FRAME_SP_OFFSET in the CIE.
21209 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
21210 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
21212 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
21213 * doc/tm.texi: Regenerated.
21215 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
21217 PR rtl-optimization/83147
21218 * lra-constraints.c (remove_inheritance_pseudos): Use
21219 lra_substitute_pseudo_within_insn.
21221 2018-01-19 Tom de Vries <tom@codesourcery.com>
21222 Cesar Philippidis <cesar@codesourcery.com>
21225 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
21227 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
21230 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
21231 spaces for function labels.
21233 2018-01-19 Martin Liska <mliska@suse.cz>
21235 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
21236 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
21237 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
21238 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
21239 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
21240 (PRED_CONST_RETURN): Change from 69 to 65.
21241 (PRED_NULL_RETURN): Change from 91 to 71.
21242 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
21243 (PRED_LOOP_GUARD): Change from 66 to 73.
21245 2018-01-19 Martin Liska <mliska@suse.cz>
21247 * predict.c (predict_insn_def): Add new assert.
21248 (struct branch_predictor): Change type to signed integer.
21249 (test_prediction_value_range): Amend test to cover
21250 PROB_UNINITIALIZED.
21251 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
21252 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
21253 (PRED_LOOP_ITERATIONS_MAX): Likewise.
21254 (PRED_LOOP_IV_COMPARE): Likewise.
21255 * predict.h (PROB_UNINITIALIZED): Define new constant.
21257 2018-01-19 Martin Liska <mliska@suse.cz>
21259 * predict.c (dump_prediction): Add new format for
21260 analyze_brprob.py script which is enabled with -details
21262 * profile-count.h (precise_p): New function.
21264 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
21266 PR tree-optimization/83922
21267 * tree-vect-loop.c (vect_verify_full_masking): Return false if
21268 there are no statements that need masking.
21269 (vect_active_double_reduction_p): New function.
21270 (vect_analyze_loop_operations): Use it when handling phis that
21271 are not in the loop header.
21273 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
21275 PR tree-optimization/83914
21276 * tree-vect-loop.c (vectorizable_induction): Don't convert
21277 init_expr or apply the peeling adjustment for inductions
21278 that are nested within the vectorized loop.
21280 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21282 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
21285 2018-01-18 Jakub Jelinek <jakub@redhat.com>
21289 * function.h (gimplify_parameters): Add gimple_seq * argument.
21290 * function.c: Include gimple.h and options.h.
21291 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
21292 for the added local temporaries if needed.
21293 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
21294 if there are any parameter cleanups, wrap whole body into a
21295 try/finally with the cleanups.
21297 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
21300 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
21301 Use GET_MODE_CLASS for scalar floating point.
21303 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
21307 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
21308 Fix call of call_cgraph_insertion_hooks.
21310 2018-01-18 Martin Sebor <msebor@redhat.com>
21312 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
21314 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
21317 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
21320 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
21323 * common.opt: (-ffile-prefix-map): New option.
21324 * opts.c (common_handle_option): Defer it.
21325 * opts-global.c (handle_common_deferred_options): Handle it.
21326 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
21327 * file-prefix-map.h: New file.
21328 (remap_debug_filename, add_debug_prefix_map): ...here.
21329 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
21330 * final.c (debug_prefix_map, add_debug_prefix_map
21331 remap_debug_filename): Move to...
21332 * file-prefix-map.c: New file.
21333 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
21334 generalize, get rid of alloca(), use strrchr() instead of strchr().
21335 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
21336 Implement in terms of add_prefix_map().
21337 (remap_macro_filename, remap_debug_filename): Implement in term of
21339 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
21340 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
21341 * dbxout.c: Include file-prefix-map.h.
21342 * varasm.c: Likewise.
21343 * vmsdbgout.c: Likewise.
21344 * xcoffout.c: Likewise.
21345 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
21346 * doc/cppopts.texi (-fmacro-prefix-map): Document.
21347 * doc/invoke.texi (-ffile-prefix-map): Document.
21348 (-fdebug-prefix-map): Update description.
21350 2018-01-18 Martin Liska <mliska@suse.cz>
21352 * config/i386/i386.c (indirect_thunk_name): Document that also
21354 (output_indirect_thunk): Document why both instructions
21355 (pause and lfence) are generated.
21357 2018-01-18 Richard Biener <rguenther@suse.de>
21359 PR tree-optimization/83887
21360 * graphite-scop-detection.c
21361 (scop_detection::get_nearest_dom_with_single_entry): Remove.
21362 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
21363 (scop_detection::merge_sese): Re-implement with a flood-fill
21364 algorithm that properly finds a SESE region if it exists.
21366 2018-01-18 Jakub Jelinek <jakub@redhat.com>
21369 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
21370 pointer_diff optimizations use view_convert instead of convert.
21372 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21374 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
21375 Generate different code for -mno-speculate-indirect-jumps.
21376 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
21377 (*call_indirect_aix<mode>): Disable for
21378 -mno-speculate-indirect-jumps.
21379 (*call_indirect_aix<mode>_nospec): New define_insn.
21380 (*call_value_indirect_aix<mode>): Disable for
21381 -mno-speculate-indirect-jumps.
21382 (*call_value_indirect_aix<mode>_nospec): New define_insn.
21383 (*sibcall_nonlocal_sysv<mode>): Generate different code for
21384 -mno-speculate-indirect-jumps.
21385 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
21387 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
21389 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
21390 long double type, set the flags for noting the default long double
21391 type, even if we don't pass or return a long double type.
21393 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
21396 * ipa-inline.c (flatten_function): Do not overwrite final inlining
21399 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
21401 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
21402 support for merge[hl].
21403 (fold_mergehl_helper): New helper function.
21404 (tree-vector-builder.h): New #include for tree_vector_builder usage.
21405 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
21406 (altivec_vmrglw_direct): Add xxmrglw insn.
21408 2018-01-17 Andrew Waterman <andrew@sifive.com>
21410 * config/riscv/riscv.c (riscv_conditional_register_usage): If
21411 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
21413 2018-01-17 David Malcolm <dmalcolm@redhat.com>
21416 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
21417 call the lto_location_cache before reading the
21418 DECL_SOURCE_LOCATION of the types.
21420 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
21421 Richard Sandiford <richard.sandiford@linaro.org>
21423 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
21424 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
21425 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
21426 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
21427 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
21429 * config/aarch64/constraints.md (aarch64_movti_operand):
21431 * config/aarch64/predicates.md (Uti): Add new constraint.
21433 2018-01-17 Carl Love <cel@us.ibm.com>
21435 * config/rs6000/vsx.md (define_expand xl_len_r,
21436 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
21437 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
21439 (define_expand, define_insn): Move the shift left from the
21440 define_insn to the define_expand for lxvl and stxvl instructions.
21441 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
21442 and XL_LEN_R definitions to PURE.
21444 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
21446 * config/i386/i386.c (indirect_thunk_name): Declare regno
21447 as unsigned int. Compare regno with INVALID_REGNUM.
21448 (output_indirect_thunk): Ditto.
21449 (output_indirect_thunk_function): Ditto.
21450 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
21451 in the call to output_indirect_thunk_function.
21453 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
21455 PR middle-end/83884
21456 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
21457 rather than the size of inner_type to determine the stack slot size
21458 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
21460 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
21463 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
21466 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
21468 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
21469 endian Linux systems to optionally enable multilibs for selecting
21470 the long double type if the user configured an explicit type.
21471 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
21472 have no long double multilibs if not defined.
21473 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
21474 warn if the user used -mabi={ieee,ibm}longdouble and we built
21475 multilibs for long double.
21476 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
21477 appropriate multilib option.
21478 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
21480 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
21481 for building long double multilibs.
21482 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
21484 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
21486 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
21489 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
21491 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
21494 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
21497 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
21500 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
21502 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
21503 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
21505 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
21507 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
21508 different rtl trees depending on TARGET_64BIT.
21509 (rs6000_gen_lvx): Likewise.
21511 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
21513 * config/visium/visium.md (nop): Tweak comment.
21514 (hazard_nop): Likewise.
21516 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21518 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
21519 -mspeculate-indirect-jumps.
21520 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
21521 for -mno-speculate-indirect-jumps.
21522 (*call_indirect_elfv2<mode>_nospec): New define_insn.
21523 (*call_value_indirect_elfv2<mode>): Disable for
21524 -mno-speculate-indirect-jumps.
21525 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
21526 (indirect_jump): Emit different RTL for
21527 -mno-speculate-indirect-jumps.
21528 (*indirect_jump<mode>): Disable for
21529 -mno-speculate-indirect-jumps.
21530 (*indirect_jump<mode>_nospec): New define_insn.
21531 (tablejump): Emit different RTL for
21532 -mno-speculate-indirect-jumps.
21533 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
21534 (tablejumpsi_nospec): New define_expand.
21535 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
21536 (tablejumpdi_nospec): New define_expand.
21537 (*tablejump<mode>_internal1): Disable for
21538 -mno-speculate-indirect-jumps.
21539 (*tablejump<mode>_internal1_nospec): New define_insn.
21540 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
21543 2018-01-16 Artyom Skrobov tyomitch@gmail.com
21545 * caller-save.c (insert_save): Drop unnecessary parameter. All
21548 2018-01-16 Jakub Jelinek <jakub@redhat.com>
21549 Richard Biener <rguenth@suse.de>
21552 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
21553 return early, inline manually is_gimple_sizepos. Make sure if we
21554 call gimplify_expr we don't end up with a gimple constant.
21555 * tree.c (variably_modified_type_p): Don't return true for
21556 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
21557 * gimplify.h (is_gimple_sizepos): Remove.
21559 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21561 PR tree-optimization/83857
21562 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
21563 vectorizable_live_operation for pure SLP statements.
21564 (vectorizable_live_operation): Handle PHIs.
21566 2018-01-16 Richard Biener <rguenther@suse.de>
21568 PR tree-optimization/83867
21569 * tree-vect-stmts.c (vect_transform_stmt): Precompute
21570 nested_in_vect_loop_p since the scalar stmt may get invalidated.
21572 2018-01-16 Jakub Jelinek <jakub@redhat.com>
21575 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
21576 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
21577 If off is not INTEGER_CST, issue a may not be aligned warning
21578 rather than isn't aligned. Use isn%'t rather than isn't.
21579 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
21581 <case MULT_EXPR>: Improve the case when bottom and one of the
21582 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
21583 operand, in that case check if the other operand is multiple of
21584 bottom divided by the INTEGER_CST operand.
21586 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21589 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
21590 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
21591 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
21592 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
21593 * config/pa/pa.c (pa_function_arg_advance): Likewise.
21594 (pa_function_arg, pa_arg_partial_bytes): Likewise.
21595 (pa_function_arg_size): New function.
21597 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21599 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
21600 in a separate statement.
21602 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21604 PR tree-optimization/83847
21605 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
21606 group gathers and scatters.
21608 2018-01-16 Jakub Jelinek <jakub@redhat.com>
21610 PR rtl-optimization/86620
21611 * params.def (max-sched-ready-insns): Bump minimum value to 1.
21613 PR rtl-optimization/83213
21614 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
21615 to last if both are JUMP_INSNs.
21617 PR tree-optimization/83843
21618 * gimple-ssa-store-merging.c
21619 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
21620 store_immediate_info for bswap/nop orig_stores.
21622 2018-01-15 Andrew Waterman <andrew@sifive.com>
21624 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
21626 <UDIV>: Increase cost if !TARGET_DIV.
21628 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
21630 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
21631 (define_attr "cr_logical_3op"): New.
21632 (cceq_ior_compare): Adjust.
21633 (cceq_ior_compare_complement): Adjust.
21634 (*cceq_rev_compare): Adjust.
21635 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
21636 (is_cracked_insn): Adjust.
21637 (insn_must_be_first_in_group): Adjust.
21638 * config/rs6000/40x.md: Adjust.
21639 * config/rs6000/440.md: Adjust.
21640 * config/rs6000/476.md: Adjust.
21641 * config/rs6000/601.md: Adjust.
21642 * config/rs6000/603.md: Adjust.
21643 * config/rs6000/6xx.md: Adjust.
21644 * config/rs6000/7450.md: Adjust.
21645 * config/rs6000/7xx.md: Adjust.
21646 * config/rs6000/8540.md: Adjust.
21647 * config/rs6000/cell.md: Adjust.
21648 * config/rs6000/e300c2c3.md: Adjust.
21649 * config/rs6000/e500mc.md: Adjust.
21650 * config/rs6000/e500mc64.md: Adjust.
21651 * config/rs6000/e5500.md: Adjust.
21652 * config/rs6000/e6500.md: Adjust.
21653 * config/rs6000/mpc.md: Adjust.
21654 * config/rs6000/power4.md: Adjust.
21655 * config/rs6000/power5.md: Adjust.
21656 * config/rs6000/power6.md: Adjust.
21657 * config/rs6000/power7.md: Adjust.
21658 * config/rs6000/power8.md: Adjust.
21659 * config/rs6000/power9.md: Adjust.
21660 * config/rs6000/rs64.md: Adjust.
21661 * config/rs6000/titan.md: Adjust.
21663 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21665 * config/i386/predicates.md (indirect_branch_operand): Rewrite
21666 ix86_indirect_branch_register logic.
21668 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21670 * config/i386/constraints.md (Bs): Update
21671 ix86_indirect_branch_register check. Don't check
21672 ix86_indirect_branch_register with GOT_memory_operand.
21674 * config/i386/predicates.md (GOT_memory_operand): Don't check
21675 ix86_indirect_branch_register here.
21676 (GOT32_symbol_operand): Likewise.
21678 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21680 * config/i386/predicates.md (constant_call_address_operand):
21681 Rewrite ix86_indirect_branch_register logic.
21682 (sibcall_insn_operand): Likewise.
21684 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21686 * config/i386/constraints.md (Bs): Replace
21687 ix86_indirect_branch_thunk_register with
21688 ix86_indirect_branch_register.
21690 * config/i386/i386.md (indirect_jump): Likewise.
21691 (tablejump): Likewise.
21692 (*sibcall_memory): Likewise.
21693 (*sibcall_value_memory): Likewise.
21694 Peepholes of indirect call and jump via memory: Likewise.
21695 * config/i386/i386.opt: Likewise.
21696 * config/i386/predicates.md (indirect_branch_operand): Likewise.
21697 (GOT_memory_operand): Likewise.
21698 (call_insn_operand): Likewise.
21699 (sibcall_insn_operand): Likewise.
21700 (GOT32_symbol_operand): Likewise.
21702 2018-01-15 Jakub Jelinek <jakub@redhat.com>
21704 PR middle-end/83837
21705 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
21706 type rather than type addr's type points to.
21707 (expand_omp_atomic_mutex): Likewise.
21708 (expand_omp_atomic): Likewise.
21710 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21713 * config/i386/i386.c (output_indirect_thunk_function): Use
21714 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
21715 for __x86_return_thunk.
21717 2018-01-15 Richard Biener <rguenther@suse.de>
21719 PR middle-end/83850
21720 * expmed.c (extract_bit_field_1): Fix typo.
21722 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21725 * config/arm/iterators.md (VF): New mode iterator.
21726 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
21727 Remove integer-related logic from pattern.
21728 (neon_vabd<mode>_3): Likewise.
21730 2018-01-15 Jakub Jelinek <jakub@redhat.com>
21732 PR middle-end/82694
21733 * common.opt (fstrict-overflow): No longer an alias.
21734 (fwrapv-pointer): New option.
21735 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
21736 also for pointer types based on flag_wrapv_pointer.
21737 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
21738 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
21739 opts->x_flag_wrapv got set.
21740 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
21741 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
21742 POINTER_TYPE_OVERFLOW_UNDEFINED.
21743 * match.pd: Likewise in address comparison pattern.
21744 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
21746 2018-01-15 Richard Biener <rguenther@suse.de>
21749 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
21750 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
21751 Reset type names to their identifier if their TYPE_DECL doesn't
21752 have linkage (and thus is used for ODR and devirt).
21753 (save_debug_info_for_decl): Remove.
21754 (save_debug_info_for_type): Likewise.
21755 (add_tree_to_fld_list): Adjust.
21756 * tree-pretty-print.c (dump_generic_node): Make dumping of
21757 type names more robust.
21759 2018-01-15 Richard Biener <rguenther@suse.de>
21761 * BASE-VER: Bump to 8.0.1.
21763 2018-01-14 Martin Sebor <msebor@redhat.com>
21766 * builtins.c (check_access): Avoid warning when the no-warning bit
21769 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
21771 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
21772 * ira-color (allocno_hard_regs_compare): Likewise.
21774 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
21777 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
21778 Use .pushsection/.popsection.
21780 2018-01-14 Martin Sebor <msebor@redhat.com>
21783 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
21785 2018-01-14 Jakub Jelinek <jakub@redhat.com>
21787 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
21788 entry from extra_headers.
21789 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
21790 extra_headers, make the list bitwise identical to the i?86-*-* one.
21792 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21794 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
21795 -mcmodel=large with -mindirect-branch=thunk,
21796 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
21797 -mfunction-return=thunk-extern.
21798 * doc/invoke.texi: Document -mcmodel=large is incompatible with
21799 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
21800 -mfunction-return=thunk and -mfunction-return=thunk-extern.
21802 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21804 * config/i386/i386.c (print_reg): Print the name of the full
21805 integer register without '%'.
21806 (ix86_print_operand): Handle 'V'.
21807 * doc/extend.texi: Document 'V' modifier.
21809 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21811 * config/i386/constraints.md (Bs): Disallow memory operand for
21812 -mindirect-branch-register.
21814 * config/i386/predicates.md (indirect_branch_operand): Likewise.
21815 (GOT_memory_operand): Likewise.
21816 (call_insn_operand): Likewise.
21817 (sibcall_insn_operand): Likewise.
21818 (GOT32_symbol_operand): Likewise.
21819 * config/i386/i386.md (indirect_jump): Call convert_memory_address
21820 for -mindirect-branch-register.
21821 (tablejump): Likewise.
21822 (*sibcall_memory): Likewise.
21823 (*sibcall_value_memory): Likewise.
21824 Disallow peepholes of indirect call and jump via memory for
21825 -mindirect-branch-register.
21826 (*call_pop): Replace m with Bw.
21827 (*call_value_pop): Likewise.
21828 (*sibcall_pop_memory): Replace m with Bs.
21829 * config/i386/i386.opt (mindirect-branch-register): New option.
21830 * doc/invoke.texi: Document -mindirect-branch-register option.
21832 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21834 * config/i386/i386-protos.h (ix86_output_function_return): New.
21835 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
21836 set function_return_type.
21837 (indirect_thunk_name): Add ret_p to indicate thunk for function
21839 (output_indirect_thunk_function): Pass false to
21840 indirect_thunk_name.
21841 (ix86_output_indirect_branch_via_reg): Likewise.
21842 (ix86_output_indirect_branch_via_push): Likewise.
21843 (output_indirect_thunk_function): Create alias for function
21844 return thunk if regno < 0.
21845 (ix86_output_function_return): New function.
21846 (ix86_handle_fndecl_attribute): Handle function_return.
21847 (ix86_attribute_table): Add function_return.
21848 * config/i386/i386.h (machine_function): Add
21849 function_return_type.
21850 * config/i386/i386.md (simple_return_internal): Use
21851 ix86_output_function_return.
21852 (simple_return_internal_long): Likewise.
21853 * config/i386/i386.opt (mfunction-return=): New option.
21854 (indirect_branch): Mention -mfunction-return=.
21855 * doc/extend.texi: Document function_return function attribute.
21856 * doc/invoke.texi: Document -mfunction-return= option.
21858 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21860 * config/i386/i386-opts.h (indirect_branch): New.
21861 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
21862 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
21863 with local indirect jump when converting indirect call and jump.
21864 (ix86_set_indirect_branch_type): New.
21865 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
21866 (indirectlabelno): New.
21867 (indirect_thunk_needed): Likewise.
21868 (indirect_thunk_bnd_needed): Likewise.
21869 (indirect_thunks_used): Likewise.
21870 (indirect_thunks_bnd_used): Likewise.
21871 (INDIRECT_LABEL): Likewise.
21872 (indirect_thunk_name): Likewise.
21873 (output_indirect_thunk): Likewise.
21874 (output_indirect_thunk_function): Likewise.
21875 (ix86_output_indirect_branch_via_reg): Likewise.
21876 (ix86_output_indirect_branch_via_push): Likewise.
21877 (ix86_output_indirect_branch): Likewise.
21878 (ix86_output_indirect_jmp): Likewise.
21879 (ix86_code_end): Call output_indirect_thunk_function if needed.
21880 (ix86_output_call_insn): Call ix86_output_indirect_branch if
21882 (ix86_handle_fndecl_attribute): Handle indirect_branch.
21883 (ix86_attribute_table): Add indirect_branch.
21884 * config/i386/i386.h (machine_function): Add indirect_branch_type
21885 and has_local_indirect_jump.
21886 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
21888 (tablejump): Likewise.
21889 (*indirect_jump): Use ix86_output_indirect_jmp.
21890 (*tablejump_1): Likewise.
21891 (simple_return_indirect_internal): Likewise.
21892 * config/i386/i386.opt (mindirect-branch=): New option.
21893 (indirect_branch): New.
21896 (thunk-inline): Likewise.
21897 (thunk-extern): Likewise.
21898 * doc/extend.texi: Document indirect_branch function attribute.
21899 * doc/invoke.texi: Document -mindirect-branch= option.
21901 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
21904 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
21906 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
21908 * ipa-inline.c (want_inline_small_function_p): Return false if
21909 inlining has already failed with CIF_FINAL_ERROR.
21910 (update_caller_keys): Call want_inline_small_function_p before
21912 (update_callee_keys): Likewise.
21914 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
21916 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
21918 (rs6000_quadword_masked_address_p): Likewise.
21919 (quad_aligned_load_p): Likewise.
21920 (quad_aligned_store_p): Likewise.
21921 (const_load_sequence_p): Add comment to describe the outer-most loop.
21922 (mimic_memory_attributes_and_flags): New function.
21923 (rs6000_gen_stvx): Likewise.
21924 (replace_swapped_aligned_store): Likewise.
21925 (rs6000_gen_lvx): Likewise.
21926 (replace_swapped_aligned_load): Likewise.
21927 (replace_swapped_load_constant): Capitalize argument name in
21928 comment describing this function.
21929 (rs6000_analyze_swaps): Add a third pass to search for vector loads
21930 and stores that access quad-word aligned addresses and replace
21931 with stvx or lvx instructions when appropriate.
21932 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
21933 New function prototype.
21934 (rs6000_quadword_masked_address_p): Likewise.
21935 (rs6000_gen_lvx): Likewise.
21936 (rs6000_gen_stvx): Likewise.
21937 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
21938 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
21939 when memory address is aligned.
21940 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
21941 this split to select lvx instruction when memory address is aligned.
21942 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
21943 instruction when memory address is aligned.
21944 (*vsx_le_perm_load_v16qi): Likewise.
21945 (four unnamed splitters): Modify to select the stvx instruction
21946 when memory is aligned.
21948 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
21950 * predict.c (determine_unlikely_bbs): Handle correctly BBs
21951 which appears in the queue multiple times.
21953 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21954 Alan Hayward <alan.hayward@arm.com>
21955 David Sherwood <david.sherwood@arm.com>
21957 * tree-vectorizer.h (vec_lower_bound): New structure.
21958 (_loop_vec_info): Add check_nonzero and lower_bounds.
21959 (LOOP_VINFO_CHECK_NONZERO): New macro.
21960 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
21961 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
21962 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
21963 fields. Make seg_len the distance travelled, not including the
21965 (dr_direction_indicator): Declare.
21966 (dr_zero_step_indicator): Likewise.
21967 (dr_known_forward_stride_p): Likewise.
21968 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
21970 (runtime_alias_check_p): Allow runtime alias checks with
21972 (operator ==): Compare access_size and align.
21973 (prune_runtime_alias_test_list): Rework for new distinction between
21974 the access_size and seg_len.
21975 (create_intersect_range_checks_index): Likewise. Cope with polynomial
21977 (get_segment_min_max): New function.
21978 (create_intersect_range_checks): Use it.
21979 (dr_step_indicator): New function.
21980 (dr_direction_indicator): Likewise.
21981 (dr_zero_step_indicator): Likewise.
21982 (dr_known_forward_stride_p): Likewise.
21983 * tree-loop-distribution.c (data_ref_segment_size): Return
21984 DR_STEP * (niters - 1).
21985 (compute_alias_check_pairs): Update call to the dr_with_seg_len
21987 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
21988 (vect_preserves_scalar_order_p): New function, split out from...
21989 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
21990 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
21991 (vect_vfa_access_size): New function.
21992 (vect_vfa_align): Likewise.
21993 (vect_compile_time_alias): Take access_size_a and access_b arguments.
21994 (dump_lower_bound): New function.
21995 (vect_check_lower_bound): Likewise.
21996 (vect_small_gap_p): Likewise.
21997 (vectorizable_with_step_bound_p): Likewise.
21998 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
21999 depencies if the vectorization factor is 1. Convert the checks
22000 for nonzero steps into checks on the bounds of DR_STEP. Try using
22001 a bunds check for variable steps if the minimum required step is
22002 relatively small. Update calls to the dr_with_seg_len
22003 constructor and to vect_compile_time_alias.
22004 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
22006 (vect_loop_versioning): Call it.
22007 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
22009 (vect_estimate_min_profitable_iters): Account for any bounds checks.
22011 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22012 Alan Hayward <alan.hayward@arm.com>
22013 David Sherwood <david.sherwood@arm.com>
22015 * doc/sourcebuild.texi (vect_scatter_store): Document.
22016 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
22018 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
22020 * genopinit.c (main): Add supports_vec_scatter_store and
22021 supports_vec_scatter_store_cached to target_optabs.
22022 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
22023 IFN_MASK_SCATTER_STORE.
22024 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
22026 * internal-fn.h (internal_store_fn_p): Declare.
22027 (internal_fn_stored_value_index): Likewise.
22028 * internal-fn.c (scatter_store_direct): New macro.
22029 (expand_scatter_store_optab_fn): New function.
22030 (direct_scatter_store_optab_supported_p): New macro.
22031 (internal_store_fn_p): New function.
22032 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
22033 IFN_MASK_SCATTER_STORE.
22034 (internal_fn_mask_index): Likewise.
22035 (internal_fn_stored_value_index): New function.
22036 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
22037 for scatter stores.
22038 * optabs-query.h (supports_vec_scatter_store_p): Declare.
22039 * optabs-query.c (supports_vec_scatter_store_p): New function.
22040 * tree-vectorizer.h (vect_get_store_rhs): Declare.
22041 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
22042 true for scatter stores.
22043 (vect_gather_scatter_fn_p): Handle scatter stores too.
22044 (vect_check_gather_scatter): Consider using scatter stores if
22045 supports_vec_scatter_store_p.
22046 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
22047 scatter stores too.
22048 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
22049 internal_fn_stored_value_index.
22050 (check_load_store_masking): Handle scatter stores too.
22051 (vect_get_store_rhs): Make public.
22052 (vectorizable_call): Use internal_store_fn_p.
22053 (vectorizable_store): Handle scatter store internal functions.
22054 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
22055 when deciding whether the end of the group has been reached.
22056 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
22057 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
22058 (mask_scatter_store<mode>): New insns.
22060 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22061 Alan Hayward <alan.hayward@arm.com>
22062 David Sherwood <david.sherwood@arm.com>
22064 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
22065 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
22066 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
22068 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
22069 Use vect_truncate_gather_scatter_offset if we can't treat the
22070 operation as a normal gather load or scatter store.
22071 (get_group_load_store_type): Take the gather_scatter_info
22072 as argument. Try using a gather load or scatter store for
22073 single-element groups.
22074 (get_load_store_type): Update calls to get_group_load_store_type
22075 and vect_use_strided_gather_scatters_p.
22077 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22078 Alan Hayward <alan.hayward@arm.com>
22079 David Sherwood <david.sherwood@arm.com>
22081 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
22082 optional tree argument.
22083 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
22085 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
22086 but continue to use the current value as a fallback.
22087 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
22088 to compare the updates.
22089 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
22090 (get_load_store_type): Use it when handling a strided access.
22091 (vect_get_strided_load_store_ops): New function.
22092 (vect_get_data_ptr_increment): Likewise.
22093 (vectorizable_load): Handle strided gather loads. Always pass
22094 a step to vect_create_data_ref_ptr and bump_vector_ptr.
22096 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22097 Alan Hayward <alan.hayward@arm.com>
22098 David Sherwood <david.sherwood@arm.com>
22100 * doc/md.texi (gather_load@var{m}): Document.
22101 (mask_gather_load@var{m}): Likewise.
22102 * genopinit.c (main): Add supports_vec_gather_load and
22103 supports_vec_gather_load_cached to target_optabs.
22104 * optabs-tree.c (init_tree_optimization_optabs): Use
22105 ggc_cleared_alloc to allocate target_optabs.
22106 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
22107 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
22109 * internal-fn.h (internal_load_fn_p): Declare.
22110 (internal_gather_scatter_fn_p): Likewise.
22111 (internal_fn_mask_index): Likewise.
22112 (internal_gather_scatter_fn_supported_p): Likewise.
22113 * internal-fn.c (gather_load_direct): New macro.
22114 (expand_gather_load_optab_fn): New function.
22115 (direct_gather_load_optab_supported_p): New macro.
22116 (direct_internal_fn_optab): New function.
22117 (internal_load_fn_p): Likewise.
22118 (internal_gather_scatter_fn_p): Likewise.
22119 (internal_fn_mask_index): Likewise.
22120 (internal_gather_scatter_fn_supported_p): Likewise.
22121 * optabs-query.c (supports_at_least_one_mode_p): New function.
22122 (supports_vec_gather_load_p): Likewise.
22123 * optabs-query.h (supports_vec_gather_load_p): Declare.
22124 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
22125 and memory_type field.
22126 (NUM_PATTERNS): Bump to 15.
22127 * tree-vect-data-refs.c: Include internal-fn.h.
22128 (vect_gather_scatter_fn_p): New function.
22129 (vect_describe_gather_scatter_call): Likewise.
22130 (vect_check_gather_scatter): Try using internal functions for
22131 gather loads. Recognize existing calls to a gather load function.
22132 (vect_analyze_data_refs): Consider using gather loads if
22133 supports_vec_gather_load_p.
22134 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
22135 (vect_get_gather_scatter_offset_type): Likewise.
22136 (vect_convert_mask_for_vectype): Likewise.
22137 (vect_add_conversion_to_patterm): Likewise.
22138 (vect_try_gather_scatter_pattern): Likewise.
22139 (vect_recog_gather_scatter_pattern): New pattern recognizer.
22140 (vect_vect_recog_func_ptrs): Add it.
22141 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
22142 internal_fn_mask_index and internal_gather_scatter_fn_p.
22143 (check_load_store_masking): Take the gather_scatter_info as an
22144 argument and handle gather loads.
22145 (vect_get_gather_scatter_ops): New function.
22146 (vectorizable_call): Check internal_load_fn_p.
22147 (vectorizable_load): Likewise. Handle gather load internal
22149 (vectorizable_store): Update call to check_load_store_masking.
22150 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
22151 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
22152 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
22153 (aarch64_gather_scale_operand_d): New predicates.
22154 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
22155 (mask_gather_load<mode>): New insns.
22157 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22158 Alan Hayward <alan.hayward@arm.com>
22159 David Sherwood <david.sherwood@arm.com>
22161 * optabs.def (fold_left_plus_optab): New optab.
22162 * doc/md.texi (fold_left_plus_@var{m}): Document.
22163 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
22164 * internal-fn.c (fold_left_direct): Define.
22165 (expand_fold_left_optab_fn): Likewise.
22166 (direct_fold_left_optab_supported_p): Likewise.
22167 * fold-const-call.c (fold_const_fold_left): New function.
22168 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
22169 * tree-parloops.c (valid_reduction_p): New function.
22170 (gather_scalar_reductions): Use it.
22171 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
22172 (vect_finish_replace_stmt): Declare.
22173 * tree-vect-loop.c (fold_left_reduction_fn): New function.
22174 (needs_fold_left_reduction_p): New function, split out from...
22175 (vect_is_simple_reduction): ...here. Accept reductions that
22176 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
22177 (vect_force_simple_reduction): Also store the reduction type in
22178 the assignment's STMT_VINFO_REDUC_TYPE.
22179 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
22180 (merge_with_identity): New function.
22181 (vect_expand_fold_left): Likewise.
22182 (vectorize_fold_left_reduction): Likewise.
22183 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
22184 scalar phi in place for it. Check for target support and reject
22185 cases that would reassociate the operation. Defer the transform
22186 phase to vectorize_fold_left_reduction.
22187 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
22188 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
22189 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
22191 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22193 * tree-if-conv.c (predicate_mem_writes): Remove redundant
22194 call to ifc_temp_var.
22196 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22197 Alan Hayward <alan.hayward@arm.com>
22198 David Sherwood <david.sherwood@arm.com>
22200 * target.def (legitimize_address_displacement): Take the original
22201 offset as a poly_int.
22202 * targhooks.h (default_legitimize_address_displacement): Update
22204 * targhooks.c (default_legitimize_address_displacement): Likewise.
22205 * doc/tm.texi: Regenerate.
22206 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
22207 as an argument, moving assert of ad->disp == ad->disp_term to...
22208 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
22209 Try calling targetm.legitimize_address_displacement before expanding
22210 the address rather than afterwards, and adjust for the new interface.
22211 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
22212 Match the new hook interface. Handle SVE addresses.
22213 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
22214 new hook interface.
22216 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22218 * Makefile.in (OBJS): Add early-remat.o.
22219 * target.def (select_early_remat_modes): New hook.
22220 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
22221 * doc/tm.texi: Regenerate.
22222 * targhooks.h (default_select_early_remat_modes): Declare.
22223 * targhooks.c (default_select_early_remat_modes): New function.
22224 * timevar.def (TV_EARLY_REMAT): New timevar.
22225 * passes.def (pass_early_remat): New pass.
22226 * tree-pass.h (make_pass_early_remat): Declare.
22227 * early-remat.c: New file.
22228 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
22230 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
22232 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22233 Alan Hayward <alan.hayward@arm.com>
22234 David Sherwood <david.sherwood@arm.com>
22236 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
22237 vfm1 with a bound_epilog parameter.
22238 (vect_do_peeling): Update calls accordingly, and move the prologue
22239 call earlier in the function. Treat the base bound_epilog as 0 for
22240 fully-masked loops and retain vf - 1 for other loops. Add 1 to
22241 this base when peeling for gaps.
22242 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
22243 with fully-masked loops.
22244 (vect_estimate_min_profitable_iters): Handle the single peeled
22245 iteration in that case.
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 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
22252 single-element interleaving even if the size is not a power of 2.
22253 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
22254 accesses for single-element interleaving if the group size is
22257 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22258 Alan Hayward <alan.hayward@arm.com>
22259 David Sherwood <david.sherwood@arm.com>
22261 * doc/md.texi (fold_extract_last_@var{m}): Document.
22262 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
22263 * optabs.def (fold_extract_last_optab): New optab.
22264 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
22265 * internal-fn.c (fold_extract_direct): New macro.
22266 (expand_fold_extract_optab_fn): Likewise.
22267 (direct_fold_extract_optab_supported_p): Likewise.
22268 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
22269 * tree-vect-loop.c (vect_model_reduction_cost): Handle
22270 EXTRACT_LAST_REDUCTION.
22271 (get_initial_def_for_reduction): Do not create an initial vector
22272 for EXTRACT_LAST_REDUCTION reductions.
22273 (vectorizable_reduction): Leave the scalar phi in place for
22274 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
22275 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
22276 epilogue code for EXTRACT_LAST_REDUCTION and defer the
22277 transform phase to vectorizable_condition.
22278 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
22280 (vect_finish_stmt_generation): ...here.
22281 (vect_finish_replace_stmt): New function.
22282 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
22283 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
22285 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
22287 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22288 Alan Hayward <alan.hayward@arm.com>
22289 David Sherwood <david.sherwood@arm.com>
22291 * doc/md.texi (extract_last_@var{m}): Document.
22292 * optabs.def (extract_last_optab): New optab.
22293 * internal-fn.def (EXTRACT_LAST): New internal function.
22294 * internal-fn.c (cond_unary_direct): New macro.
22295 (expand_cond_unary_optab_fn): Likewise.
22296 (direct_cond_unary_optab_supported_p): Likewise.
22297 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
22298 loops using EXTRACT_LAST.
22299 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
22300 (extract_last_<mode>): ...this optab.
22301 (vec_extract<mode><Vel>): Update accordingly.
22303 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22304 Alan Hayward <alan.hayward@arm.com>
22305 David Sherwood <david.sherwood@arm.com>
22307 * target.def (empty_mask_is_expensive): New hook.
22308 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
22309 * doc/tm.texi: Regenerate.
22310 * targhooks.h (default_empty_mask_is_expensive): Declare.
22311 * targhooks.c (default_empty_mask_is_expensive): New function.
22312 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
22313 if the target says that empty masks are expensive.
22314 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
22316 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
22318 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22319 Alan Hayward <alan.hayward@arm.com>
22320 David Sherwood <david.sherwood@arm.com>
22322 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
22323 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
22324 (vect_use_loop_mask_for_alignment_p): New function.
22325 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
22326 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
22327 niters_skip argument. Make sure that the first niters_skip elements
22328 of the first iteration are inactive.
22329 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
22330 Update call to vect_set_loop_masks_directly.
22331 (get_misalign_in_elems): New function, split out from...
22332 (vect_gen_prolog_loop_niters): ...here.
22333 (vect_update_init_of_dr): Take a code argument that specifies whether
22334 the adjustment should be added or subtracted.
22335 (vect_update_init_of_drs): Likewise.
22336 (vect_prepare_for_masked_peels): New function.
22337 (vect_do_peeling): Skip prologue peeling if we're using a mask
22338 instead. Update call to vect_update_inits_of_drs.
22339 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
22341 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
22342 alignment. Do not include the number of peeled iterations in
22343 the minimum threshold in that case.
22344 (vectorizable_induction): Adjust the start value down by
22345 LOOP_VINFO_MASK_SKIP_NITERS iterations.
22346 (vect_transform_loop): Call vect_prepare_for_masked_peels.
22347 Take the number of skipped iterations into account when calculating
22349 * tree-vect-stmts.c (vect_gen_while_not): New function.
22351 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22352 Alan Hayward <alan.hayward@arm.com>
22353 David Sherwood <david.sherwood@arm.com>
22355 * doc/sourcebuild.texi (vect_fully_masked): Document.
22356 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
22357 default value to 0.
22358 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
22360 (vect_analyze_loop_2): ...here. Don't check the vectorization
22361 factor against the number of loop iterations if the loop is
22364 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22365 Alan Hayward <alan.hayward@arm.com>
22366 David Sherwood <david.sherwood@arm.com>
22368 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
22369 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
22370 (dump_groups): Update accordingly.
22371 (iv_use::mem_type): New member variable.
22372 (address_p): New function.
22373 (record_use): Add a mem_type argument and initialize the new
22375 (record_group_use): Add a mem_type argument. Use address_p.
22376 Remove obsolete null checks of base_object. Update call to record_use.
22377 (find_interesting_uses_op): Update call to record_group_use.
22378 (find_interesting_uses_cond): Likewise.
22379 (find_interesting_uses_address): Likewise.
22380 (get_mem_type_for_internal_fn): New function.
22381 (find_address_like_use): Likewise.
22382 (find_interesting_uses_stmt): Try find_address_like_use before
22383 calling find_interesting_uses_op.
22384 (addr_offset_valid_p): Use the iv mem_type field as the type
22385 of the addressed memory.
22386 (add_autoinc_candidates): Likewise.
22387 (get_address_cost): Likewise.
22388 (split_small_address_groups_p): Use address_p.
22389 (split_address_groups): Likewise.
22390 (add_iv_candidate_for_use): Likewise.
22391 (autoinc_possible_for_pair): Likewise.
22392 (rewrite_groups): Likewise.
22393 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
22394 (determine_group_iv_cost): Update after split of USE_ADDRESS.
22395 (get_alias_ptr_type_for_ptr_address): New function.
22396 (rewrite_use_address): Rewrite address uses in calls that were
22397 identified by find_address_like_use.
22399 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22400 Alan Hayward <alan.hayward@arm.com>
22401 David Sherwood <david.sherwood@arm.com>
22403 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
22405 * gimple-expr.h (is_gimple_addressable: Likewise.
22406 * gimple-expr.c (is_gimple_address): Likewise.
22407 * internal-fn.c (expand_call_mem_ref): New function.
22408 (expand_mask_load_optab_fn): Use it.
22409 (expand_mask_store_optab_fn): Likewise.
22411 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22412 Alan Hayward <alan.hayward@arm.com>
22413 David Sherwood <david.sherwood@arm.com>
22415 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
22416 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
22417 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
22418 (cond_umax@var{mode}): Document.
22419 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
22420 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
22421 (cond_umin_optab, cond_umax_optab): New optabs.
22422 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
22423 (COND_IOR, COND_XOR): New internal functions.
22424 * internal-fn.h (get_conditional_internal_fn): Declare.
22425 * internal-fn.c (cond_binary_direct): New macro.
22426 (expand_cond_binary_optab_fn): Likewise.
22427 (direct_cond_binary_optab_supported_p): Likewise.
22428 (get_conditional_internal_fn): New function.
22429 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
22430 Cope with reduction statements that are vectorized as calls rather
22432 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
22433 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
22434 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
22435 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
22436 (UNSPEC_COND_EOR): New unspecs.
22437 (optab): Add mappings for them.
22438 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
22439 (sve_int_op, sve_fp_op): New int attributes.
22441 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22442 Alan Hayward <alan.hayward@arm.com>
22443 David Sherwood <david.sherwood@arm.com>
22445 * optabs.def (while_ult_optab): New optab.
22446 * doc/md.texi (while_ult@var{m}@var{n}): Document.
22447 * internal-fn.def (WHILE_ULT): New internal function.
22448 * internal-fn.h (direct_internal_fn_supported_p): New override
22449 that takes two types as argument.
22450 * internal-fn.c (while_direct): New macro.
22451 (expand_while_optab_fn): New function.
22452 (convert_optab_supported_p): Likewise.
22453 (direct_while_optab_supported_p): New macro.
22454 * wide-int.h (wi::udiv_ceil): New function.
22455 * tree-vectorizer.h (rgroup_masks): New structure.
22456 (vec_loop_masks): New typedef.
22457 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
22458 and fully_masked_p.
22459 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
22460 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
22461 (vect_max_vf): New function.
22462 (slpeel_make_loop_iterate_ntimes): Delete.
22463 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
22464 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
22465 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
22466 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
22467 internal-fn.h, stor-layout.h and optabs-query.h.
22468 (vect_set_loop_mask): New function.
22469 (add_preheader_seq): Likewise.
22470 (add_header_seq): Likewise.
22471 (interleave_supported_p): Likewise.
22472 (vect_maybe_permute_loop_masks): Likewise.
22473 (vect_set_loop_masks_directly): Likewise.
22474 (vect_set_loop_condition_masked): Likewise.
22475 (vect_set_loop_condition_unmasked): New function, split out from
22476 slpeel_make_loop_iterate_ntimes.
22477 (slpeel_make_loop_iterate_ntimes): Rename to..
22478 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
22479 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
22480 (vect_do_peeling): Update call accordingly.
22481 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
22483 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
22484 mask_compare_type, can_fully_mask_p and fully_masked_p.
22485 (release_vec_loop_masks): New function.
22486 (_loop_vec_info): Use it to free the loop masks.
22487 (can_produce_all_loop_masks_p): New function.
22488 (vect_get_max_nscalars_per_iter): Likewise.
22489 (vect_verify_full_masking): Likewise.
22490 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
22491 retries, and free the mask rgroups before retrying. Check loop-wide
22492 reasons for disallowing fully-masked loops. Make the final decision
22493 about whether use a fully-masked loop or not.
22494 (vect_estimate_min_profitable_iters): Do not assume that peeling
22495 for the number of iterations will be needed for fully-masked loops.
22496 (vectorizable_reduction): Disable fully-masked loops.
22497 (vectorizable_live_operation): Likewise.
22498 (vect_halve_mask_nunits): New function.
22499 (vect_double_mask_nunits): Likewise.
22500 (vect_record_loop_mask): Likewise.
22501 (vect_get_loop_mask): Likewise.
22502 (vect_transform_loop): Handle the case in which the final loop
22503 iteration might handle a partial vector. Call vect_set_loop_condition
22504 instead of slpeel_make_loop_iterate_ntimes.
22505 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
22506 (check_load_store_masking): New function.
22507 (prepare_load_store_mask): Likewise.
22508 (vectorizable_store): Handle fully-masked loops.
22509 (vectorizable_load): Likewise.
22510 (supportable_widening_operation): Use vect_halve_mask_nunits for
22512 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
22513 (vect_gen_while): New function.
22514 * config/aarch64/aarch64.md (umax<mode>3): New expander.
22515 (aarch64_uqdec<mode>): New insn.
22517 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22518 Alan Hayward <alan.hayward@arm.com>
22519 David Sherwood <david.sherwood@arm.com>
22521 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
22522 (reduc_xor_scal_optab): New optabs.
22523 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
22524 (reduc_xor_scal_@var{m}): Document.
22525 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
22526 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
22527 internal functions.
22528 * fold-const-call.c (fold_const_call): Handle them.
22529 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
22530 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
22531 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
22532 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
22533 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
22534 (UNSPEC_XORV): New unspecs.
22535 (optab): Add entries for them.
22536 (BITWISEV): New int iterator.
22537 (bit_reduc_op): New int attributes.
22539 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22540 Alan Hayward <alan.hayward@arm.com>
22541 David Sherwood <david.sherwood@arm.com>
22543 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
22544 * internal-fn.def (VEC_SHL_INSERT): New internal function.
22545 * optabs.def (vec_shl_insert_optab): New optab.
22546 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
22547 (duplicate_and_interleave): Likewise.
22548 * tree-vect-loop.c: Include internal-fn.h.
22549 (neutral_op_for_slp_reduction): New function, split out from
22550 get_initial_defs_for_reduction.
22551 (get_initial_def_for_reduction): Handle option 2 for variable-length
22552 vectors by loading the neutral value into a vector and then shifting
22553 the initial value into element 0.
22554 (get_initial_defs_for_reduction): Replace the code argument with
22555 the neutral value calculated by neutral_op_for_slp_reduction.
22556 Use gimple_build_vector for constant-length vectors.
22557 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
22558 but the first group_size elements have a neutral value.
22559 Use duplicate_and_interleave otherwise.
22560 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
22561 Update call to get_initial_defs_for_reduction. Handle SLP
22562 reductions for variable-length vectors by creating one vector
22563 result for each scalar result, with the elements associated
22564 with other scalar results stubbed out with the neutral value.
22565 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
22566 Require IFN_VEC_SHL_INSERT for double reductions on
22567 variable-length vectors, or SLP reductions that have
22568 a neutral value. Require can_duplicate_and_interleave_p
22569 support for variable-length unchained SLP reductions if there
22570 is no neutral value, such as for MIN/MAX reductions. Also require
22571 the number of vector elements to be a multiple of the number of
22572 SLP statements when doing variable-length unchained SLP reductions.
22573 Update call to vect_create_epilog_for_reduction.
22574 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
22575 and remove initial values.
22576 (duplicate_and_interleave): Make public.
22577 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
22578 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
22580 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22581 Alan Hayward <alan.hayward@arm.com>
22582 David Sherwood <david.sherwood@arm.com>
22584 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
22585 (can_duplicate_and_interleave_p): New function.
22586 (vect_get_and_check_slp_defs): Take the vector of statements
22587 rather than just the current one. Remove excess parentheses.
22588 Restriction rejectinon of vect_constant_def and vect_external_def
22589 for variable-length vectors to boolean types, or types for which
22590 can_duplicate_and_interleave_p is false.
22591 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
22592 (duplicate_and_interleave): New function.
22593 (vect_get_constant_vectors): Use gimple_build_vector for
22594 constant-length vectors and suitable variable-length constant
22595 vectors. Use duplicate_and_interleave for other variable-length
22596 vectors. Don't defer the update when inserting new statements.
22598 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22599 Alan Hayward <alan.hayward@arm.com>
22600 David Sherwood <david.sherwood@arm.com>
22602 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
22603 min_profitable_iters doesn't go negative.
22605 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22606 Alan Hayward <alan.hayward@arm.com>
22607 David Sherwood <david.sherwood@arm.com>
22609 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
22610 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
22611 * optabs.def (vec_mask_load_lanes_optab): New optab.
22612 (vec_mask_store_lanes_optab): Likewise.
22613 * internal-fn.def (MASK_LOAD_LANES): New internal function.
22614 (MASK_STORE_LANES): Likewise.
22615 * internal-fn.c (mask_load_lanes_direct): New macro.
22616 (mask_store_lanes_direct): Likewise.
22617 (expand_mask_load_optab_fn): Handle masked operations.
22618 (expand_mask_load_lanes_optab_fn): New macro.
22619 (expand_mask_store_optab_fn): Handle masked operations.
22620 (expand_mask_store_lanes_optab_fn): New macro.
22621 (direct_mask_load_lanes_optab_supported_p): Likewise.
22622 (direct_mask_store_lanes_optab_supported_p): Likewise.
22623 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
22625 (vect_load_lanes_supported): Likewise.
22626 * tree-vect-data-refs.c (strip_conversion): New function.
22627 (can_group_stmts_p): Likewise.
22628 (vect_analyze_data_ref_accesses): Use it instead of checking
22629 for a pair of assignments.
22630 (vect_store_lanes_supported): Take a masked_p parameter.
22631 (vect_load_lanes_supported): Likewise.
22632 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
22633 vect_store_lanes_supported and vect_load_lanes_supported.
22634 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
22635 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
22636 parameter. Don't allow gaps for masked accesses.
22637 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
22638 and vect_load_lanes_supported.
22639 (get_load_store_type): Take a masked_p parameter and update
22640 call to get_group_load_store_type.
22641 (vectorizable_store): Update call to get_load_store_type.
22642 Handle IFN_MASK_STORE_LANES.
22643 (vectorizable_load): Update call to get_load_store_type.
22644 Handle IFN_MASK_LOAD_LANES.
22646 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22647 Alan Hayward <alan.hayward@arm.com>
22648 David Sherwood <david.sherwood@arm.com>
22650 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
22652 * config/aarch64/aarch64-protos.h
22653 (aarch64_sve_struct_memory_operand_p): Declare.
22654 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
22655 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
22656 (VPRED, vpred): Handle SVE structure modes.
22657 * config/aarch64/constraints.md (Utx): New constraint.
22658 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
22659 (aarch64_sve_struct_nonimmediate_operand): New predicates.
22660 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
22661 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
22662 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
22663 structure modes. Split into pieces after RA.
22664 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
22665 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
22667 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
22668 SVE structure modes.
22669 (aarch64_classify_address): Likewise.
22670 (sizetochar): Move earlier in file.
22671 (aarch64_print_operand): Handle SVE register lists.
22672 (aarch64_array_mode): New function.
22673 (aarch64_sve_struct_memory_operand_p): Likewise.
22674 (TARGET_ARRAY_MODE): Redefine.
22676 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22677 Alan Hayward <alan.hayward@arm.com>
22678 David Sherwood <david.sherwood@arm.com>
22680 * target.def (array_mode): New target hook.
22681 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
22682 * doc/tm.texi: Regenerate.
22683 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
22684 * hooks.c (hook_optmode_mode_uhwi_none): New function.
22685 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
22686 targetm.array_mode.
22687 * stor-layout.c (mode_for_array): Likewise. Support polynomial
22690 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22691 Alan Hayward <alan.hayward@arm.com>
22692 David Sherwood <david.sherwood@arm.com>
22694 * fold-const.c (fold_binary_loc): Check the argument types
22695 rather than the result type when testing for a vector operation.
22697 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22699 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
22700 * doc/tm.texi: Regenerate.
22702 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22703 Alan Hayward <alan.hayward@arm.com>
22704 David Sherwood <david.sherwood@arm.com>
22706 * doc/invoke.texi (-msve-vector-bits=): Document new option.
22707 (sve): Document new AArch64 extension.
22708 * doc/md.texi (w): Extend the description of the AArch64
22709 constraint to include SVE vectors.
22710 (Upl, Upa): Document new AArch64 predicate constraints.
22711 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
22713 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
22714 (msve-vector-bits=): New option.
22715 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
22716 SVE when these are disabled.
22717 (sve): New extension.
22718 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
22719 modes. Adjust their number of units based on aarch64_sve_vg.
22720 (MAX_BITSIZE_MODE_ANY_MODE): Define.
22721 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
22722 aarch64_addr_query_type.
22723 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
22724 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
22725 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
22726 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
22727 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
22728 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
22729 (aarch64_simd_imm_zero_p): Delete.
22730 (aarch64_check_zero_based_sve_index_immediate): Declare.
22731 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
22732 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
22733 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
22734 (aarch64_sve_float_mul_immediate_p): Likewise.
22735 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
22736 rather than an rtx.
22737 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
22738 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
22739 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
22740 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
22741 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
22742 (aarch64_regmode_natural_size): Likewise.
22743 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
22744 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
22746 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
22747 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
22748 for VG and the SVE predicate registers.
22749 (V_ALIASES): Add a "z"-prefixed alias.
22750 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
22751 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
22752 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
22753 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
22754 (REG_CLASS_NAMES): Add entries for them.
22755 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
22756 and the predicate registers.
22757 (aarch64_sve_vg): Declare.
22758 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
22759 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
22760 (REGMODE_NATURAL_SIZE): Define.
22761 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
22763 * config/aarch64/aarch64.c: Include cfgrtl.h.
22764 (simd_immediate_info): Add a constructor for series vectors,
22765 and an associated step field.
22766 (aarch64_sve_vg): New variable.
22767 (aarch64_dbx_register_number): Handle VG and the predicate registers.
22768 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
22769 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
22770 (VEC_ANY_DATA, VEC_STRUCT): New constants.
22771 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
22772 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
22773 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
22774 (aarch64_get_mask_mode): New functions.
22775 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
22776 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
22777 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
22778 predicate modes and predicate registers. Explicitly restrict
22779 GPRs to modes of 16 bytes or smaller. Only allow FP registers
22780 to store a vector mode if it is recognized by
22781 aarch64_classify_vector_mode.
22782 (aarch64_regmode_natural_size): New function.
22783 (aarch64_hard_regno_caller_save_mode): Return the original mode
22785 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
22786 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
22787 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
22788 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
22790 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
22791 does not overlap dest if the function is frame-related. Handle
22793 (aarch64_split_add_offset): New function.
22794 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
22795 them aarch64_add_offset.
22796 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
22797 and update call to aarch64_sub_sp.
22798 (aarch64_add_cfa_expression): New function.
22799 (aarch64_expand_prologue): Pass extra temporary registers to the
22800 functions above. Handle the case in which we need to emit new
22801 DW_CFA_expressions for registers that were originally saved
22802 relative to the stack pointer, but now have to be expressed
22803 relative to the frame pointer.
22804 (aarch64_output_mi_thunk): Pass extra temporary registers to the
22806 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
22807 IP0 and IP1 values for SVE frames.
22808 (aarch64_expand_vec_series): New function.
22809 (aarch64_expand_sve_widened_duplicate): Likewise.
22810 (aarch64_expand_sve_const_vector): Likewise.
22811 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
22812 Handle SVE constants. Use emit_move_insn to move a force_const_mem
22813 into the register, rather than emitting a SET directly.
22814 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
22815 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
22816 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
22817 (offset_9bit_signed_scaled_p): New functions.
22818 (aarch64_replicate_bitmask_imm): New function.
22819 (aarch64_bitmask_imm): Use it.
22820 (aarch64_cannot_force_const_mem): Reject expressions involving
22821 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
22822 (aarch64_classify_index): Handle SVE indices, by requiring
22823 a plain register index with a scale that matches the element size.
22824 (aarch64_classify_address): Handle SVE addresses. Assert that
22825 the mode of the address is VOIDmode or an integer mode.
22826 Update call to aarch64_classify_symbol.
22827 (aarch64_classify_symbolic_expression): Update call to
22828 aarch64_classify_symbol.
22829 (aarch64_const_vec_all_in_range_p): New function.
22830 (aarch64_print_vector_float_operand): Likewise.
22831 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
22832 "vN" for FP registers with SVE modes. Handle (const ...) vectors
22833 and the FP immediates 1.0 and 0.5.
22834 (aarch64_print_address_internal): Handle SVE addresses.
22835 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
22836 (aarch64_regno_regclass): Handle predicate registers.
22837 (aarch64_secondary_reload): Handle big-endian reloads of SVE
22839 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
22840 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
22841 (aarch64_convert_sve_vector_bits): New function.
22842 (aarch64_override_options): Use it to handle -msve-vector-bits=.
22843 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
22844 rather than an rtx.
22845 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
22846 Handle SVE vector and predicate modes. Accept VL-based constants
22847 that need only one temporary register, and VL offsets that require
22848 no temporary registers.
22849 (aarch64_conditional_register_usage): Mark the predicate registers
22850 as fixed if SVE isn't available.
22851 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
22852 Return true for SVE vector and predicate modes.
22853 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
22854 rather than an unsigned int. Handle SVE modes.
22855 (aarch64_preferred_simd_mode): Update call accordingly. Handle
22857 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
22859 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
22860 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
22861 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
22862 (aarch64_sve_float_mul_immediate_p): New functions.
22863 (aarch64_sve_valid_immediate): New function.
22864 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
22865 Explicitly reject structure modes. Check for INDEX constants.
22866 Handle PTRUE and PFALSE constants.
22867 (aarch64_check_zero_based_sve_index_immediate): New function.
22868 (aarch64_simd_imm_zero_p): Delete.
22869 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
22870 vector modes. Accept constants in the range of CNT[BHWD].
22871 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
22872 ask for an Advanced SIMD mode.
22873 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
22874 (aarch64_simd_vector_alignment): Handle SVE predicates.
22875 (aarch64_vectorize_preferred_vector_alignment): New function.
22876 (aarch64_simd_vector_alignment_reachable): Use it instead of
22878 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
22879 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
22881 (MAX_VECT_LEN): Delete.
22882 (expand_vec_perm_d): Add a vec_flags field.
22883 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
22884 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
22885 (aarch64_evpc_ext): Don't apply a big-endian lane correction
22887 (aarch64_evpc_rev): Rename to...
22888 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
22889 (aarch64_evpc_rev_global): New function.
22890 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
22891 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
22893 (aarch64_evpc_sve_tbl): New function.
22894 (aarch64_expand_vec_perm_const_1): Update after rename of
22895 aarch64_evpc_rev. Handle SVE permutes too, trying
22896 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
22897 than aarch64_evpc_tbl.
22898 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
22899 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
22900 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
22901 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
22902 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
22903 (aarch64_expand_sve_vcond): New functions.
22904 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
22905 of aarch64_vector_mode_p.
22906 (aarch64_dwarf_poly_indeterminate_value): New function.
22907 (aarch64_compute_pressure_classes): Likewise.
22908 (aarch64_can_change_mode_class): Likewise.
22909 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
22910 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
22911 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
22912 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
22913 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
22914 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
22915 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
22916 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
22918 (Dn, Dl, Dr): Accept const as well as const_vector.
22919 (Dz): Likewise. Compare against CONST0_RTX.
22920 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
22921 of "vector" where appropriate.
22922 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
22923 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
22924 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
22925 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
22926 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
22927 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
22928 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
22929 (v_int_equiv): Extend to SVE modes.
22930 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
22932 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
22933 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
22934 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
22935 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
22936 (SVE_COND_FP_CMP): New int iterators.
22937 (perm_hilo): Handle the new unpack unspecs.
22938 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
22940 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
22941 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
22942 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
22943 (aarch64_equality_operator, aarch64_constant_vector_operand)
22944 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
22945 (aarch64_sve_nonimmediate_operand): Likewise.
22946 (aarch64_sve_general_operand): Likewise.
22947 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
22948 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
22949 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
22950 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
22951 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
22952 (aarch64_sve_float_arith_immediate): Likewise.
22953 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
22954 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
22955 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
22956 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
22957 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
22958 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
22959 (aarch64_sve_float_arith_operand): Likewise.
22960 (aarch64_sve_float_arith_with_sub_operand): Likewise.
22961 (aarch64_sve_float_mul_operand): Likewise.
22962 (aarch64_sve_vec_perm_operand): Likewise.
22963 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
22964 (aarch64_mov_operand): Accept const_poly_int and const_vector.
22965 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
22966 as well as const_vector.
22967 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
22968 in file. Use CONST0_RTX and CONSTM1_RTX.
22969 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
22970 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
22971 Use aarch64_simd_imm_zero.
22972 * config/aarch64/aarch64-sve.md: New file.
22973 * config/aarch64/aarch64.md: Include it.
22974 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
22975 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
22976 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
22977 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
22978 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
22979 (sve): New attribute.
22980 (enabled): Disable instructions with the sve attribute unless
22982 (movqi, movhi): Pass CONST_POLY_INT operaneds through
22983 aarch64_expand_mov_immediate.
22984 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
22985 CNT[BHSD] immediates.
22986 (movti): Split CONST_POLY_INT moves into two halves.
22987 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
22988 Split additions that need a temporary here if the destination
22989 is the stack pointer.
22990 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
22991 (*add<mode>3_poly_1): New instruction.
22992 (set_clobber_cc): New expander.
22994 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22996 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
22997 parameter and use it instead of GET_MODE_SIZE (innermode). Use
22998 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
22999 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
23000 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
23001 Change innermode from fixed_mode_size to machine_mode.
23002 (simplify_subreg): Update call accordingly. Handle a constant-sized
23003 subreg of a variable-length CONST_VECTOR.
23005 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23006 Alan Hayward <alan.hayward@arm.com>
23007 David Sherwood <david.sherwood@arm.com>
23009 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
23010 (add_offset_to_base): New function, split out from...
23011 (create_mem_ref): ...here. When handling a scale other than 1,
23012 check first whether the address is valid without the offset.
23013 Add it into the base if so, leaving the index and scale as-is.
23015 2018-01-12 Jakub Jelinek <jakub@redhat.com>
23018 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
23019 fold_for_warn before checking if arg2 is INTEGER_CST.
23021 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
23023 * config/rs6000/predicates.md (load_multiple_operation): Delete.
23024 (store_multiple_operation): Delete.
23025 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
23026 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
23027 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
23028 guarded by TARGET_STRING.
23029 (rs6000_output_load_multiple): Delete.
23030 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
23031 OPTION_MASK_STRING / TARGET_STRING handling.
23032 (print_operand) <'N', 'O'>: Add comment that these are unused now.
23033 (const rs6000_opt_masks) <"string">: Change mask to 0.
23034 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
23035 (MASK_STRING): Delete.
23036 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
23038 (load_multiple): Delete.
23045 (store_multiple): Delete.
23052 (movmemsi_8reg): Delete.
23053 (corresponding unnamed define_insn): Delete.
23054 (movmemsi_6reg): Delete.
23055 (corresponding unnamed define_insn): Delete.
23056 (movmemsi_4reg): Delete.
23057 (corresponding unnamed define_insn): Delete.
23058 (movmemsi_2reg): Delete.
23059 (corresponding unnamed define_insn): Delete.
23060 (movmemsi_1reg): Delete.
23061 (corresponding unnamed define_insn): Delete.
23062 * config/rs6000/rs6000.opt (mno-string): New.
23063 (mstring): Replace by deprecation warning stub.
23064 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
23066 2018-01-12 Jakub Jelinek <jakub@redhat.com>
23068 * regrename.c (regrename_do_replace): If replacing the same
23069 reg multiple times, try to reuse last created gen_raw_REG.
23072 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
23073 main to workaround a bug in GDB.
23075 2018-01-12 Tom de Vries <tom@codesourcery.com>
23078 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
23080 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
23082 PR rtl-optimization/80481
23083 * ira-color.c (get_cap_member): New function.
23084 (allocnos_conflict_by_live_ranges_p): Use it.
23085 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
23086 (setup_slot_coalesced_allocno_live_ranges): Ditto.
23088 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
23091 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
23092 (*saddl_se_1): Ditto.
23093 (*ssubsi_1): Ditto.
23094 (*ssubl_se_1): Ditto.
23096 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
23098 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
23099 rather than wi::to_widest for DR_INITs.
23100 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
23101 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
23102 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
23104 (vect_analyze_group_access_1): Note that here.
23106 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
23108 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
23109 polynomial type sizes.
23111 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
23113 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
23114 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
23115 (gimple_add_tmp_var): Likewise.
23117 2018-01-12 Martin Liska <mliska@suse.cz>
23119 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
23120 (gimple_alloc_sizes): Likewise.
23121 (dump_gimple_statistics): Use PRIu64 in printf format.
23122 * gimple.h: Change uint64_t to int.
23124 2018-01-12 Martin Liska <mliska@suse.cz>
23126 * tree-core.h: Use uint64_t instead of int.
23127 * tree.c (tree_node_counts): Likewise.
23128 (tree_node_sizes): Likewise.
23129 (dump_tree_statistics): Use PRIu64 in printf format.
23131 2018-01-12 Martin Liska <mliska@suse.cz>
23133 * Makefile.in: As qsort_chk is implemented in vec.c, add
23134 vec.o to linkage of gencfn-macros.
23135 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
23136 passing the info to record_node_allocation_statistics.
23137 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
23139 * ggc-common.c (struct ggc_usage): Add operator== and use
23140 it in operator< and compare function.
23141 * mem-stats.h (struct mem_usage): Likewise.
23142 * vec.c (struct vec_usage): Remove operator< and compare
23143 function. Can be simply inherited.
23145 2018-01-12 Martin Jambor <mjambor@suse.cz>
23148 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
23149 * tree-ssa-math-opts.c: Include domwalk.h.
23150 (convert_mult_to_fma_1): New function.
23151 (fma_transformation_info): New type.
23152 (fma_deferring_state): Likewise.
23153 (cancel_fma_deferring): New function.
23154 (result_of_phi): Likewise.
23155 (last_fma_candidate_feeds_initial_phi): Likewise.
23156 (convert_mult_to_fma): Added deferring logic, split actual
23157 transformation to convert_mult_to_fma_1.
23158 (math_opts_dom_walker): New type.
23159 (math_opts_dom_walker::after_dom_children): New method, body moved
23160 here from pass_optimize_widening_mul::execute, added deferring logic
23162 (pass_optimize_widening_mul::execute): Moved most of code to
23163 math_opts_dom_walker::after_dom_children.
23164 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
23165 * config/i386/i386.c (ix86_option_override_internal): Added
23166 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
23168 2018-01-12 Richard Biener <rguenther@suse.de>
23171 * dwarf2out.c (gen_variable_die): Do not reset old_die for
23172 inline instance vars.
23174 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
23177 * config/rx/rx.c (rx_is_restricted_memory_address):
23178 Handle SUBREG case.
23180 2018-01-12 Richard Biener <rguenther@suse.de>
23182 PR tree-optimization/80846
23183 * target.def (split_reduction): New target hook.
23184 * targhooks.c (default_split_reduction): New function.
23185 * targhooks.h (default_split_reduction): Declare.
23186 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
23187 target requests first reduce vectors by combining low and high
23189 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
23190 (get_vectype_for_scalar_type_and_size): Export.
23191 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
23192 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
23193 * doc/tm.texi: Regenerate.
23194 * config/i386/i386.c (ix86_split_reduction): Implement
23195 TARGET_VECTORIZE_SPLIT_REDUCTION.
23197 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
23200 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
23201 in PIC mode except for TARGET_VXWORKS_RTP.
23202 * config/sparc/sparc.c: Include cfgrtl.h.
23203 (TARGET_INIT_PIC_REG): Define.
23204 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
23205 (sparc_pic_register_p): New predicate.
23206 (sparc_legitimate_address_p): Use it.
23207 (sparc_legitimize_pic_address): Likewise.
23208 (sparc_delegitimize_address): Likewise.
23209 (sparc_mode_dependent_address_p): Likewise.
23210 (gen_load_pcrel_sym): Remove 4th parameter.
23211 (load_got_register): Adjust call to above. Remove obsolete stuff.
23212 (sparc_expand_prologue): Do not call load_got_register here.
23213 (sparc_flat_expand_prologue): Likewise.
23214 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
23215 (sparc_use_pseudo_pic_reg): New function.
23216 (sparc_init_pic_reg): Likewise.
23217 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
23218 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
23220 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
23222 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
23223 Add item for branch_cost.
23225 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
23227 PR rtl-optimization/83565
23228 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
23229 not extend the result to a larger mode for rotate operations.
23230 (num_sign_bit_copies1): Likewise.
23232 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
23235 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
23237 Use values-Xc.o for -pedantic.
23238 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
23240 2018-01-12 Martin Liska <mliska@suse.cz>
23243 * ipa-devirt.c (final_warning_record::grow_type_warnings):
23245 (possible_polymorphic_call_targets): Use it.
23246 (ipa_devirt): Likewise.
23248 2018-01-12 Martin Liska <mliska@suse.cz>
23250 * profile-count.h (enum profile_quality): Use 0 as invalid
23251 enum value of profile_quality.
23253 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
23255 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
23256 -mext-string options.
23258 2018-01-12 Richard Biener <rguenther@suse.de>
23260 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
23261 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
23262 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
23264 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
23266 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
23268 * configure.ac (--with-long-double-format): Add support for the
23269 configuration option to change the default long double format on
23271 * config.gcc (powerpc*-linux*-*): Likewise.
23272 * configure: Regenerate.
23273 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
23274 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
23275 used without modification.
23277 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23279 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
23280 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
23281 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
23282 MISC_BUILTIN_SPEC_BARRIER.
23283 (rs6000_init_builtins): Likewise.
23284 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
23286 (speculation_barrier): New define_insn.
23287 * doc/extend.texi: Document __builtin_speculation_barrier.
23289 2018-01-11 Jakub Jelinek <jakub@redhat.com>
23292 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
23293 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
23294 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
23296 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
23297 integral modes instead of "ss" and "sd".
23298 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
23299 vectors with 32-bit and 64-bit elements.
23300 (vecdupssescalarmodesuffix): New mode attribute.
23301 (vec_dup<mode>): Use it.
23303 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
23306 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
23307 frame if argument is passed on stack.
23309 2018-01-11 Jakub Jelinek <jakub@redhat.com>
23312 * ree.c (combine_reaching_defs): Optimize also
23313 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
23314 reg2=any_extend(exp); reg1=reg2;, formatting fix.
23316 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
23318 PR middle-end/83189
23319 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
23321 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
23323 PR middle-end/83718
23324 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
23325 after they are computed.
23327 2018-01-11 Bin Cheng <bin.cheng@arm.com>
23329 PR tree-optimization/83695
23330 * gimple-loop-linterchange.cc
23331 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
23332 reset cached scev information after interchange.
23333 (pass_linterchange::execute): Remove call to scev_reset_htab.
23335 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23337 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
23338 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
23339 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
23340 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
23341 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
23342 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
23343 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
23344 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
23345 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
23346 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
23347 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
23348 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
23349 (V_lane_reg): Likewise.
23350 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
23352 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
23353 (vfmal_lane_low<mode>_intrinsic,
23354 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
23355 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
23356 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
23357 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
23358 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
23359 vfmsl_lane_high<mode>_intrinsic): New define_insns.
23361 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23363 * config/arm/arm-cpus.in (fp16fml): New feature.
23364 (ALL_SIMD): Add fp16fml.
23365 (armv8.2-a): Add fp16fml as an option.
23366 (armv8.3-a): Likewise.
23367 (armv8.4-a): Add fp16fml as part of fp16.
23368 * config/arm/arm.h (TARGET_FP16FML): Define.
23369 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
23371 * config/arm/arm-modes.def (V2HF): Define.
23372 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
23373 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
23374 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
23375 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
23376 vfmsl_low, vfmsl_high): New set of builtins.
23377 * config/arm/iterators.md (PLUSMINUS): New code iterator.
23378 (vfml_op): New code attribute.
23379 (VFMLHALVES): New int iterator.
23380 (VFML, VFMLSEL): New mode attributes.
23381 (V_reg): Define mapping for V2HF.
23382 (V_hi, V_lo): New mode attributes.
23383 (VF_constraint): Likewise.
23384 (vfml_half, vfml_half_selector): New int attributes.
23385 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
23387 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
23388 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
23390 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
23391 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
23392 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
23393 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
23395 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
23396 Document new effective target and option set.
23398 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23400 * config/arm/arm-cpus.in (armv8_4): New feature.
23401 (ARMv8_4a): New fgroup.
23402 (armv8.4-a): New arch.
23403 * config/arm/arm-tables.opt: Regenerate.
23404 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
23405 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
23406 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
23407 Add matching rules for -march=armv8.4-a and extensions.
23408 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
23410 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
23413 * config/rx/rx.md (BW): New mode attribute.
23414 (sync_lock_test_and_setsi): Add mode suffix to insn output.
23416 2018-01-11 Richard Biener <rguenther@suse.de>
23418 PR tree-optimization/83435
23419 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
23420 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
23421 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
23423 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23424 Alan Hayward <alan.hayward@arm.com>
23425 David Sherwood <david.sherwood@arm.com>
23427 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
23429 (aarch64_classify_address): Initialize it. Track polynomial offsets.
23430 (aarch64_print_address_internal): Use it to check for a zero offset.
23432 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23433 Alan Hayward <alan.hayward@arm.com>
23434 David Sherwood <david.sherwood@arm.com>
23436 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
23437 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
23438 Return a poly_int64 rather than a HOST_WIDE_INT.
23439 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
23440 rather than a HOST_WIDE_INT.
23441 * config/aarch64/aarch64.h (aarch64_frame): Protect with
23442 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
23443 hard_fp_offset, frame_size, initial_adjust, callee_offset and
23444 final_offset from HOST_WIDE_INT to poly_int64.
23445 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
23446 to_constant when getting the number of units in an Advanced SIMD
23448 (aarch64_builtin_vectorized_function): Check for a constant number
23450 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
23452 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
23453 attribute instead of GET_MODE_NUNITS.
23454 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
23455 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
23456 GET_MODE_SIZE for fixed-size registers.
23457 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
23458 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
23459 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
23460 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
23461 (aarch64_print_operand, aarch64_print_address_internal)
23462 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
23463 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
23464 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
23465 Handle polynomial GET_MODE_SIZE.
23466 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
23467 wider than SImode without modification.
23468 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
23469 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
23470 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
23471 passing and returning SVE modes.
23472 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
23473 rather than GEN_INT.
23474 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
23475 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
23476 (aarch64_allocate_and_probe_stack_space): Likewise.
23477 (aarch64_layout_frame): Cope with polynomial offsets.
23478 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
23479 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
23480 polynomial offsets.
23481 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
23482 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
23483 poly_int64 rather than a HOST_WIDE_INT.
23484 (aarch64_get_separate_components, aarch64_process_components)
23485 (aarch64_expand_prologue, aarch64_expand_epilogue)
23486 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
23487 (aarch64_anchor_offset): New function, split out from...
23488 (aarch64_legitimize_address): ...here.
23489 (aarch64_builtin_vectorization_cost): Handle polynomial
23490 TYPE_VECTOR_SUBPARTS.
23491 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
23493 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
23494 number of elements from the PARALLEL rather than the mode.
23495 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
23496 rather than GET_MODE_BITSIZE.
23497 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
23498 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
23499 (aarch64_expand_vec_perm_const_1): Handle polynomial
23500 d->perm.length () and d->perm elements.
23501 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
23502 Apply to_constant to d->perm elements.
23503 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
23504 polynomial CONST_VECTOR_NUNITS.
23505 (aarch64_move_pointer): Take amount as a poly_int64 rather
23507 (aarch64_progress_pointer): Avoid temporary variable.
23508 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
23509 the mode attribute instead of GET_MODE.
23511 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23512 Alan Hayward <alan.hayward@arm.com>
23513 David Sherwood <david.sherwood@arm.com>
23515 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
23516 x exists before using it.
23517 (aarch64_add_constant_internal): Rename to...
23518 (aarch64_add_offset_1): ...this. Replace regnum with separate
23519 src and dest rtxes. Handle the case in which they're different,
23520 including when the offset is zero. Replace scratchreg with an rtx.
23521 Use 2 additions if there is no spare register into which we can
23522 move a 16-bit constant.
23523 (aarch64_add_constant): Delete.
23524 (aarch64_add_offset): Replace reg with separate src and dest
23525 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
23526 Use aarch64_add_offset_1.
23527 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
23528 an rtx rather than an int. Take the delta as a poly_int64
23529 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
23530 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
23531 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
23532 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
23533 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
23534 and aarch64_add_sp.
23535 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
23536 aarch64_add_constant.
23538 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23540 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
23541 Use scalar_float_mode.
23543 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23545 * config/aarch64/aarch64-simd.md
23546 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
23547 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
23548 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
23549 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
23550 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
23551 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
23552 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
23553 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
23554 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
23555 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
23557 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23560 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
23561 targ_options->x_arm_arch_string is non NULL.
23563 2018-01-11 Tamar Christina <tamar.christina@arm.com>
23565 * config/aarch64/aarch64.h
23566 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
23568 2018-01-11 Sudakshina Das <sudi.das@arm.com>
23571 * expmed.c (emit_store_flag_force): Swap if const op0
23572 and change VOIDmode to mode of op0.
23574 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23576 PR rtl-optimization/83761
23577 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
23578 than bytes to mode_for_size.
23580 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
23582 PR middle-end/83189
23583 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
23584 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
23587 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
23589 PR middle-end/83575
23590 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
23591 when in layout mode.
23592 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
23593 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
23596 2018-01-10 Michael Collison <michael.collison@arm.com>
23598 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
23599 * config/aarch64/aarch64-option-extension.def: Add
23600 AARCH64_OPT_EXTENSION of 'fp16fml'.
23601 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23602 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
23603 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
23604 * config/aarch64/constraints.md (Ui7): New constraint.
23605 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
23606 (VFMLA_SEL_W): Ditto.
23609 (VFMLA16_LOW): New int iterator.
23610 (VFMLA16_HIGH): Ditto.
23611 (UNSPEC_FMLAL): New unspec.
23612 (UNSPEC_FMLSL): Ditto.
23613 (UNSPEC_FMLAL2): Ditto.
23614 (UNSPEC_FMLSL2): Ditto.
23615 (f16mac): New code attribute.
23616 * config/aarch64/aarch64-simd-builtins.def
23617 (aarch64_fmlal_lowv2sf): Ditto.
23618 (aarch64_fmlsl_lowv2sf): Ditto.
23619 (aarch64_fmlalq_lowv4sf): Ditto.
23620 (aarch64_fmlslq_lowv4sf): Ditto.
23621 (aarch64_fmlal_highv2sf): Ditto.
23622 (aarch64_fmlsl_highv2sf): Ditto.
23623 (aarch64_fmlalq_highv4sf): Ditto.
23624 (aarch64_fmlslq_highv4sf): Ditto.
23625 (aarch64_fmlal_lane_lowv2sf): Ditto.
23626 (aarch64_fmlsl_lane_lowv2sf): Ditto.
23627 (aarch64_fmlal_laneq_lowv2sf): Ditto.
23628 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
23629 (aarch64_fmlalq_lane_lowv4sf): Ditto.
23630 (aarch64_fmlsl_lane_lowv4sf): Ditto.
23631 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
23632 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
23633 (aarch64_fmlal_lane_highv2sf): Ditto.
23634 (aarch64_fmlsl_lane_highv2sf): Ditto.
23635 (aarch64_fmlal_laneq_highv2sf): Ditto.
23636 (aarch64_fmlsl_laneq_highv2sf): Ditto.
23637 (aarch64_fmlalq_lane_highv4sf): Ditto.
23638 (aarch64_fmlsl_lane_highv4sf): Ditto.
23639 (aarch64_fmlalq_laneq_highv4sf): Ditto.
23640 (aarch64_fmlsl_laneq_highv4sf): Ditto.
23641 * config/aarch64/aarch64-simd.md:
23642 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
23643 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
23644 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
23645 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
23646 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
23647 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
23648 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
23649 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
23650 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
23651 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
23652 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
23653 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
23654 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
23655 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
23656 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
23657 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
23658 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
23659 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
23660 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
23661 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
23662 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
23663 (vfmlsl_low_u32): Ditto.
23664 (vfmlalq_low_u32): Ditto.
23665 (vfmlslq_low_u32): Ditto.
23666 (vfmlal_high_u32): Ditto.
23667 (vfmlsl_high_u32): Ditto.
23668 (vfmlalq_high_u32): Ditto.
23669 (vfmlslq_high_u32): Ditto.
23670 (vfmlal_lane_low_u32): Ditto.
23671 (vfmlsl_lane_low_u32): Ditto.
23672 (vfmlal_laneq_low_u32): Ditto.
23673 (vfmlsl_laneq_low_u32): Ditto.
23674 (vfmlalq_lane_low_u32): Ditto.
23675 (vfmlslq_lane_low_u32): Ditto.
23676 (vfmlalq_laneq_low_u32): Ditto.
23677 (vfmlslq_laneq_low_u32): Ditto.
23678 (vfmlal_lane_high_u32): Ditto.
23679 (vfmlsl_lane_high_u32): Ditto.
23680 (vfmlal_laneq_high_u32): Ditto.
23681 (vfmlsl_laneq_high_u32): Ditto.
23682 (vfmlalq_lane_high_u32): Ditto.
23683 (vfmlslq_lane_high_u32): Ditto.
23684 (vfmlalq_laneq_high_u32): Ditto.
23685 (vfmlslq_laneq_high_u32): Ditto.
23686 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
23687 (AARCH64_FL_FOR_ARCH8_4): New.
23688 (AARCH64_ISA_F16FML): New ISA flag.
23689 (TARGET_F16FML): New feature flag for fp16fml.
23690 (doc/invoke.texi): Document new fp16fml option.
23692 2018-01-10 Michael Collison <michael.collison@arm.com>
23694 * config/aarch64/aarch64-builtins.c:
23695 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
23696 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23697 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
23698 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
23699 (AARCH64_ISA_SHA3): New ISA flag.
23700 (TARGET_SHA3): New feature flag for sha3.
23701 * config/aarch64/iterators.md (sha512_op): New int attribute.
23702 (CRYPTO_SHA512): New int iterator.
23703 (UNSPEC_SHA512H): New unspec.
23704 (UNSPEC_SHA512H2): Ditto.
23705 (UNSPEC_SHA512SU0): Ditto.
23706 (UNSPEC_SHA512SU1): Ditto.
23707 * config/aarch64/aarch64-simd-builtins.def
23708 (aarch64_crypto_sha512hqv2di): New builtin.
23709 (aarch64_crypto_sha512h2qv2di): Ditto.
23710 (aarch64_crypto_sha512su0qv2di): Ditto.
23711 (aarch64_crypto_sha512su1qv2di): Ditto.
23712 (aarch64_eor3qv8hi): Ditto.
23713 (aarch64_rax1qv2di): Ditto.
23714 (aarch64_xarqv2di): Ditto.
23715 (aarch64_bcaxqv8hi): Ditto.
23716 * config/aarch64/aarch64-simd.md:
23717 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
23718 (aarch64_crypto_sha512su0qv2di): Ditto.
23719 (aarch64_crypto_sha512su1qv2di): Ditto.
23720 (aarch64_eor3qv8hi): Ditto.
23721 (aarch64_rax1qv2di): Ditto.
23722 (aarch64_xarqv2di): Ditto.
23723 (aarch64_bcaxqv8hi): Ditto.
23724 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
23725 (vsha512h2q_u64): Ditto.
23726 (vsha512su0q_u64): Ditto.
23727 (vsha512su1q_u64): Ditto.
23728 (veor3q_u16): Ditto.
23729 (vrax1q_u64): Ditto.
23730 (vxarq_u64): Ditto.
23731 (vbcaxq_u16): Ditto.
23732 * config/arm/types.md (crypto_sha512): New type attribute.
23733 (crypto_sha3): Ditto.
23734 (doc/invoke.texi): Document new sha3 option.
23736 2018-01-10 Michael Collison <michael.collison@arm.com>
23738 * config/aarch64/aarch64-builtins.c:
23739 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
23740 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23741 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
23742 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
23743 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
23744 (AARCH64_ISA_SM4): New ISA flag.
23745 (TARGET_SM4): New feature flag for sm4.
23746 * config/aarch64/aarch64-simd-builtins.def
23747 (aarch64_sm3ss1qv4si): Ditto.
23748 (aarch64_sm3tt1aq4si): Ditto.
23749 (aarch64_sm3tt1bq4si): Ditto.
23750 (aarch64_sm3tt2aq4si): Ditto.
23751 (aarch64_sm3tt2bq4si): Ditto.
23752 (aarch64_sm3partw1qv4si): Ditto.
23753 (aarch64_sm3partw2qv4si): Ditto.
23754 (aarch64_sm4eqv4si): Ditto.
23755 (aarch64_sm4ekeyqv4si): Ditto.
23756 * config/aarch64/aarch64-simd.md:
23757 (aarch64_sm3ss1qv4si): Ditto.
23758 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
23759 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
23760 (aarch64_sm4eqv4si): Ditto.
23761 (aarch64_sm4ekeyqv4si): Ditto.
23762 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
23763 (sm3part_op): Ditto.
23764 (CRYPTO_SM3TT): Ditto.
23765 (CRYPTO_SM3PART): Ditto.
23766 (UNSPEC_SM3SS1): New unspec.
23767 (UNSPEC_SM3TT1A): Ditto.
23768 (UNSPEC_SM3TT1B): Ditto.
23769 (UNSPEC_SM3TT2A): Ditto.
23770 (UNSPEC_SM3TT2B): Ditto.
23771 (UNSPEC_SM3PARTW1): Ditto.
23772 (UNSPEC_SM3PARTW2): Ditto.
23773 (UNSPEC_SM4E): Ditto.
23774 (UNSPEC_SM4EKEY): Ditto.
23775 * config/aarch64/constraints.md (Ui2): New constraint.
23776 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
23777 * config/arm/types.md (crypto_sm3): New type attribute.
23778 (crypto_sm4): Ditto.
23779 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
23780 (vsm3tt1aq_u32): Ditto.
23781 (vsm3tt1bq_u32): Ditto.
23782 (vsm3tt2aq_u32): Ditto.
23783 (vsm3tt2bq_u32): Ditto.
23784 (vsm3partw1q_u32): Ditto.
23785 (vsm3partw2q_u32): Ditto.
23786 (vsm4eq_u32): Ditto.
23787 (vsm4ekeyq_u32): Ditto.
23788 (doc/invoke.texi): Document new sm4 option.
23790 2018-01-10 Michael Collison <michael.collison@arm.com>
23792 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
23793 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
23794 (AARCH64_FL_FOR_ARCH8_4): New.
23795 (AARCH64_FL_V8_4): New flag.
23796 (doc/invoke.texi): Document new armv8.4-a option.
23798 2018-01-10 Michael Collison <michael.collison@arm.com>
23800 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23801 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
23802 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
23803 * config/aarch64/aarch64-option-extension.def: Add
23804 AARCH64_OPT_EXTENSION of 'sha2'.
23805 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
23806 (crypto): Disable sha2 and aes if crypto disabled.
23807 (crypto): Enable aes and sha2 if enabled.
23808 (simd): Disable sha2 and aes if simd disabled.
23809 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
23811 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
23812 (TARGET_SHA2): New feature flag for sha2.
23813 (TARGET_AES): New feature flag for aes.
23814 * config/aarch64/aarch64-simd.md:
23815 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
23816 conditional on TARGET_AES.
23817 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
23818 (aarch64_crypto_sha1hsi): Make pattern conditional
23820 (aarch64_crypto_sha1hv4si): Ditto.
23821 (aarch64_be_crypto_sha1hv4si): Ditto.
23822 (aarch64_crypto_sha1su1v4si): Ditto.
23823 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
23824 (aarch64_crypto_sha1su0v4si): Ditto.
23825 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
23826 (aarch64_crypto_sha256su0v4si): Ditto.
23827 (aarch64_crypto_sha256su1v4si): Ditto.
23828 (doc/invoke.texi): Document new aes and sha2 options.
23830 2018-01-10 Martin Sebor <msebor@redhat.com>
23832 PR tree-optimization/83781
23833 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
23836 2018-01-11 Martin Sebor <msebor@gmail.com>
23837 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23839 PR tree-optimization/83501
23840 PR tree-optimization/81703
23842 * tree-ssa-strlen.c (get_string_cst): Rename...
23843 (get_string_len): ...to this. Handle global constants.
23844 (handle_char_store): Adjust.
23846 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
23847 Jim Wilson <jimw@sifive.com>
23849 * config/riscv/riscv-protos.h (riscv_output_return): New.
23850 * config/riscv/riscv.c (struct machine_function): New naked_p field.
23851 (riscv_attribute_table, riscv_output_return),
23852 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
23853 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
23854 (riscv_compute_frame_info): Only compute frame->mask if not a naked
23856 (riscv_expand_prologue): Add early return for naked function.
23857 (riscv_expand_epilogue): Likewise.
23858 (riscv_function_ok_for_sibcall): Return false for naked function.
23859 (riscv_set_current_function): New.
23860 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
23861 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
23862 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
23863 * doc/extend.texi (RISC-V Function Attributes): New.
23865 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
23867 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
23868 check for 128-bit long double before checking TCmode.
23869 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
23870 128-bit long doubles before checking TFmode or TCmode.
23871 (FLOAT128_IBM_P): Likewise.
23873 2018-01-10 Martin Sebor <msebor@redhat.com>
23875 PR tree-optimization/83671
23876 * builtins.c (c_strlen): Unconditionally return zero for the empty
23878 Use -Warray-bounds for warnings.
23879 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
23880 for non-constant array indices with COMPONENT_REF, arrays of
23881 arrays, and pointers to arrays.
23882 (gimple_fold_builtin_strlen): Determine and set length range for
23883 non-constant character arrays.
23885 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
23887 PR middle-end/81897
23888 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
23891 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
23893 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
23895 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
23898 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
23899 VECTOR_MEM_ALTIVEC_OR_VSX_P.
23900 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
23901 indexed_or_indirect_operand predicate.
23902 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
23903 (*vsx_le_perm_load_v8hi): Likewise.
23904 (*vsx_le_perm_load_v16qi): Likewise.
23905 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
23906 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
23907 (*vsx_le_perm_store_v8hi): Likewise.
23908 (*vsx_le_perm_store_v16qi): Likewise.
23909 (eight unnamed splitters): Likewise.
23911 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
23913 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
23914 * config/rs6000/emmintrin.h: Likewise.
23915 * config/rs6000/mmintrin.h: Likewise.
23916 * config/rs6000/xmmintrin.h: Likewise.
23918 2018-01-10 David Malcolm <dmalcolm@redhat.com>
23921 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
23923 * tree.c (tree_nop_conversion): Return true for location wrapper
23925 (maybe_wrap_with_location): New function.
23926 (selftest::check_strip_nops): New function.
23927 (selftest::test_location_wrappers): New function.
23928 (selftest::tree_c_tests): Call it.
23929 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
23930 (maybe_wrap_with_location): New decl.
23931 (EXPR_LOCATION_WRAPPER_P): New macro.
23932 (location_wrapper_p): New inline function.
23933 (tree_strip_any_location_wrapper): New inline function.
23935 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
23938 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
23939 stack_realign_offset for the largest alignment of stack slot
23941 (ix86_find_max_used_stack_alignment): New function.
23942 (ix86_finalize_stack_frame_flags): Use it. Set
23943 max_used_stack_alignment if we don't realign stack.
23944 * config/i386/i386.h (machine_function): Add
23945 max_used_stack_alignment.
23947 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
23949 * config/arm/arm.opt (-mbranch-cost): New option.
23950 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
23953 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
23956 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
23957 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
23959 2018-01-10 Richard Biener <rguenther@suse.de>
23962 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
23963 early out so it also covers the case where we have a non-NULL
23966 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
23968 PR tree-optimization/83753
23969 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
23970 for non-strided grouped accesses if the number of elements is 1.
23972 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
23975 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
23976 * i386.h (TARGET_USE_GATHER): Define.
23977 * x86-tune.def (X86_TUNE_USE_GATHER): New.
23979 2018-01-10 Martin Liska <mliska@suse.cz>
23982 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
23983 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
23985 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
23986 CLEANUP_NO_PARTITIONING is not set.
23988 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
23990 * doc/rtl.texi: Remove documentation of (const ...) wrappers
23991 for vectors, as a partial revert of r254296.
23992 * rtl.h (const_vec_p): Delete.
23993 (const_vec_duplicate_p): Don't test for vector CONSTs.
23994 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
23995 * expmed.c (make_tree): Likewise.
23998 * common.md (E, F): Use CONSTANT_P instead of checking for
24000 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
24001 checking for CONST_VECTOR.
24003 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
24005 PR middle-end/83575
24006 * predict.c (force_edge_cold): Handle in more sane way edges
24007 with no prediction.
24009 2018-01-09 Carl Love <cel@us.ibm.com>
24011 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
24013 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
24014 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
24015 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
24016 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
24017 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
24018 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
24019 * config/rs6000/rs6000-protos.h: Add extern defition for
24020 rs6000_generate_float2_double_code.
24021 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
24023 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
24024 (float2_v2df): Add define_expand.
24026 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
24029 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
24030 op_mode in the force_to_mode call.
24032 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
24034 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
24035 instead of checking each element individually.
24036 (aarch64_evpc_uzp): Likewise.
24037 (aarch64_evpc_zip): Likewise.
24038 (aarch64_evpc_ext): Likewise.
24039 (aarch64_evpc_rev): Likewise.
24040 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
24041 instead of checking each element individually. Return true without
24043 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
24044 whether all selected elements come from the same input, instead of
24045 checking each element individually. Remove calls to gen_rtx_REG,
24046 start_sequence and end_sequence and instead assert that no rtl is
24049 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
24051 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
24052 order of HIGH and CONST checks.
24054 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
24056 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
24057 if the destination isn't an SSA_NAME.
24059 2018-01-09 Richard Biener <rguenther@suse.de>
24061 PR tree-optimization/83668
24062 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
24064 (canonicalize_loop_form): ... here, renamed from ...
24065 (canonicalize_loop_closed_ssa_form): ... this and amended to
24066 swap successor edges for loop exit blocks to make us use
24067 the RPO order we need for initial schedule generation.
24069 2018-01-09 Joseph Myers <joseph@codesourcery.com>
24071 PR tree-optimization/64811
24072 * match.pd: When optimizing comparisons with Inf, avoid
24073 introducing or losing exceptions from comparisons with NaN.
24075 2018-01-09 Martin Liska <mliska@suse.cz>
24078 * asan.c (shadow_mem_size): Add gcc_assert.
24080 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
24082 Don't save registers in main().
24085 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
24086 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
24087 * config/avr/avr.c (avr_set_current_function): Don't error if
24088 naked, OS_task or OS_main are specified at the same time.
24089 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
24091 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
24093 * common/config/avr/avr-common.c (avr_option_optimization_table):
24094 Switch on -mmain-is-OS_task for optimizing compilations.
24096 2018-01-09 Richard Biener <rguenther@suse.de>
24098 PR tree-optimization/83572
24099 * graphite.c: Include cfganal.h.
24100 (graphite_transform_loops): Connect infinite loops to exit
24101 and remove fake edges at the end.
24103 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
24105 * ipa-inline.c (edge_badness): Revert accidental checkin.
24107 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
24110 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
24111 symbols; not inline clones.
24113 2018-01-09 Jakub Jelinek <jakub@redhat.com>
24116 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
24117 hard registers. Formatting fixes.
24119 PR preprocessor/83722
24120 * gcc.c (try_generate_repro): Pass
24121 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
24122 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
24125 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
24126 Kito Cheng <kito.cheng@gmail.com>
24128 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
24129 (riscv_leaf_function_p): Delete.
24130 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
24132 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
24134 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
24136 (do_ifelse): New function.
24137 (do_isel): New function.
24138 (do_sub3): New function.
24139 (do_add3): New function.
24140 (do_load_mask_compare): New function.
24141 (do_overlap_load_compare): New function.
24142 (expand_compare_loop): New function.
24143 (expand_block_compare): Call expand_compare_loop() when appropriate.
24144 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
24145 option description.
24146 (-mblock-compare-inline-loop-limit): New option.
24148 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
24151 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
24152 Reverse order of second and third operands in first alternative.
24153 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
24154 of first and second elements in UNSPEC_VPERMR vector.
24155 (altivec_expand_vec_perm_le): Likewise.
24157 2018-01-08 Jeff Law <law@redhat.com>
24159 PR rtl-optimizatin/81308
24160 * tree-switch-conversion.c (cfg_altered): New file scoped static.
24161 (process_switch): If group_case_labels makes a change, then set
24163 (pass_convert_switch::execute): If a switch is converted, then
24164 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
24166 PR rtl-optimization/81308
24167 * recog.c (split_all_insns): Conditionally cleanup the CFG after
24170 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
24172 PR target/83663 - Revert r255946
24173 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
24174 generation for cases where splatting a value is not useful.
24175 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
24176 across a vec_duplicate and a paradoxical subreg forming a vector
24177 mode to a vec_concat.
24179 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24181 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
24182 -march=armv8.3-a variants.
24183 * config/arm/t-multilib: Likewise.
24184 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
24186 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
24188 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
24190 (cceq_ior_compare_complement): Give it a name so I can use it, and
24191 change boolean_or_operator predicate to boolean_operator so it can
24192 be used to generate a crand.
24193 (eqne): New code iterator.
24194 (bd/bd_neg): New code_attrs.
24195 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
24196 a single define_insn.
24197 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
24198 decrement (bdnzt/bdnzf/bdzt/bdzf).
24199 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
24200 with the new names of the branch decrement patterns, and added the
24201 names of the branch decrement conditional patterns.
24203 2018-01-08 Richard Biener <rguenther@suse.de>
24205 PR tree-optimization/83563
24206 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
24209 2018-01-08 Richard Biener <rguenther@suse.de>
24211 PR middle-end/83713
24212 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
24214 2018-01-08 Richard Biener <rguenther@suse.de>
24216 PR tree-optimization/83685
24217 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
24218 references to abnormals.
24220 2018-01-08 Richard Biener <rguenther@suse.de>
24223 * dwarf2out.c (output_indirect_strings): Handle empty
24224 skeleton_debug_str_hash.
24225 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
24227 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
24229 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
24230 (emit_store_direct): Likewise.
24231 (arc_trampoline_adjust_address): Likewise.
24232 (arc_asm_trampoline_template): New function.
24233 (arc_initialize_trampoline): Use asm_trampoline_template.
24234 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
24235 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
24236 * config/arc/arc.md (flush_icache): Delete pattern.
24238 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
24240 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
24241 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
24244 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
24247 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
24248 by not USED_FOR_TARGET.
24249 (make_pass_resolve_sw_modes): Likewise.
24251 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
24253 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
24256 2018-01-08 Richard Biener <rguenther@suse.de>
24258 PR middle-end/83580
24259 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
24261 2018-01-08 Richard Biener <rguenther@suse.de>
24263 PR middle-end/83517
24264 * match.pd ((t * 2) / 2) -> t): Add missing :c.
24266 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
24268 PR middle-end/81897
24269 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
24270 basic blocks with a small number of successors.
24271 (convert_control_dep_chain_into_preds): Improve handling of
24273 (dump_predicates): Split apart into...
24274 (dump_pred_chain): ...here...
24275 (dump_pred_info): ...and here.
24276 (can_one_predicate_be_invalidated_p): Add debugging printfs.
24277 (can_chain_union_be_invalidated_p): Improve check for invalidation
24279 (uninit_uses_cannot_happen): Avoid unnecessary if
24280 convert_control_dep_chain_into_preds yielded nothing.
24282 2018-01-06 Martin Sebor <msebor@redhat.com>
24284 PR tree-optimization/83640
24285 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
24286 subtracting negative offset from size.
24287 (builtin_access::overlap): Adjust offset bounds of the access to fall
24288 within the size of the object if possible.
24290 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
24292 PR rtl-optimization/83699
24293 * expmed.c (extract_bit_field_1): Restrict the vector usage of
24294 extract_bit_field_as_subreg to cases in which the extracted
24295 value is also a vector.
24297 * lra-constraints.c (process_alt_operands): Test for the equivalence
24298 substitutions when detecting a possible reload cycle.
24300 2018-01-06 Jakub Jelinek <jakub@redhat.com>
24303 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
24304 by default if flag_selective_schedling{,2}. Formatting fixes.
24306 PR rtl-optimization/83682
24307 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
24308 if it has non-VECTOR_MODE element mode.
24309 (vec_duplicate_p): Likewise.
24311 PR middle-end/83694
24312 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
24313 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
24315 2018-01-05 Jakub Jelinek <jakub@redhat.com>
24318 * config/i386/i386-builtin.def
24319 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
24320 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
24321 Require also OPTION_MASK_ISA_AVX512F in addition to
24322 OPTION_MASK_ISA_GFNI.
24323 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
24324 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
24325 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
24326 to OPTION_MASK_ISA_GFNI.
24327 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
24328 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
24329 OPTION_MASK_ISA_AVX512BW.
24330 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
24331 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
24332 addition to OPTION_MASK_ISA_GFNI.
24333 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
24334 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
24335 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
24336 to OPTION_MASK_ISA_GFNI.
24337 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
24338 a requirement for all ISAs rather than any of them with a few
24340 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
24342 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
24343 bitmasks to be enabled with 3 exceptions, instead of requiring any
24344 enabled ISA with lots of exceptions.
24345 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
24346 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
24347 Change avx512bw in isa attribute to avx512f.
24348 * config/i386/sgxintrin.h: Add license boilerplate.
24349 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
24350 to __AVX512F__ and __AVX512VL to __AVX512VL__.
24351 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
24352 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
24354 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
24355 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
24356 temporarily sse2 rather than sse if not enabled already.
24359 * config/i386/sse.md (VI248_VLBW): Rename to ...
24360 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
24361 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
24362 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
24363 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
24364 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
24365 mode iterator instead of VI248_VLBW.
24367 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
24369 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
24370 (record_modified): Skip clobbers; add debug output.
24371 (param_change_prob): Use sreal frequencies.
24373 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
24375 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
24376 punt for user-aligned variables.
24378 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
24380 * tree-chrec.c (chrec_contains_symbols): Return true for
24383 2018-01-05 Sudakshina Das <sudi.das@arm.com>
24386 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
24387 of (x|y) == x for BICS pattern.
24389 2018-01-05 Jakub Jelinek <jakub@redhat.com>
24391 PR tree-optimization/83605
24392 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
24393 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
24396 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
24398 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
24399 * config/epiphany/rtems.h: New file.
24401 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24402 Uros Bizjak <ubizjak@gmail.com>
24405 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
24406 QIreg_operand instead of register_operand predicate.
24407 * config/i386/i386.c (ix86_rop_should_change_byte_p,
24408 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
24409 comments instead of -fmitigate[-_]rop.
24411 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
24414 * cgraphunit.c (symbol_table::compile): Switch to text_section
24415 before calling assembly_start debug hook.
24416 * run-rtl-passes.c (run_rtl_passes): Likewise.
24419 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24421 * tree-vrp.c (extract_range_from_binary_expr_1): Check
24422 range_int_cst_p rather than !symbolic_range_p before calling
24423 extract_range_from_multiplicative_op_1.
24425 2018-01-04 Jeff Law <law@redhat.com>
24427 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
24428 redundant test in assertion.
24430 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24432 * doc/rtl.texi: Document machine_mode wrapper classes.
24434 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24436 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
24437 using tree_to_uhwi.
24439 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24441 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
24442 the VEC_PERM_EXPR fold to fail.
24444 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24447 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
24448 to switched_sections.
24450 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24453 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
24454 test for d.testing.
24456 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
24459 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
24460 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
24462 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24465 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
24466 is BLKmode and bitpos not zero or mode change is needed.
24468 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24471 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
24474 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
24477 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
24478 instead of MULT rtx. Update all corresponding splitters.
24479 (*saddl_se): Ditto.
24480 (*ssub<modesuffix>): Ditto.
24481 (*ssubl_se): Ditto.
24482 (*cmp_sadd_di): Update split patterns.
24483 (*cmp_sadd_si): Ditto.
24484 (*cmp_sadd_sidi): Ditto.
24485 (*cmp_ssub_di): Ditto.
24486 (*cmp_ssub_si): Ditto.
24487 (*cmp_ssub_sidi): Ditto.
24488 * config/alpha/predicates.md (const23_operand): New predicate.
24489 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
24490 Look for ASHIFT, not MULT inner operand.
24491 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
24493 2018-01-04 Martin Liska <mliska@suse.cz>
24495 PR gcov-profile/83669
24496 * gcov.c (output_intermediate_file): Add version to intermediate
24498 * doc/gcov.texi: Document new field 'version' in intermediate
24499 file format. Fix location of '-k' option of gcov command.
24501 2018-01-04 Martin Liska <mliska@suse.cz>
24504 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
24506 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24508 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
24510 2018-01-03 Martin Sebor <msebor@redhat.com>
24512 PR tree-optimization/83655
24513 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
24514 checking calls with invalid arguments.
24516 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24518 * tree-vect-stmts.c (vect_get_store_rhs): New function.
24519 (vectorizable_mask_load_store): Delete.
24520 (vectorizable_call): Return false for masked loads and stores.
24521 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
24522 instead of gimple_assign_rhs1.
24523 (vectorizable_load): Handle IFN_MASK_LOAD.
24524 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
24526 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24528 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
24530 (vectorizable_mask_load_store): ...here.
24531 (vectorizable_load): ...and here.
24533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24535 * tree-vect-stmts.c (vect_build_all_ones_mask)
24536 (vect_build_zero_merge_argument): New functions, split out from...
24537 (vectorizable_load): ...here.
24539 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24541 * tree-vect-stmts.c (vect_check_store_rhs): New function,
24543 (vectorizable_mask_load_store): ...here.
24544 (vectorizable_store): ...and here.
24546 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24548 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
24550 (vectorizable_mask_load_store): ...here.
24552 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24554 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
24555 (vect_model_store_cost): Take a vec_load_store_type instead of a
24557 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
24558 (vect_model_store_cost): Take a vec_load_store_type instead of a
24560 (vectorizable_mask_load_store): Update accordingly.
24561 (vectorizable_store): Likewise.
24562 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
24564 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24566 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
24567 IFN_MASK_LOAD calls here rather than...
24568 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
24570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24571 Alan Hayward <alan.hayward@arm.com>
24572 David Sherwood <david.sherwood@arm.com>
24574 * expmed.c (extract_bit_field_1): For vector extracts,
24575 fall back to extract_bit_field_as_subreg if vec_extract
24578 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24579 Alan Hayward <alan.hayward@arm.com>
24580 David Sherwood <david.sherwood@arm.com>
24582 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
24583 they are variable or constant sized.
24584 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
24585 slots for constant-sized data.
24587 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24588 Alan Hayward <alan.hayward@arm.com>
24589 David Sherwood <david.sherwood@arm.com>
24591 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
24592 handling COND_EXPRs with boolean comparisons, try to find a better
24593 basis for the mask type than the boolean itself.
24595 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24597 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
24598 is calculated and how it can be overridden.
24599 * genmodes.c (max_bitsize_mode_any_mode): New variable.
24600 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
24602 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
24605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24606 Alan Hayward <alan.hayward@arm.com>
24607 David Sherwood <david.sherwood@arm.com>
24609 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
24610 Remove the mode argument.
24611 (aarch64_simd_valid_immediate): Remove the mode and inverse
24613 * config/aarch64/iterators.md (bitsize): New iterator.
24614 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
24615 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
24616 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
24617 aarch64_simd_valid_immediate.
24618 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
24619 (aarch64_reg_or_bic_imm): Likewise.
24620 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
24621 with an insn_type enum and msl with a modifier_type enum.
24622 Replace element_width with a scalar_mode. Change the shift
24623 to unsigned int. Add constructors for scalar_float_mode and
24624 scalar_int_mode elements.
24625 (aarch64_vect_float_const_representable_p): Delete.
24626 (aarch64_can_const_movi_rtx_p)
24627 (aarch64_simd_scalar_immediate_valid_for_move)
24628 (aarch64_simd_make_constant): Update call to
24629 aarch64_simd_valid_immediate.
24630 (aarch64_advsimd_valid_immediate_hs): New function.
24631 (aarch64_advsimd_valid_immediate): Likewise.
24632 (aarch64_simd_valid_immediate): Remove mode and inverse
24633 arguments. Rewrite to use the above. Use const_vec_duplicate_p
24634 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
24635 and aarch64_float_const_representable_p on the result.
24636 (aarch64_output_simd_mov_immediate): Remove mode argument.
24637 Update call to aarch64_simd_valid_immediate and use of
24638 simd_immediate_info.
24639 (aarch64_output_scalar_simd_mov_immediate): Update call
24642 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24643 Alan Hayward <alan.hayward@arm.com>
24644 David Sherwood <david.sherwood@arm.com>
24646 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
24647 (mode_nunits): Likewise CONST_MODE_NUNITS.
24648 * machmode.def (ADJUST_NUNITS): Document.
24649 * genmodes.c (mode_data::need_nunits_adj): New field.
24650 (blank_mode): Update accordingly.
24651 (adj_nunits): New variable.
24652 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
24654 (emit_mode_size_inline): Set need_bytesize_adj for all modes
24655 listed in adj_nunits.
24656 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
24657 listed in adj_nunits. Don't emit case statements for such modes.
24658 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
24659 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
24660 nothing if adj_nunits is nonnull.
24661 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
24662 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
24663 (emit_mode_fbit): Update use of print_maybe_const_decl.
24664 (emit_move_size): Likewise. Treat the array as non-const
24666 (emit_mode_adjustments): Handle adj_nunits.
24668 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24670 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
24671 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
24672 (VECTOR_MODES): Use it.
24673 (make_vector_modes): Take the prefix as an argument.
24675 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24676 Alan Hayward <alan.hayward@arm.com>
24677 David Sherwood <david.sherwood@arm.com>
24679 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
24680 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
24681 for MODE_VECTOR_BOOL.
24682 * machmode.def (VECTOR_BOOL_MODE): Document.
24683 * genmodes.c (VECTOR_BOOL_MODE): New macro.
24684 (make_vector_bool_mode): New function.
24685 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
24687 * lto-streamer-in.c (lto_input_mode_table): Likewise.
24688 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
24690 * stor-layout.c (int_mode_for_mode): Likewise.
24691 * tree.c (build_vector_type_for_mode): Likewise.
24692 * varasm.c (output_constant_pool_2): Likewise.
24693 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
24694 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
24695 for MODE_VECTOR_BOOL.
24696 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
24697 of mode class checks.
24698 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
24699 instead of a list of mode class checks.
24700 (expand_vector_scalar_condition): Likewise.
24701 (type_for_widest_vector_mode): Handle BImode as an inner mode.
24703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24704 Alan Hayward <alan.hayward@arm.com>
24705 David Sherwood <david.sherwood@arm.com>
24707 * machmode.h (mode_size): Change from unsigned short to
24709 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
24710 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
24711 or if measurement_type is not polynomial.
24712 (fixed_size_mode::includes_p): Check for constant-sized modes.
24713 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
24714 return a poly_uint16 rather than an unsigned short.
24715 (emit_mode_size): Change the type of mode_size from unsigned short
24716 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
24717 (emit_mode_adjustments): Cope with polynomial vector sizes.
24718 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24720 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24722 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
24723 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
24724 * caller-save.c (setup_save_areas): Likewise.
24725 (replace_reg_with_saved_mem): Likewise.
24726 * calls.c (emit_library_call_value_1): Likewise.
24727 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
24728 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
24729 (gen_lowpart_for_combine): Likewise.
24730 * convert.c (convert_to_integer_1): Likewise.
24731 * cse.c (equiv_constant, cse_insn): Likewise.
24732 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
24733 (cselib_subst_to_values): Likewise.
24734 * dce.c (word_dce_process_block): Likewise.
24735 * df-problems.c (df_word_lr_mark_ref): Likewise.
24736 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
24737 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
24738 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
24739 (rtl_for_decl_location): Likewise.
24740 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
24741 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
24742 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
24743 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
24744 (expand_expr_real_1): Likewise.
24745 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
24746 (pad_below): Likewise.
24747 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
24748 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
24749 * ira.c (get_subreg_tracking_sizes): Likewise.
24750 * ira-build.c (ira_create_allocno_objects): Likewise.
24751 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
24752 (ira_sort_regnos_for_alter_reg): Likewise.
24753 * ira-costs.c (record_operand_costs): Likewise.
24754 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
24755 (resolve_simple_move): Likewise.
24756 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
24757 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
24758 (lra_constraints): Likewise.
24759 (CONST_POOL_OK_P): Reject variable-sized modes.
24760 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
24761 (add_pseudo_to_slot, lra_spill): Likewise.
24762 * omp-low.c (omp_clause_aligned_alignment): Likewise.
24763 * optabs-query.c (get_best_extraction_insn): Likewise.
24764 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
24765 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
24766 (expand_mult_highpart, valid_multiword_target_p): Likewise.
24767 * recog.c (offsettable_address_addr_space_p): Likewise.
24768 * regcprop.c (maybe_mode_change): Likewise.
24769 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
24770 * regrename.c (build_def_use): Likewise.
24771 * regstat.c (dump_reg_info): Likewise.
24772 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
24773 (find_reloads, find_reloads_subreg_address): Likewise.
24774 * reload1.c (eliminate_regs_1): Likewise.
24775 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
24776 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
24777 (simplify_binary_operation_1, simplify_subreg): Likewise.
24778 * targhooks.c (default_function_arg_padding): Likewise.
24779 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
24780 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
24781 (verify_gimple_assign_ternary): Likewise.
24782 * tree-inline.c (estimate_move_cost): Likewise.
24783 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24784 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
24785 (get_address_cost_ainc): Likewise.
24786 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
24787 (vect_supportable_dr_alignment): Likewise.
24788 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
24789 (vectorizable_reduction): Likewise.
24790 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
24791 (vectorizable_operation, vectorizable_load): Likewise.
24792 * tree.c (build_same_sized_truth_vector_type): Likewise.
24793 * valtrack.c (cleanup_auto_inc_dec): Likewise.
24794 * var-tracking.c (emit_note_insn_var_location): Likewise.
24795 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
24796 (ADDR_VEC_ALIGN): Likewise.
24798 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24799 Alan Hayward <alan.hayward@arm.com>
24800 David Sherwood <david.sherwood@arm.com>
24802 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
24804 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
24805 or if measurement_type is polynomial.
24806 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
24807 * combine.c (make_extraction): Likewise.
24808 * dse.c (find_shift_sequence): Likewise.
24809 * dwarf2out.c (mem_loc_descriptor): Likewise.
24810 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
24811 (extract_bit_field, extract_low_bits): Likewise.
24812 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
24813 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
24814 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
24815 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
24816 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
24817 * reload.c (find_reloads): Likewise.
24818 * reload1.c (alter_reg): Likewise.
24819 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
24820 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
24821 * tree-if-conv.c (predicate_mem_writes): Likewise.
24822 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
24823 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
24824 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
24825 * valtrack.c (dead_debug_insert_temp): Likewise.
24826 * varasm.c (mergeable_constant_section): Likewise.
24827 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
24829 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24830 Alan Hayward <alan.hayward@arm.com>
24831 David Sherwood <david.sherwood@arm.com>
24833 * expr.c (expand_assignment): Cope with polynomial mode sizes
24834 when assigning to a CONCAT.
24836 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24837 Alan Hayward <alan.hayward@arm.com>
24838 David Sherwood <david.sherwood@arm.com>
24840 * machmode.h (mode_precision): Change from unsigned short to
24842 (mode_to_precision): Return a poly_uint16 rather than an unsigned
24844 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
24845 or if measurement_type is not polynomial.
24846 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
24847 in which the mode is already known to be a scalar_int_mode.
24848 * genmodes.c (emit_mode_precision): Change the type of mode_precision
24849 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
24851 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24852 for GET_MODE_PRECISION.
24853 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24854 for GET_MODE_PRECISION.
24855 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
24857 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
24858 (expand_field_assignment, make_extraction): Likewise.
24859 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
24860 (get_last_value): Likewise.
24861 * convert.c (convert_to_integer_1): Likewise.
24862 * cse.c (cse_insn): Likewise.
24863 * expr.c (expand_expr_real_1): Likewise.
24864 * lra-constraints.c (simplify_operand_subreg): Likewise.
24865 * optabs-query.c (can_atomic_load_p): Likewise.
24866 * optabs.c (expand_atomic_load): Likewise.
24867 (expand_atomic_store): Likewise.
24868 * ree.c (combine_reaching_defs): Likewise.
24869 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
24870 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
24871 * tree.h (type_has_mode_precision_p): Likewise.
24872 * ubsan.c (instrument_si_overflow): Likewise.
24874 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24875 Alan Hayward <alan.hayward@arm.com>
24876 David Sherwood <david.sherwood@arm.com>
24878 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
24879 polynomial numbers of units.
24880 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
24881 (valid_vector_subparts_p): New function.
24882 (build_vector_type): Remove temporary shim and take the number
24883 of units as a poly_uint64 rather than an int.
24884 (build_opaque_vector_type): Take the number of units as a
24885 poly_uint64 rather than an int.
24886 * tree.c (build_vector_from_ctor): Handle polynomial
24887 TYPE_VECTOR_SUBPARTS.
24888 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
24889 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
24890 (build_vector_from_val): If the number of units is variable,
24891 use build_vec_duplicate_cst for constant operands and
24892 VEC_DUPLICATE_EXPR otherwise.
24893 (make_vector_type): Remove temporary is_constant ().
24894 (build_vector_type, build_opaque_vector_type): Take the number of
24895 units as a poly_uint64 rather than an int.
24896 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
24898 * cfgexpand.c (expand_debug_expr): Likewise.
24899 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
24900 (store_constructor, expand_expr_real_1): Likewise.
24901 (const_scalar_mask_from_tree): Likewise.
24902 * fold-const-call.c (fold_const_reduction): Likewise.
24903 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
24904 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
24905 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
24906 (fold_relational_const): Likewise.
24907 (native_interpret_vector): Likewise. Change the size from an
24908 int to an unsigned int.
24909 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
24910 TYPE_VECTOR_SUBPARTS.
24911 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
24912 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
24913 duplicating a non-constant operand into a variable-length vector.
24914 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
24915 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
24916 * ipa-icf.c (sem_variable::equals): Likewise.
24917 * match.pd: Likewise.
24918 * omp-simd-clone.c (simd_clone_subparts): Likewise.
24919 * print-tree.c (print_node): Likewise.
24920 * stor-layout.c (layout_type): Likewise.
24921 * targhooks.c (default_builtin_vectorization_cost): Likewise.
24922 * tree-cfg.c (verify_gimple_comparison): Likewise.
24923 (verify_gimple_assign_binary): Likewise.
24924 (verify_gimple_assign_ternary): Likewise.
24925 (verify_gimple_assign_single): Likewise.
24926 * tree-pretty-print.c (dump_generic_node): Likewise.
24927 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24928 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
24929 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
24930 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
24931 (vect_shift_permute_load_chain): Likewise.
24932 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
24933 (expand_vector_condition, optimize_vector_constructor): Likewise.
24934 (lower_vec_perm, get_compute_type): Likewise.
24935 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
24936 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
24937 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
24938 (vect_recog_mask_conversion_pattern): Likewise.
24939 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
24940 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
24941 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
24942 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
24943 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
24944 (vectorizable_shift, vectorizable_operation, vectorizable_store)
24945 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
24946 (supportable_widening_operation): Likewise.
24947 (supportable_narrowing_operation): Likewise.
24948 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
24950 * varasm.c (output_constant): 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 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
24957 so that both the length == 3 and length != 3 cases set up their
24958 own permute vectors. Add comments explaining why we know the
24959 number of elements is constant.
24960 (vect_permute_load_chain): Likewise.
24962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24963 Alan Hayward <alan.hayward@arm.com>
24964 David Sherwood <david.sherwood@arm.com>
24966 * machmode.h (mode_nunits): Change from unsigned char to
24968 (ONLY_FIXED_SIZE_MODES): New macro.
24969 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
24970 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
24971 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
24973 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
24974 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
24975 or if measurement_type is not polynomial.
24976 * genmodes.c (ZERO_COEFFS): New macro.
24977 (emit_mode_nunits_inline): Make mode_nunits_inline return a
24979 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
24980 Use ZERO_COEFFS when emitting initializers.
24981 * data-streamer.h (bp_pack_poly_value): New function.
24982 (bp_unpack_poly_value): Likewise.
24983 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24984 for GET_MODE_NUNITS.
24985 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24986 for GET_MODE_NUNITS.
24987 * tree.c (make_vector_type): Remove temporary shim and make
24988 the real function take the number of units as a poly_uint64
24989 rather than an int.
24990 (build_vector_type_for_mode): Handle polynomial nunits.
24991 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
24992 * emit-rtl.c (const_vec_series_p_1): Likewise.
24993 (gen_rtx_CONST_VECTOR): Likewise.
24994 * fold-const.c (test_vec_duplicate_folding): Likewise.
24995 * genrecog.c (validate_pattern): Likewise.
24996 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
24997 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
24998 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
24999 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
25000 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
25001 * rtlanal.c (subreg_get_info): Likewise.
25002 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
25003 (vect_grouped_load_supported): Likewise.
25004 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
25005 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
25006 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
25007 (simplify_const_unary_operation, simplify_binary_operation_1)
25008 (simplify_const_binary_operation, simplify_ternary_operation)
25009 (test_vector_ops_duplicate, test_vector_ops): Likewise.
25010 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
25011 instead of CONST_VECTOR_NUNITS.
25012 * varasm.c (output_constant_pool_2): Likewise.
25013 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
25014 explicit-encoded elements in the XVEC for variable-length vectors.
25016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25018 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
25020 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25021 Alan Hayward <alan.hayward@arm.com>
25022 David Sherwood <david.sherwood@arm.com>
25024 * coretypes.h (fixed_size_mode): Declare.
25025 (fixed_size_mode_pod): New typedef.
25026 * builtins.h (target_builtins::x_apply_args_mode)
25027 (target_builtins::x_apply_result_mode): Change type to
25028 fixed_size_mode_pod.
25029 * builtins.c (apply_args_size, apply_result_size, result_vector)
25030 (expand_builtin_apply_args_1, expand_builtin_apply)
25031 (expand_builtin_return): Update accordingly.
25033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25035 * cse.c (hash_rtx_cb): Hash only the encoded elements.
25036 * cselib.c (cselib_hash_rtx): Likewise.
25037 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
25038 CONST_VECTOR encoding.
25040 2018-01-03 Jakub Jelinek <jakub@redhat.com>
25041 Jeff Law <law@redhat.com>
25044 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
25045 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
25046 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
25047 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
25050 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
25051 explicitly probe *sp in a noreturn function if there were any callee
25052 register saves or frame pointer is needed.
25054 2018-01-03 Jakub Jelinek <jakub@redhat.com>
25057 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
25058 BLKmode for ternary, binary or unary expressions.
25061 * var-tracking.c (delete_vta_debug_insn): New inline function.
25062 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
25063 insns from get_insns () to NULL instead of each bb separately.
25064 Use delete_vta_debug_insn. No longer static.
25065 (vt_debug_insns_local, variable_tracking_main_1): Adjust
25066 delete_vta_debug_insns callers.
25067 * rtl.h (delete_vta_debug_insns): Declare.
25068 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
25069 instead of variable_tracking_main.
25071 2018-01-03 Martin Sebor <msebor@redhat.com>
25073 PR tree-optimization/83603
25074 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
25075 arguments past the endof the argument list in functions declared
25076 without a prototype.
25077 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
25078 Avoid checking when arguments are null.
25080 2018-01-03 Martin Sebor <msebor@redhat.com>
25083 * doc/extend.texi (attribute const): Fix a typo.
25084 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
25085 issuing -Wsuggest-attribute for void functions.
25087 2018-01-03 Martin Sebor <msebor@redhat.com>
25089 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
25090 offset_int::from instead of wide_int::to_shwi.
25091 (maybe_diag_overlap): Remove assertion.
25092 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
25093 * gimple-ssa-sprintf.c (format_directive): Same.
25094 (parse_directive): Same.
25095 (sprintf_dom_walker::compute_format_length): Same.
25096 (try_substitute_return_value): Same.
25098 2018-01-03 Jeff Law <law@redhat.com>
25100 PR middle-end/83654
25101 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
25102 non-constant residual for zero at runtime and avoid probing in
25103 that case. Reorganize code for trailing problem to mirror handling
25106 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
25108 PR tree-optimization/83501
25109 * tree-ssa-strlen.c (get_string_cst): New.
25110 (handle_char_store): Call get_string_cst.
25112 2018-01-03 Martin Liska <mliska@suse.cz>
25114 PR tree-optimization/83593
25115 * tree-ssa-strlen.c: Include tree-cfg.h.
25116 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
25117 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
25118 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
25120 (strlen_dom_walker::before_dom_children): Call
25121 gimple_purge_dead_eh_edges. Dump tranformation with details
25123 (strlen_dom_walker::before_dom_children): Update call by adding
25124 new argument cleanup_eh.
25125 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
25127 2018-01-03 Martin Liska <mliska@suse.cz>
25130 * cif-code.def (VARIADIC_THUNK): New enum value.
25131 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
25134 2018-01-03 Jan Beulich <jbeulich@suse.com>
25136 * sse.md (mov<mode>_internal): Tighten condition for when to use
25137 vmovdqu<ssescalarsize> for TI and OI modes.
25139 2018-01-03 Jakub Jelinek <jakub@redhat.com>
25141 Update copyright years.
25143 2018-01-03 Martin Liska <mliska@suse.cz>
25146 * ipa-visibility.c (function_and_variable_visibility): Skip
25147 functions with noipa attribure.
25149 2018-01-03 Jakub Jelinek <jakub@redhat.com>
25151 * gcc.c (process_command): Update copyright notice dates.
25152 * gcov-dump.c (print_version): Ditto.
25153 * gcov.c (print_version): Ditto.
25154 * gcov-tool.c (print_version): Ditto.
25155 * gengtype.c (create_file): Ditto.
25156 * doc/cpp.texi: Bump @copying's copyright year.
25157 * doc/cppinternals.texi: Ditto.
25158 * doc/gcc.texi: Ditto.
25159 * doc/gccint.texi: Ditto.
25160 * doc/gcov.texi: Ditto.
25161 * doc/install.texi: Ditto.
25162 * doc/invoke.texi: Ditto.
25164 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25166 * vector-builder.h (vector_builder::m_full_nelts): Change from
25167 unsigned int to poly_uint64.
25168 (vector_builder::full_nelts): Update prototype accordingly.
25169 (vector_builder::new_vector): Likewise.
25170 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
25171 (vector_builder::operator ==): Likewise.
25172 (vector_builder::finalize): Likewise.
25173 * int-vector-builder.h (int_vector_builder::int_vector_builder):
25174 Take the number of elements as a poly_uint64 rather than an
25176 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
25177 from unsigned int to poly_uint64.
25178 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
25179 (vec_perm_indices::new_vector): Likewise.
25180 (vec_perm_indices::length): Likewise.
25181 (vec_perm_indices::nelts_per_input): Likewise.
25182 (vec_perm_indices::input_nelts): Likewise.
25183 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
25184 number of elements per input as a poly_uint64 rather than an
25185 unsigned int. Use the original encoding for variable-length
25186 vectors, rather than clamping each individual element.
25187 For the second and subsequent elements in each pattern,
25188 clamp the step and base before clamping their sum.
25189 (vec_perm_indices::series_p): Handle polynomial element counts.
25190 (vec_perm_indices::all_in_range_p): Likewise.
25191 (vec_perm_indices_to_tree): Likewise.
25192 (vec_perm_indices_to_rtx): Likewise.
25193 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
25194 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
25195 (tree_vector_builder::new_binary_operation): Handle polynomial
25196 element counts. Return false if we need to know the number
25197 of elements at compile time.
25198 * fold-const.c (fold_vec_perm): Punt if the number of elements
25199 isn't known at compile time.
25201 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25203 * vec-perm-indices.h (vec_perm_builder): Change element type
25204 from HOST_WIDE_INT to poly_int64.
25205 (vec_perm_indices::element_type): Update accordingly.
25206 (vec_perm_indices::clamp): Handle polynomial element_types.
25207 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
25208 (vec_perm_indices::all_in_range_p): Likewise.
25209 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
25211 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
25212 polynomial vec_perm_indices element types.
25213 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
25214 * fold-const.c (fold_vec_perm): Likewise.
25215 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
25216 * tree-vect-generic.c (lower_vec_perm): Likewise.
25217 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
25218 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
25219 element type to HOST_WIDE_INT.
25221 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25222 Alan Hayward <alan.hayward@arm.com>
25223 David Sherwood <david.sherwood@arm.com>
25225 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
25226 rather than an int. Use plus_constant.
25227 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
25228 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
25230 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25231 Alan Hayward <alan.hayward@arm.com>
25232 David Sherwood <david.sherwood@arm.com>
25234 * calls.c (emit_call_1, expand_call): Change struct_value_size from
25235 a HOST_WIDE_INT to a poly_int64.
25237 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25238 Alan Hayward <alan.hayward@arm.com>
25239 David Sherwood <david.sherwood@arm.com>
25241 * calls.c (load_register_parameters): Cope with polynomial
25242 mode sizes. Require a constant size for BLKmode parameters
25243 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
25244 forces a parameter to be padded at the lsb end in order to
25245 fill a complete number of words, require the parameter size
25246 to be ordered wrt UNITS_PER_WORD.
25248 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25249 Alan Hayward <alan.hayward@arm.com>
25250 David Sherwood <david.sherwood@arm.com>
25252 * reload1.c (spill_stack_slot_width): Change element type
25253 from unsigned int to poly_uint64_pod.
25254 (alter_reg): Treat mode sizes as polynomial.
25256 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25257 Alan Hayward <alan.hayward@arm.com>
25258 David Sherwood <david.sherwood@arm.com>
25260 * reload.c (complex_word_subreg_p): New function.
25261 (reload_inner_reg_of_subreg, push_reload): Use it.
25263 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25264 Alan Hayward <alan.hayward@arm.com>
25265 David Sherwood <david.sherwood@arm.com>
25267 * lra-constraints.c (process_alt_operands): Reject matched
25268 operands whose sizes aren't ordered.
25269 (match_reload): Refer to this check here.
25271 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25272 Alan Hayward <alan.hayward@arm.com>
25273 David Sherwood <david.sherwood@arm.com>
25275 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
25276 that the mode size is in the set {1, 2, 4, 8, 16}.
25278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25279 Alan Hayward <alan.hayward@arm.com>
25280 David Sherwood <david.sherwood@arm.com>
25282 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
25283 Use plus_constant instead of gen_rtx_PLUS.
25285 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25286 Alan Hayward <alan.hayward@arm.com>
25287 David Sherwood <david.sherwood@arm.com>
25289 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
25290 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
25291 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
25292 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
25293 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
25294 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
25295 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
25296 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
25297 * config/i386/i386.c (ix86_push_rounding): ...this new function.
25298 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
25300 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
25301 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
25302 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
25303 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
25304 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
25305 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
25306 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
25307 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
25308 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
25309 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
25311 * expr.c (emit_move_resolve_push): Treat the input and result
25312 of PUSH_ROUNDING as a poly_int64.
25313 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
25314 (emit_push_insn): Likewise.
25315 * lra-eliminations.c (mark_not_eliminable): Likewise.
25316 * recog.c (push_operand): Likewise.
25317 * reload1.c (elimination_effects): Likewise.
25318 * rtlanal.c (nonzero_bits1): Likewise.
25319 * calls.c (store_one_arg): Likewise. Require the padding to be
25320 known at compile time.
25322 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25323 Alan Hayward <alan.hayward@arm.com>
25324 David Sherwood <david.sherwood@arm.com>
25326 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
25327 Use plus_constant instead of gen_rtx_PLUS.
25329 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25330 Alan Hayward <alan.hayward@arm.com>
25331 David Sherwood <david.sherwood@arm.com>
25333 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
25334 rather than an int.
25336 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25337 Alan Hayward <alan.hayward@arm.com>
25338 David Sherwood <david.sherwood@arm.com>
25340 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
25341 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
25342 via stack temporaries. Treat the mode size as polynomial too.
25344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25345 Alan Hayward <alan.hayward@arm.com>
25346 David Sherwood <david.sherwood@arm.com>
25348 * expr.c (expand_expr_real_2): When handling conversions involving
25349 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
25350 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
25351 as a poly_uint64 too.
25353 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25354 Alan Hayward <alan.hayward@arm.com>
25355 David Sherwood <david.sherwood@arm.com>
25357 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
25359 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25360 Alan Hayward <alan.hayward@arm.com>
25361 David Sherwood <david.sherwood@arm.com>
25363 * combine.c (can_change_dest_mode): Handle polynomial
25364 REGMODE_NATURAL_SIZE.
25365 * expmed.c (store_bit_field_1): Likewise.
25366 * expr.c (store_constructor): Likewise.
25367 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
25368 and polynomial REGMODE_NATURAL_SIZE.
25369 (gen_lowpart_common): Likewise.
25370 * reginfo.c (record_subregs_of_mode): Likewise.
25371 * rtlanal.c (read_modify_subreg_p): Likewise.
25373 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25374 Alan Hayward <alan.hayward@arm.com>
25375 David Sherwood <david.sherwood@arm.com>
25377 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
25378 numbers of elements.
25380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25381 Alan Hayward <alan.hayward@arm.com>
25382 David Sherwood <david.sherwood@arm.com>
25384 * match.pd: Cope with polynomial numbers of vector elements.
25386 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25387 Alan Hayward <alan.hayward@arm.com>
25388 David Sherwood <david.sherwood@arm.com>
25390 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
25391 in a POINTER_PLUS_EXPR.
25393 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25394 Alan Hayward <alan.hayward@arm.com>
25395 David Sherwood <david.sherwood@arm.com>
25397 * omp-simd-clone.c (simd_clone_subparts): New function.
25398 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
25399 (ipa_simd_modify_function_body): Likewise.
25401 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25402 Alan Hayward <alan.hayward@arm.com>
25403 David Sherwood <david.sherwood@arm.com>
25405 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
25406 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
25407 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
25408 (expand_vector_condition, vector_element): Likewise.
25409 (subparts_gt): New function.
25410 (get_compute_type): Use subparts_gt.
25411 (count_type_subparts): Delete.
25412 (expand_vector_operations_1): Use subparts_gt instead of
25413 count_type_subparts.
25415 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25416 Alan Hayward <alan.hayward@arm.com>
25417 David Sherwood <david.sherwood@arm.com>
25419 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
25420 (vect_compile_time_alias): ...this new function. Do the calculation
25421 on poly_ints rather than trees.
25422 (vect_prune_runtime_alias_test_list): Update call accordingly.
25424 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25425 Alan Hayward <alan.hayward@arm.com>
25426 David Sherwood <david.sherwood@arm.com>
25428 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
25430 (vect_schedule_slp_instance): Likewise.
25432 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25433 Alan Hayward <alan.hayward@arm.com>
25434 David Sherwood <david.sherwood@arm.com>
25436 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
25437 constant and extern definitions for variable-length vectors.
25438 (vect_get_constant_vectors): Note that the number of units
25439 is known to be constant.
25441 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25442 Alan Hayward <alan.hayward@arm.com>
25443 David Sherwood <david.sherwood@arm.com>
25445 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
25446 of units as polynomial. Choose between WIDE and NARROW based
25449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25450 Alan Hayward <alan.hayward@arm.com>
25451 David Sherwood <david.sherwood@arm.com>
25453 * tree-vect-stmts.c (simd_clone_subparts): New function.
25454 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
25456 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25457 Alan Hayward <alan.hayward@arm.com>
25458 David Sherwood <david.sherwood@arm.com>
25460 * tree-vect-stmts.c (vectorizable_call): Treat the number of
25461 vectors as polynomial. Use build_index_vector for
25462 IFN_GOMP_SIMD_LANE.
25464 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25465 Alan Hayward <alan.hayward@arm.com>
25466 David Sherwood <david.sherwood@arm.com>
25468 * tree-vect-stmts.c (get_load_store_type): Treat the number of
25469 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
25470 for variable-length vectors.
25471 (vectorizable_mask_load_store): Treat the number of units as
25472 polynomial, asserting that it is constant if the condition has
25473 already been enforced.
25474 (vectorizable_store, vectorizable_load): Likewise.
25476 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25477 Alan Hayward <alan.hayward@arm.com>
25478 David Sherwood <david.sherwood@arm.com>
25480 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
25481 of units as polynomial. Punt if we can't tell at compile time
25482 which vector contains the final result.
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-vect-loop.c (vectorizable_induction): Treat the number
25489 of units as polynomial. Punt on SLP inductions. Use an integer
25490 VEC_SERIES_EXPR for variable-length integer reductions. Use a
25491 cast of such a series for variable-length floating-point
25494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25495 Alan Hayward <alan.hayward@arm.com>
25496 David Sherwood <david.sherwood@arm.com>
25498 * tree.h (build_index_vector): Declare.
25499 * tree.c (build_index_vector): New function.
25500 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
25501 of units as polynomial, forcibly converting it to a constant if
25502 vectorizable_reduction has already enforced the condition.
25503 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
25504 to create a {1,2,3,...} vector.
25505 (vectorizable_reduction): Treat the number of units as polynomial.
25506 Choose vectype_in based on the largest scalar element size rather
25507 than the smallest number of units. Enforce the restrictions
25510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25511 Alan Hayward <alan.hayward@arm.com>
25512 David Sherwood <david.sherwood@arm.com>
25514 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
25515 number of units as polynomial.
25517 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25518 Alan Hayward <alan.hayward@arm.com>
25519 David Sherwood <david.sherwood@arm.com>
25521 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
25522 * target.def (autovectorize_vector_sizes): Return the vector sizes
25523 by pointer, using vector_sizes rather than a bitmask.
25524 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
25525 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
25526 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
25528 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
25529 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
25530 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
25531 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
25532 * omp-general.c (omp_max_vf): Likewise.
25533 * omp-low.c (omp_clause_aligned_alignment): Likewise.
25534 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
25535 * tree-vect-loop.c (vect_analyze_loop): Likewise.
25536 * tree-vect-slp.c (vect_slp_bb): Likewise.
25537 * doc/tm.texi: Regenerate.
25538 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
25540 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
25541 the vector size as a poly_uint64 rather than an unsigned int.
25542 (current_vector_size): Change from an unsigned int to a poly_uint64.
25543 (get_vectype_for_scalar_type): Update accordingly.
25544 * tree.h (build_truth_vector_type): Take the size and number of
25545 units as a poly_uint64 rather than an unsigned int.
25546 (build_vector_type): Add a temporary overload that takes
25547 the number of units as a poly_uint64 rather than an unsigned int.
25548 * tree.c (make_vector_type): Likewise.
25549 (build_truth_vector_type): Take the number of units as a poly_uint64
25550 rather than an unsigned int.
25552 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25553 Alan Hayward <alan.hayward@arm.com>
25554 David Sherwood <david.sherwood@arm.com>
25556 * target.def (get_mask_mode): Take the number of units and length
25557 as poly_uint64s rather than unsigned ints.
25558 * targhooks.h (default_get_mask_mode): Update accordingly.
25559 * targhooks.c (default_get_mask_mode): Likewise.
25560 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
25561 * doc/tm.texi: Regenerate.
25563 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25564 Alan Hayward <alan.hayward@arm.com>
25565 David Sherwood <david.sherwood@arm.com>
25567 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
25568 * omp-general.c (omp_max_vf): Likewise.
25569 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
25570 (expand_omp_simd): Handle polynomial safelen.
25571 * omp-low.c (omplow_simd_context): Add a default constructor.
25572 (omplow_simd_context::max_vf): Change from int to poly_uint64.
25573 (lower_rec_simd_input_clauses): Update accordingly.
25574 (lower_rec_input_clauses): Likewise.
25576 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25577 Alan Hayward <alan.hayward@arm.com>
25578 David Sherwood <david.sherwood@arm.com>
25580 * tree-vectorizer.h (vect_nunits_for_cost): New function.
25581 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
25582 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
25583 (vect_analyze_slp_cost): Likewise.
25584 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
25585 (vect_model_load_cost): Likewise.
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-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
25592 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
25593 from an unsigned int * to a poly_uint64_pod *.
25594 (calculate_unrolling_factor): New function.
25595 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
25597 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25598 Alan Hayward <alan.hayward@arm.com>
25599 David Sherwood <david.sherwood@arm.com>
25601 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
25602 from an unsigned int to a poly_uint64.
25603 (_loop_vec_info::slp_unrolling_factor): Likewise.
25604 (_loop_vec_info::vectorization_factor): Change from an int
25606 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
25607 (vect_get_num_vectors): New function.
25608 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
25609 (vect_get_num_copies): Use vect_get_num_vectors.
25610 (vect_analyze_data_ref_dependences): Change max_vf from an int *
25611 to an unsigned int *.
25612 (vect_analyze_data_refs): Change min_vf from an int * to a
25614 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
25615 than an unsigned HOST_WIDE_INT.
25616 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
25617 (vect_analyze_data_ref_dependence): Change max_vf from an int *
25618 to an unsigned int *.
25619 (vect_analyze_data_ref_dependences): Likewise.
25620 (vect_compute_data_ref_alignment): Handle polynomial vf.
25621 (vect_enhance_data_refs_alignment): Likewise.
25622 (vect_prune_runtime_alias_test_list): Likewise.
25623 (vect_shift_permute_load_chain): Likewise.
25624 (vect_supportable_dr_alignment): Likewise.
25625 (dependence_distance_ge_vf): Take the vectorization factor as a
25626 poly_uint64 rather than an unsigned HOST_WIDE_INT.
25627 (vect_analyze_data_refs): Change min_vf from an int * to a
25629 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
25630 vfm1 as a poly_uint64 rather than an int. Make the same change
25631 for the returned bound_scalar.
25632 (vect_gen_vector_loop_niters): Handle polynomial vf.
25633 (vect_do_peeling): Likewise. Update call to
25634 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
25635 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
25637 * tree-vect-loop.c (vect_determine_vectorization_factor)
25638 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
25639 (vect_get_known_peeling_cost): Likewise.
25640 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
25641 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
25642 (vect_transform_loop): Likewise. Use the lowest possible VF when
25643 updating the upper bounds of the loop.
25644 (vect_min_worthwhile_factor): Make static. Return an unsigned int
25645 rather than an int.
25646 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
25647 polynomial unroll factors.
25648 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
25649 (vect_make_slp_decision): Likewise.
25650 (vect_supported_load_permutation_p): Likewise, and polynomial
25652 (vect_analyze_slp_cost): Handle polynomial vf.
25653 (vect_slp_analyze_node_operations): Likewise.
25654 (vect_slp_analyze_bb_1): Likewise.
25655 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
25656 than an unsigned HOST_WIDE_INT.
25657 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
25658 (vectorizable_load): Handle polynomial vf.
25659 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
25661 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
25663 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25664 Alan Hayward <alan.hayward@arm.com>
25665 David Sherwood <david.sherwood@arm.com>
25667 * match.pd: Handle bit operations involving three constants
25668 and try to fold one pair.
25670 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25672 * tree-vect-loop-manip.c: Include gimple-fold.h.
25673 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
25674 niters_maybe_zero parameters. Handle other cases besides a step of 1.
25675 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
25676 Add a path that uses a step of VF instead of 1, but disable it
25678 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
25679 and niters_no_overflow parameters. Update calls to
25680 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
25681 Create a new SSA name if the latter choses to use a ste other
25682 than zero, and return it via niters_vector_mult_vf_var.
25683 * tree-vect-loop.c (vect_transform_loop): Update calls to
25684 vect_do_peeling, vect_gen_vector_loop_niters and
25685 slpeel_make_loop_iterate_ntimes.
25686 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
25687 (vect_gen_vector_loop_niters): Update declarations after above changes.
25689 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
25691 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
25692 128-bit round to integer instructions.
25693 (ceil<mode>2): Likewise.
25694 (btrunc<mode>2): Likewise.
25695 (round<mode>2): Likewise.
25697 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
25699 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
25700 unaligned VSX load/store on P8/P9.
25701 (expand_block_clear): Allow the use of unaligned VSX
25702 load/store on P8/P9.
25704 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
25706 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
25708 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
25709 swap associated with both a load and a store.
25711 2018-01-02 Andrew Waterman <andrew@sifive.com>
25713 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
25714 * config/riscv/riscv.md (clear_cache): Use it.
25716 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
25718 * web.c: Remove out-of-date comment.
25720 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25722 * expr.c (fixup_args_size_notes): Check that any existing
25723 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
25724 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
25725 (emit_single_push_insn): ...here.
25727 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25729 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
25730 (const_vector_encoded_nelts): New function.
25731 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
25732 (const_vector_int_elt, const_vector_elt): Declare.
25733 * emit-rtl.c (const_vector_int_elt_1): New function.
25734 (const_vector_elt): Likewise.
25735 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
25736 of CONST_VECTOR_ELT.
25738 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25740 * expr.c: Include rtx-vector-builder.h.
25741 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
25742 directly on the tree encoding.
25743 (const_vector_from_tree): Likewise.
25744 * optabs.c: Include rtx-vector-builder.h.
25745 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
25746 sequence of "u" values.
25747 * vec-perm-indices.c: Include rtx-vector-builder.h.
25748 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
25749 directly on the vec_perm_indices encoding.
25751 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25753 * doc/rtl.texi (const_vector): Describe new encoding scheme.
25754 * Makefile.in (OBJS): Add rtx-vector-builder.o.
25755 * rtx-vector-builder.h: New file.
25756 * rtx-vector-builder.c: Likewise.
25757 * rtl.h (rtx_def::u2): Add a const_vector field.
25758 (CONST_VECTOR_NPATTERNS): New macro.
25759 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
25760 (CONST_VECTOR_DUPLICATE_P): Likewise.
25761 (CONST_VECTOR_STEPPED_P): Likewise.
25762 (CONST_VECTOR_ENCODED_ELT): Likewise.
25763 (const_vec_duplicate_p): Check for a duplicated vector encoding.
25764 (unwrap_const_vec_duplicate): Likewise.
25765 (const_vec_series_p): Check for a non-duplicated vector encoding.
25766 Say that the function only returns true for integer vectors.
25767 * emit-rtl.c: Include rtx-vector-builder.h.
25768 (gen_const_vec_duplicate_1): Delete.
25769 (gen_const_vector): Call gen_const_vec_duplicate instead of
25770 gen_const_vec_duplicate_1.
25771 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
25772 (gen_const_vec_duplicate): Use rtx_vector_builder.
25773 (gen_const_vec_series): Likewise.
25774 (gen_rtx_CONST_VECTOR): Likewise.
25775 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
25776 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
25777 Build a new vector rather than modifying a CONST_VECTOR in-place.
25778 (handle_special_swappables): Update call accordingly.
25779 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
25780 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
25781 Build a new vector rather than modifying a CONST_VECTOR in-place.
25782 (handle_special_swappables): Update call accordingly.
25784 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25786 * simplify-rtx.c (simplify_const_binary_operation): Use
25787 CONST_VECTOR_ELT instead of XVECEXP.
25789 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25791 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
25792 the selector elements to be different from the data elements
25793 if the selector is a VECTOR_CST.
25794 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
25795 ssizetype for the selector.
25797 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25799 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
25800 before testing each element individually.
25801 * tree-vect-generic.c (lower_vec_perm): Likewise.
25803 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25805 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
25806 * selftest-run-tests.c (selftest::run_tests): Call it.
25807 * vector-builder.h (vector_builder::operator ==): New function.
25808 (vector_builder::operator !=): Likewise.
25809 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
25810 (vec_perm_indices::all_from_input_p): New function.
25811 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
25812 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
25813 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
25814 instead of reading the VECTOR_CST directly. Detect whether both
25815 vector inputs are the same before constructing the vec_perm_indices,
25816 and update the number of inputs argument accordingly. Use the
25817 utility functions added above. Only construct sel2 if we need to.
25819 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25821 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
25822 the broadcast of the low byte.
25823 (expand_mult_highpart): Use an explicit encoding for the permutes.
25824 * optabs-query.c (can_mult_highpart_p): Likewise.
25825 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
25826 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25827 (vectorizable_bswap): Likewise.
25828 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
25829 explicit encoding for the power-of-2 permutes.
25830 (vect_permute_store_chain): Likewise.
25831 (vect_grouped_load_supported): Likewise.
25832 (vect_permute_load_chain): Likewise.
25834 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25836 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
25837 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
25838 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
25839 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
25840 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
25841 (vect_gen_perm_mask_any): Likewise.
25843 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25845 * int-vector-builder.h: New file.
25846 * vec-perm-indices.h: Include int-vector-builder.h.
25847 (vec_perm_indices): Redefine as an int_vector_builder.
25848 (auto_vec_perm_indices): Delete.
25849 (vec_perm_builder): Redefine as a stand-alone class.
25850 (vec_perm_indices::vec_perm_indices): New function.
25851 (vec_perm_indices::clamp): Likewise.
25852 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
25853 (vec_perm_indices::new_vector): New function.
25854 (vec_perm_indices::new_expanded_vector): Update for new
25855 vec_perm_indices class.
25856 (vec_perm_indices::rotate_inputs): New function.
25857 (vec_perm_indices::all_in_range_p): Operate directly on the
25858 encoded form, without computing elided elements.
25859 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
25860 encoding. Update for new vec_perm_indices class.
25861 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
25862 the given vec_perm_builder.
25863 (expand_vec_perm_var): Update vec_perm_builder constructor.
25864 (expand_mult_highpart): Use vec_perm_builder instead of
25865 auto_vec_perm_indices.
25866 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
25867 vec_perm_indices instead of auto_vec_perm_indices. Use a single
25868 or double series encoding as appropriate.
25869 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
25870 vec_perm_indices instead of auto_vec_perm_indices.
25871 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
25872 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
25873 (vect_permute_store_chain): Likewise.
25874 (vect_grouped_load_supported): Likewise.
25875 (vect_permute_load_chain): Likewise.
25876 (vect_shift_permute_load_chain): Likewise.
25877 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
25878 (vect_transform_slp_perm_load): Likewise.
25879 (vect_schedule_slp_instance): Likewise.
25880 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25881 (vectorizable_mask_load_store): Likewise.
25882 (vectorizable_bswap): Likewise.
25883 (vectorizable_store): Likewise.
25884 (vectorizable_load): Likewise.
25885 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
25886 vec_perm_indices instead of auto_vec_perm_indices. Use
25887 tree_to_vec_perm_builder to read the vector from a tree.
25888 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
25889 vec_perm_builder instead of a vec_perm_indices.
25890 (have_whole_vector_shift): Use vec_perm_builder and
25891 vec_perm_indices instead of auto_vec_perm_indices. Leave the
25892 truncation to calc_vec_perm_mask_for_shift.
25893 (vect_create_epilog_for_reduction): Likewise.
25894 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
25895 from auto_vec_perm_indices to vec_perm_indices.
25896 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
25897 instead of changing individual elements.
25898 (aarch64_vectorize_vec_perm_const): Use new_vector to install
25899 the vector in d.perm.
25900 * config/arm/arm.c (expand_vec_perm_d::perm): Change
25901 from auto_vec_perm_indices to vec_perm_indices.
25902 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
25903 instead of changing individual elements.
25904 (arm_vectorize_vec_perm_const): Use new_vector to install
25905 the vector in d.perm.
25906 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
25907 Update vec_perm_builder constructor.
25908 (rs6000_expand_interleave): Likewise.
25909 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
25910 (rs6000_expand_interleave): Likewise.
25912 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25914 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
25915 to qimode could truncate the indices.
25916 * optabs.c (expand_vec_perm_var): Likewise.
25918 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25920 * Makefile.in (OBJS): Add vec-perm-indices.o.
25921 * vec-perm-indices.h: New file.
25922 * vec-perm-indices.c: Likewise.
25923 * target.h (vec_perm_indices): Replace with a forward class
25925 (auto_vec_perm_indices): Move to vec-perm-indices.h.
25926 * optabs.h: Include vec-perm-indices.h.
25927 (expand_vec_perm): Delete.
25928 (selector_fits_mode_p, expand_vec_perm_var): Declare.
25929 (expand_vec_perm_const): Declare.
25930 * target.def (vec_perm_const_ok): Replace with...
25931 (vec_perm_const): ...this new hook.
25932 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
25933 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
25934 * doc/tm.texi: Regenerate.
25935 * optabs.def (vec_perm_const): Delete.
25936 * doc/md.texi (vec_perm_const): Likewise.
25937 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
25938 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
25939 expand_vec_perm for constant permutation vectors. Assert that
25940 the mode of variable permutation vectors is the integer equivalent
25941 of the mode that is being permuted.
25942 * optabs-query.h (selector_fits_mode_p): Declare.
25943 * optabs-query.c: Include vec-perm-indices.h.
25944 (selector_fits_mode_p): New function.
25945 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
25946 is defined, instead of checking whether the vec_perm_const_optab
25947 exists. Use targetm.vectorize.vec_perm_const instead of
25948 targetm.vectorize.vec_perm_const_ok. Check whether the indices
25949 fit in the vector mode before using a variable permute.
25950 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
25951 vec_perm_indices instead of an rtx.
25952 (expand_vec_perm): Replace with...
25953 (expand_vec_perm_const): ...this new function. Take the selector
25954 as a vec_perm_indices rather than an rtx. Also take the mode of
25955 the selector. Update call to shift_amt_for_vec_perm_mask.
25956 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
25957 Use vec_perm_indices::new_expanded_vector to expand the original
25958 selector into bytes. Check whether the indices fit in the vector
25959 mode before using a variable permute.
25960 (expand_vec_perm_var): Make global.
25961 (expand_mult_highpart): Use expand_vec_perm_const.
25962 * fold-const.c: Includes vec-perm-indices.h.
25963 * tree-ssa-forwprop.c: Likewise.
25964 * tree-vect-data-refs.c: Likewise.
25965 * tree-vect-generic.c: Likewise.
25966 * tree-vect-loop.c: Likewise.
25967 * tree-vect-slp.c: Likewise.
25968 * tree-vect-stmts.c: Likewise.
25969 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
25971 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
25972 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
25973 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
25974 (aarch64_vectorize_vec_perm_const): ...this new function.
25975 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25976 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25977 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
25978 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
25979 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25980 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25981 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
25983 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
25984 check for NEON modes.
25985 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
25986 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
25987 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
25988 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
25990 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
25991 the old VEC_PERM_CONST conditions.
25992 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
25993 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
25994 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
25995 (ia64_vectorize_vec_perm_const_ok): Merge into...
25996 (ia64_vectorize_vec_perm_const): ...this new function.
25997 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
25998 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
25999 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
26000 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
26001 * config/mips/mips.c (mips_expand_vec_perm_const)
26002 (mips_vectorize_vec_perm_const_ok): Merge into...
26003 (mips_vectorize_vec_perm_const): ...this new function.
26004 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
26005 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
26006 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
26007 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
26008 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
26009 (rs6000_expand_vec_perm_const): Delete.
26010 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
26012 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26013 (altivec_expand_vec_perm_const_le): Take each operand individually.
26014 Operate on constant selectors rather than rtxes.
26015 (altivec_expand_vec_perm_const): Likewise. Update call to
26016 altivec_expand_vec_perm_const_le.
26017 (rs6000_expand_vec_perm_const): Delete.
26018 (rs6000_vectorize_vec_perm_const_ok): Delete.
26019 (rs6000_vectorize_vec_perm_const): New function.
26020 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
26021 an element count and rtx array.
26022 (rs6000_expand_extract_even): Update call accordingly.
26023 (rs6000_expand_interleave): Likewise.
26024 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
26025 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
26026 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
26027 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
26028 (rs6000_expand_vec_perm_const): Delete.
26029 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
26030 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26031 (altivec_expand_vec_perm_const_le): Take each operand individually.
26032 Operate on constant selectors rather than rtxes.
26033 (altivec_expand_vec_perm_const): Likewise. Update call to
26034 altivec_expand_vec_perm_const_le.
26035 (rs6000_expand_vec_perm_const): Delete.
26036 (rs6000_vectorize_vec_perm_const_ok): Delete.
26037 (rs6000_vectorize_vec_perm_const): New function. Remove stray
26038 reference to the SPE evmerge intructions.
26039 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
26040 an element count and rtx array.
26041 (rs6000_expand_extract_even): Update call accordingly.
26042 (rs6000_expand_interleave): Likewise.
26043 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
26044 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
26046 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26048 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26050 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
26051 vector mode and that that mode matches the mode of the data
26053 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
26054 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
26055 directly using expand_vec_perm_1 when forcing selectors into
26057 (expand_vec_perm_var): New function, split out from expand_vec_perm.
26059 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26061 * optabs-query.h (can_vec_perm_p): Delete.
26062 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
26063 * optabs-query.c (can_vec_perm_p): Split into...
26064 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
26065 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
26066 particular selector is valid.
26067 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26068 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
26069 (vect_grouped_load_supported): Likewise.
26070 (vect_shift_permute_load_chain): Likewise.
26071 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
26072 (vect_transform_slp_perm_load): Likewise.
26073 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26074 (vectorizable_bswap): Likewise.
26075 (vect_gen_perm_mask_checked): Likewise.
26076 * fold-const.c (fold_ternary_loc): Likewise. Don't take
26077 implementations of variable permutation vectors into account
26078 when deciding which selector to use.
26079 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
26080 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
26081 with a false third argument.
26082 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
26083 to test whether the constant selector is valid and can_vec_perm_var_p
26084 to test whether a variable selector is valid.
26086 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26088 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
26089 * optabs-query.c (can_vec_perm_p): Likewise.
26090 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
26091 instead of vec_perm_indices.
26092 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
26093 (vect_gen_perm_mask_checked): Likewise,
26094 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
26095 (vect_gen_perm_mask_checked): Likewise,
26097 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26099 * optabs-query.h (qimode_for_vec_perm): Declare.
26100 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
26101 (qimode_for_vec_perm): ...this new function.
26102 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
26104 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
26106 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
26107 does not have a conditional at the top.
26109 2018-01-02 Richard Biener <rguenther@suse.de>
26111 * ipa-inline.c (big_speedup_p): Fix expression.
26113 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
26116 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
26119 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
26123 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
26124 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
26125 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
26126 cond_taken_branch_cost 3->4.
26128 2018-01-01 Jakub Jelinek <jakub@redhat.com>
26130 PR tree-optimization/83581
26131 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
26132 TODO_cleanup_cfg if any changes have been made.
26134 PR middle-end/83608
26135 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
26136 convert_modes if target mode has the right side, but different mode
26139 PR middle-end/83609
26140 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
26141 last argument when extracting from CONCAT. If either from_real or
26142 from_imag is NULL, use expansion through memory. If result is not
26143 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
26144 the parts directly to inner mode, if even that fails, use expansion
26147 PR middle-end/83623
26148 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
26149 check for bswap in mode rather than HImode and use that in expand_unop
26152 Copyright (C) 2018 Free Software Foundation, Inc.
26154 Copying and distribution of this file, with or without modification,
26155 are permitted in any medium without royalty provided the copyright
26156 notice and this notice are preserved.